How Finite Capacity Scheduling Actually Works in Manufacturing

1
00:00:00,000 --> 00:00:03,400
Your ERP releases production orders and assigns a planned finish date,

2
00:00:03,400 --> 00:00:04,960
and it all looks fine on paper.

3
00:00:04,960 --> 00:00:08,600
But then you walk the shop floor and find the machine that's supposed to make those parts

4
00:00:08,600 --> 00:00:12,480
is already buried under more work than it can physically finish in the time available.

5
00:00:12,480 --> 00:00:13,760
Now here's the real challenge.

6
00:00:13,760 --> 00:00:15,200
That gap shows up everywhere.

7
00:00:15,200 --> 00:00:18,960
The system has an order date, but it doesn't track time on the machine itself,

8
00:00:18,960 --> 00:00:22,080
and that machine might need a specific operator who's already booked,

9
00:00:22,080 --> 00:00:24,520
or a fixture tied up on another job.

10
00:00:24,520 --> 00:00:26,840
Material might exist as a stock number in ERP,

11
00:00:26,840 --> 00:00:30,240
but not as inspected released material sitting at the machine ready to run.

12
00:00:30,240 --> 00:00:33,640
So imagine one loaded machine loses four hours during the day.

13
00:00:33,640 --> 00:00:38,080
By itself, that's not a disaster, but it forces a real question, which order moves first.

14
00:00:38,080 --> 00:00:40,840
The one due tomorrow, the repair part for a customer waiting,

15
00:00:40,840 --> 00:00:42,520
the job that shares the current setup,

16
00:00:42,520 --> 00:00:46,560
or the one feeding an assembly cell where people are already booked for the afternoon.

17
00:00:46,560 --> 00:00:49,000
Most factories answer that question with a mix of experience,

18
00:00:49,000 --> 00:00:53,120
phone calls, whiteboards, and often a spreadsheet that survived several software projects.

19
00:00:53,120 --> 00:00:54,480
That's not a criticism.

20
00:00:54,480 --> 00:00:58,480
It's what people do when the plan doesn't carry enough of the factory's real limits.

21
00:00:58,480 --> 00:01:01,920
Finite capacity scheduling puts the plan through a tougher test.

22
00:01:01,920 --> 00:01:05,120
It asks whether work can actually fit into the available time

23
00:01:05,120 --> 00:01:08,960
with the people, machines, tools, material, and rules needed to run it.

24
00:01:08,960 --> 00:01:11,080
It doesn't promise every order will meet its date,

25
00:01:11,080 --> 00:01:14,000
but it tells you sooner when that promise can't hold.

26
00:01:14,000 --> 00:01:16,320
And that moves us past the late order symptom

27
00:01:16,320 --> 00:01:20,600
to look at the planning assumption that created the impossible plan in the first place.

28
00:01:20,600 --> 00:01:22,280
Infinite capacity planning.

29
00:01:22,280 --> 00:01:24,440
Let's cut through the hype for a second.

30
00:01:24,440 --> 00:01:29,200
A lot of production planning starts with an assumption that sounds odd when you say it out loud.

31
00:01:29,200 --> 00:01:32,840
Capacity can stretch as far as demand needs it to go.

32
00:01:32,840 --> 00:01:35,600
That approach is called infinite capacity planning,

33
00:01:35,600 --> 00:01:38,160
where the system places demand against dates,

34
00:01:38,160 --> 00:01:40,960
calculates requirements, and creates planned work

35
00:01:40,960 --> 00:01:44,600
without forcing every operation to fit within the real limits of every machine,

36
00:01:44,600 --> 00:01:47,000
person, tool, and shift.

37
00:01:47,000 --> 00:01:48,240
This isn't bad planning.

38
00:01:48,240 --> 00:01:49,560
It has a specific job.

39
00:01:49,560 --> 00:01:53,600
Your ERP system handles the commercial and transactional side of manufacturing,

40
00:01:53,600 --> 00:01:57,960
sales orders, work orders, bills of material, inventory, purchasing,

41
00:01:57,960 --> 00:02:00,520
supplier dates, and financial data.

42
00:02:00,520 --> 00:02:03,280
Material requirements planning, or MRP,

43
00:02:03,280 --> 00:02:06,800
uses that info to figure out what materials and components you need,

44
00:02:06,800 --> 00:02:08,280
and when you might need them.

45
00:02:08,280 --> 00:02:12,480
For that kind of planning, speed matters because you might need to assess demand across months,

46
00:02:12,480 --> 00:02:15,880
across plans, across suppliers, and across thousands of part numbers.

47
00:02:15,880 --> 00:02:18,720
An ERP system can build a broad plan quickly,

48
00:02:18,720 --> 00:02:22,000
and it can expose material shortages before they become surprises.

49
00:02:22,000 --> 00:02:25,200
But a transaction plan and a runnable shop flow schedule are two different things.

50
00:02:25,200 --> 00:02:27,760
A transaction plan can say a work order should finish next Thursday

51
00:02:27,760 --> 00:02:31,240
because the order, routing, and due date support that result in the system,

52
00:02:31,240 --> 00:02:34,160
while a runnable schedule must answer a more physical question.

53
00:02:34,160 --> 00:02:36,120
Which resource will run the first operation?

54
00:02:36,120 --> 00:02:38,160
At what time? After which prior job?

55
00:02:38,160 --> 00:02:41,480
With which operator? And with what material and tooling ready?

56
00:02:41,480 --> 00:02:44,560
Those aren't small details, but that's the actual work.

57
00:02:44,560 --> 00:02:47,240
Think about a simple case where your ERP sees demand

58
00:02:47,240 --> 00:02:52,760
for 40 hours of machining next week, but the work center calendar only shows 20 working hours.

59
00:02:52,760 --> 00:02:55,600
If the planning setup treats the resource as infinite,

60
00:02:55,600 --> 00:03:00,320
the system can put all 40 hours into that week without breaking a single rule.

61
00:03:00,320 --> 00:03:03,760
The dates might look tidy, but the load doesn't fit.

62
00:03:03,760 --> 00:03:06,560
That isn't a software bug. It's a design assumption.

63
00:03:06,560 --> 00:03:10,120
Infinite planning sticks around because it keeps the planning model manageable.

64
00:03:10,120 --> 00:03:13,640
It can work with rough capacities rather than every shift detail,

65
00:03:13,640 --> 00:03:18,960
and it can plan long horizons without trying to predict every machine stop or operator absence.

66
00:03:18,960 --> 00:03:23,720
And in many plans, the master data needed for a strict finite schedule simply isn't ready yet.

67
00:03:23,720 --> 00:03:26,320
Maybe the routing says an operation takes 30 minutes,

68
00:03:26,320 --> 00:03:28,720
but production usually needs closer to 50.

69
00:03:28,720 --> 00:03:31,480
And maybe the work center calendar assumes three shifts,

70
00:03:31,480 --> 00:03:33,600
but the plant currently runs two.

71
00:03:33,600 --> 00:03:35,720
And then there's the alternate machine that exists on paper,

72
00:03:35,720 --> 00:03:37,440
but hasn't been approved for that part.

73
00:03:37,440 --> 00:03:40,920
If you force a detailed schedule onto week data, you don't get control.

74
00:03:40,920 --> 00:03:42,880
You get a very precise version of a guess.

75
00:03:42,880 --> 00:03:47,280
So there's a practical reason to use broad, infinite planning at the ERP level.

76
00:03:47,280 --> 00:03:50,240
It helps you see demand by material and create work.

77
00:03:50,240 --> 00:03:53,200
But the problem starts when people treat that broad plan as proof

78
00:03:53,200 --> 00:03:55,000
that the shop floor can deliver those dates.

79
00:03:55,000 --> 00:03:58,760
A date from MRP usually means this is when the system expects the work.

80
00:03:58,760 --> 00:04:02,560
Not a finite set of resources can complete the work by then.

81
00:04:02,560 --> 00:04:04,960
That distinction changes how you read overload.

82
00:04:04,960 --> 00:04:08,560
An overloaded work center isn't necessarily a failure of the planner or the ERP.

83
00:04:08,560 --> 00:04:12,080
It's information that demand exceeds the capacity currently modeled for that period.

84
00:04:12,080 --> 00:04:16,360
The plant now needs a decision, move demand, add capacity, change the route,

85
00:04:16,360 --> 00:04:19,160
alter the promise, or accept the late delivery risk.

86
00:04:19,160 --> 00:04:24,240
Without finite scheduling, that decision stays hidden until the queue builds in front of the machine.

87
00:04:24,240 --> 00:04:29,840
And the supervisor sees the conflict first even though it existed in the plan days or weeks earlier.

88
00:04:29,840 --> 00:04:33,400
Let's follow one production order now as it leaves ERP with a due date and a routing,

89
00:04:33,400 --> 00:04:36,960
then reaches the machine that must turn that plan into a physical part.

90
00:04:36,960 --> 00:04:38,640
A factory scenario.

91
00:04:38,640 --> 00:04:41,640
With one bottleneck, here's a scenario I see all the time.

92
00:04:41,640 --> 00:04:44,720
Picture a plant that makes machined components and assembled units.

93
00:04:44,720 --> 00:04:48,400
Two shifts and most workflows through a shared CNC machining center

94
00:04:48,400 --> 00:04:50,240
before hitting assembly cells downstream.

95
00:04:50,240 --> 00:04:52,520
That machine is the bottleneck we're about to dig into.

96
00:04:52,520 --> 00:04:55,640
Now here's the thing, this plant doesn't just produce one part type.

97
00:04:55,640 --> 00:04:58,000
You've got orders needing a short milling operation,

98
00:04:58,000 --> 00:04:59,680
others needing drilling or turning,

99
00:04:59,680 --> 00:05:03,480
and some requiring a longer machining sequence before assembly can even start.

100
00:05:03,480 --> 00:05:05,240
Every order has its own due date,

101
00:05:05,240 --> 00:05:08,320
and your balancing standard customer orders, urgent service parts,

102
00:05:08,320 --> 00:05:13,520
and internal assembly orders that already have people, material, and test capacity lined up.

103
00:05:13,520 --> 00:05:16,320
From the planner's view, everything looks released.

104
00:05:16,320 --> 00:05:20,560
The ERP has the orders, routing steps, quantities, and dates.

105
00:05:20,560 --> 00:05:24,800
Material planning triggered purchase needs and production has a list of work waiting to run.

106
00:05:24,800 --> 00:05:28,040
But walk to the machining area and you see a completely different picture.

107
00:05:28,040 --> 00:05:31,240
A real queue sits physically in front of that CNC machine,

108
00:05:31,240 --> 00:05:36,680
carts, pallets, paperwork, material containers, and operators asking, which job is next.

109
00:05:36,680 --> 00:05:38,000
Here's where it gets interesting.

110
00:05:38,000 --> 00:05:40,720
Some of those orders share a setup family.

111
00:05:40,720 --> 00:05:44,600
Same material grade, similar tool package, same fixture arrangement.

112
00:05:44,600 --> 00:05:46,880
Grouping them together saves change over time,

113
00:05:46,880 --> 00:05:50,480
which sounds sensible until one of those later orders has a tighter customer date.

114
00:05:50,480 --> 00:05:53,400
Other orders could use an alternate machine at least in theory,

115
00:05:53,400 --> 00:05:57,640
but that alternate machine runs a different product family only one operator knows the setup,

116
00:05:57,640 --> 00:05:59,960
or the fixture is already committed elsewhere.

117
00:05:59,960 --> 00:06:03,120
So the CNC machining center is where all those planning assumptions

118
00:06:03,120 --> 00:06:06,640
hit physical reality, upstream material might arrive on time,

119
00:06:06,640 --> 00:06:09,080
and downstream assembly might have open capacity,

120
00:06:09,080 --> 00:06:12,800
but none of that helps if the machined component can't leave the bottleneck.

121
00:06:12,800 --> 00:06:14,840
Let me be clear about what a bottleneck actually is.

122
00:06:14,840 --> 00:06:17,640
It's the resource that limits the pace of the whole flow,

123
00:06:17,640 --> 00:06:19,720
doesn't have to be the most expensive machine,

124
00:06:19,720 --> 00:06:22,120
and it doesn't need to run at 100% utilization.

125
00:06:22,120 --> 00:06:26,200
It earns that label when demand for its time consistently exceeds,

126
00:06:26,200 --> 00:06:29,320
or nearly exceeds, the usable time you can get from it.

127
00:06:29,320 --> 00:06:32,440
That machine dictates the cadence for everything downstream,

128
00:06:32,440 --> 00:06:35,320
when it completes fewer parts than planned assembly weights.

129
00:06:35,320 --> 00:06:39,800
When it runs the wrong sequence, an urgent order sits behind lower priority work.

130
00:06:39,800 --> 00:06:42,880
When it spends too long on changeovers between unrelated jobs,

131
00:06:42,880 --> 00:06:45,400
that lost time ripples through the whole plant.

132
00:06:45,400 --> 00:06:49,600
Late starts, idle people, missed inspections, changed delivery dates.

133
00:06:49,600 --> 00:06:52,680
Here's why a planner and a supervisor can both look at the same orders

134
00:06:52,680 --> 00:06:54,800
and reach completely different conclusions.

135
00:06:54,800 --> 00:06:57,400
The planner sees released orders with due dates,

136
00:06:57,400 --> 00:07:01,920
the supervisor sees the machine queue and knows the next decision changes the entire shift.

137
00:07:01,920 --> 00:07:04,240
In practical terms, here's what that looks like.

138
00:07:04,240 --> 00:07:08,000
At the start of the morning shift, the machining center has three orders waiting.

139
00:07:08,000 --> 00:07:11,000
Order A supports an assembly cell later that day.

140
00:07:11,000 --> 00:07:13,200
Order B has the earliest customer due date,

141
00:07:13,200 --> 00:07:16,560
order C shares the same setup family as the job that just finished,

142
00:07:16,560 --> 00:07:19,120
sought only by due date and order B runs first.

143
00:07:19,120 --> 00:07:20,680
That protects one customer promise,

144
00:07:20,680 --> 00:07:22,240
but it requires a full changeover,

145
00:07:22,240 --> 00:07:25,200
and pushes order A past the point where assembly can use it.

146
00:07:25,200 --> 00:07:27,800
Keep the existing setup and run order C first,

147
00:07:27,800 --> 00:07:30,160
and the machining center produces more efficiently,

148
00:07:30,160 --> 00:07:32,640
but the earliest due order stays in the queue,

149
00:07:32,640 --> 00:07:35,000
and your customer risk grows.

150
00:07:35,000 --> 00:07:36,480
Run order A first,

151
00:07:36,480 --> 00:07:38,120
and you protect the assembly cell,

152
00:07:38,120 --> 00:07:40,520
but delay both external customer orders.

153
00:07:40,520 --> 00:07:43,040
None of these choices comes from a lack of effort.

154
00:07:43,040 --> 00:07:47,280
The plant has a real conflict and someone needs to decide which consequence to accept.

155
00:07:47,280 --> 00:07:49,560
This is where a simple order list stops helping.

156
00:07:49,560 --> 00:07:51,440
Sorting by due date gives you a rule,

157
00:07:51,440 --> 00:07:54,800
but it doesn't tell you whether the order can run on the required machine,

158
00:07:54,800 --> 00:07:57,520
whether the current setup changes available time,

159
00:07:57,520 --> 00:07:59,280
whether an operator has the skill,

160
00:07:59,280 --> 00:08:02,920
or what downstream work will wait if you pick one job over another.

161
00:08:02,920 --> 00:08:05,840
The queue has turned planning into a scheduling problem.

162
00:08:05,840 --> 00:08:07,600
What finite actually means?

163
00:08:07,600 --> 00:08:11,360
Let's cut through the jargon and talk about what finite actually means.

164
00:08:11,360 --> 00:08:14,200
Finite capacity scheduling starts with a simple rule,

165
00:08:14,200 --> 00:08:17,640
only put work where the time and resources actually exist to do it.

166
00:08:17,640 --> 00:08:21,080
That sounds obvious, but on a shop floor it changes the whole conversation.

167
00:08:21,080 --> 00:08:23,320
Instead of asking which orders should finish this week,

168
00:08:23,320 --> 00:08:24,080
you ask,

169
00:08:24,080 --> 00:08:25,680
given the real limits of this plant,

170
00:08:25,680 --> 00:08:28,040
which operations can run this week in which sequence,

171
00:08:28,040 --> 00:08:30,480
and what will that choice push out?

172
00:08:30,480 --> 00:08:33,760
A finite schedule treats capacity as a limit, not a wish.

173
00:08:33,760 --> 00:08:35,200
Here's a common mistake, I see.

174
00:08:35,200 --> 00:08:36,680
Someone looks at the machine calendar,

175
00:08:36,680 --> 00:08:38,360
sees two shifts Monday through Friday,

176
00:08:38,360 --> 00:08:40,720
and calls that 80 available machine hours.

177
00:08:40,720 --> 00:08:43,600
But nobody actually produces for 80 clean hours,

178
00:08:43,600 --> 00:08:44,960
just because the building is open.

179
00:08:44,960 --> 00:08:47,440
You've got breaks, plan maintenance, handover, setup,

180
00:08:47,440 --> 00:08:48,640
first piece checks, cleaning,

181
00:08:48,640 --> 00:08:51,000
and all the small stops that happen in normal production.

182
00:08:51,000 --> 00:08:53,440
Some machines run unattended for part of a shift,

183
00:08:53,440 --> 00:08:55,640
others need a trained person present.

184
00:08:55,640 --> 00:08:58,880
The capacity that matters is usable time, not calendar time.

185
00:08:58,880 --> 00:09:01,400
Usable time depends on more than just the machine.

186
00:09:01,400 --> 00:09:03,880
A work order might need a specific CNC machine

187
00:09:03,880 --> 00:09:06,080
for its size, tolerance, or program,

188
00:09:06,080 --> 00:09:08,840
but it also needs an operator with the right approval,

189
00:09:08,840 --> 00:09:10,600
a fixture that fits the part,

190
00:09:10,600 --> 00:09:14,440
cutting tools with enough remaining life, material from an approved lot,

191
00:09:14,440 --> 00:09:17,320
and an inspection step before the next operation.

192
00:09:17,320 --> 00:09:20,600
If any one of those things is missing, the work can't run.

193
00:09:20,600 --> 00:09:22,120
An open hour on the machine doesn't help.

194
00:09:22,120 --> 00:09:25,320
This is where people often misunderstand finite capacity scheduling.

195
00:09:25,320 --> 00:09:29,000
They hear finite, and assume the schedule becomes rigid.

196
00:09:29,000 --> 00:09:31,520
A timetable nobody can touch once it's published.

197
00:09:31,520 --> 00:09:32,840
But it doesn't work like that.

198
00:09:32,840 --> 00:09:37,120
A finite schedule is a current model of what can run under the constraints, you know, right now.

199
00:09:37,120 --> 00:09:39,760
When a constraint changes, the schedule should respond.

200
00:09:39,760 --> 00:09:43,200
A machine stops, material arrives early, a quality hold clears,

201
00:09:43,200 --> 00:09:46,560
an operator calls in sick, or maintenance needs to bring forward a repair.

202
00:09:46,560 --> 00:09:48,840
The plan changes because the factory changed.

203
00:09:48,840 --> 00:09:52,640
That doesn't mean you should rebuild every schedule from scratch for a minor change.

204
00:09:52,640 --> 00:09:56,680
A schedule that changes constantly becomes hard to execute, and people stop trusting it.

205
00:09:56,680 --> 00:09:59,880
But the system needs a way to show the effect of a real change,

206
00:09:59,880 --> 00:10:04,120
rather than leaving an old plan in place and hoping the shift absorbs the difference.

207
00:10:04,120 --> 00:10:08,600
There's also a critical distinction between a feasible schedule and a commercial priority.

208
00:10:08,600 --> 00:10:11,080
A feasible schedule answers a physical question.

209
00:10:11,080 --> 00:10:13,840
Can this operation run on this resource at this time

210
00:10:13,840 --> 00:10:17,320
with the required material tool, person and process conditions?

211
00:10:17,320 --> 00:10:19,680
Commercial priority answers a different question.

212
00:10:19,680 --> 00:10:23,840
Which order deserves preference when more work wants the same constraint capacity?

213
00:10:23,840 --> 00:10:25,160
You need both.

214
00:10:25,160 --> 00:10:28,400
Sales has a customer commitment that puts one order ahead of another.

215
00:10:28,400 --> 00:10:30,600
Service needs a repair part immediately.

216
00:10:30,600 --> 00:10:32,560
Production needs to protect an assembly run

217
00:10:32,560 --> 00:10:36,120
because stopping that sell waste far more capacity than one additional setup.

218
00:10:36,120 --> 00:10:40,320
Those priorities belong in the decision, but they don't erase physical limits.

219
00:10:40,320 --> 00:10:42,800
If the urgent order needs six hours on a machine,

220
00:10:42,800 --> 00:10:44,880
with only two usable hours left today,

221
00:10:44,880 --> 00:10:46,720
the schedule shouldn't pretend otherwise.

222
00:10:46,720 --> 00:10:48,040
It should show the consequence.

223
00:10:48,040 --> 00:10:50,800
Another order moves, over time becomes an option,

224
00:10:50,800 --> 00:10:52,720
or the customer date needs a new conversation.

225
00:10:52,720 --> 00:10:54,920
That is a much more useful form of bad news.

226
00:10:54,920 --> 00:10:57,560
Finite capacity scheduling doesn't create capacity.

227
00:10:57,560 --> 00:11:01,320
It makes the competition for capacity visible while you still have time to choose.

228
00:11:01,320 --> 00:11:03,800
A commercial promise can shape the sequence,

229
00:11:03,800 --> 00:11:07,000
but it can't turn an unavailable fixture into an available one,

230
00:11:07,000 --> 00:11:10,400
or put a qualified operator on a shift where none exists.

231
00:11:10,400 --> 00:11:14,040
So before a scheduler can place even one operation into a real time slot,

232
00:11:14,040 --> 00:11:16,880
it needs to know far more than an order number and a due date.

233
00:11:16,880 --> 00:11:20,560
It needs the data that describes the product, the work, the resources,

234
00:11:20,560 --> 00:11:23,160
and the conditions for that work to actually run.

235
00:11:23,160 --> 00:11:24,640
The scheduling data model.

236
00:11:24,640 --> 00:11:26,520
Here's what a finite schedule actually needs.

237
00:11:26,520 --> 00:11:29,640
A data model that describes work in enough detail to place it in time.

238
00:11:29,640 --> 00:11:31,320
You don't need perfect data whistles.

239
00:11:31,320 --> 00:11:32,760
I've never seen a plant that has it,

240
00:11:32,760 --> 00:11:36,520
but the model has to be honest enough that the scheduler doesn't invent capacity,

241
00:11:36,520 --> 00:11:40,120
routing options, or readiness that production can't support.

242
00:11:40,120 --> 00:11:42,720
Start with the product and its root through the plant.

243
00:11:42,720 --> 00:11:44,320
For each part or product family,

244
00:11:44,320 --> 00:11:47,240
the model needs every operation required to produce it,

245
00:11:47,240 --> 00:11:49,960
and the sequence matters because the scheduler has to understand

246
00:11:49,960 --> 00:11:51,960
what happens first and what waits.

247
00:11:51,960 --> 00:11:53,920
That root also needs planning times,

248
00:11:53,920 --> 00:11:56,800
including run time depending on quantity and cycle time,

249
00:11:56,800 --> 00:11:58,640
setup time before production,

250
00:11:58,640 --> 00:12:02,080
and often batch rules that decide whether everything moves together

251
00:12:02,080 --> 00:12:05,720
or partial quantities transfer as soon as they pass inspection.

252
00:12:05,720 --> 00:12:07,840
Those differences change the dates completely.

253
00:12:07,840 --> 00:12:10,240
A 10 hour order doesn't always block out 10 straight hours

254
00:12:10,240 --> 00:12:12,080
because it might use a setup, run in batches,

255
00:12:12,080 --> 00:12:15,440
wait for approval, and release smaller quantities to the next step.

256
00:12:15,440 --> 00:12:18,480
If the root only gives you a broad duration and a work center name,

257
00:12:18,480 --> 00:12:20,040
the scheduler can produce a date,

258
00:12:20,040 --> 00:12:21,960
but that date doesn't carry much confidence.

259
00:12:21,960 --> 00:12:23,840
Now let's talk about the resource model.

260
00:12:23,840 --> 00:12:25,520
Resources might be specific machines,

261
00:12:25,520 --> 00:12:27,160
groups of similar machines, work centers,

262
00:12:27,160 --> 00:12:28,920
or shared assets that limit throughput,

263
00:12:28,920 --> 00:12:31,800
and the scheduler needs calendars for each, shifts, breaks,

264
00:12:31,800 --> 00:12:33,320
shutdowns, maintenance,

265
00:12:33,320 --> 00:12:36,600
along with an understanding of what each resource can actually do.

266
00:12:36,600 --> 00:12:38,720
A machine name alone tells you very little,

267
00:12:38,720 --> 00:12:41,760
considered two CNC machines in the same ERP work center

268
00:12:41,760 --> 00:12:43,680
that might look identical on paper,

269
00:12:43,680 --> 00:12:45,280
but one has the spindle range,

270
00:12:45,280 --> 00:12:47,160
probing setup, program version,

271
00:12:47,160 --> 00:12:51,240
and fixture interface for a specific part while the other doesn't.

272
00:12:51,240 --> 00:12:54,640
An alternate machine only helps if it's a valid option under current rules

273
00:12:54,640 --> 00:12:56,040
and tooling matters too.

274
00:12:56,040 --> 00:12:58,280
A machine without the right fixture or tool package

275
00:12:58,280 --> 00:13:00,400
isn't available for that job.

276
00:13:00,400 --> 00:13:01,920
Then you need the order itself,

277
00:13:01,920 --> 00:13:05,200
which tells the scheduler how much work it must place when it can start

278
00:13:05,200 --> 00:13:06,640
and when it needs to finish.

279
00:13:06,640 --> 00:13:09,880
Quantity drives runtime, the release date may block early starts

280
00:13:09,880 --> 00:13:11,760
and the due date gives a target.

281
00:13:11,760 --> 00:13:16,360
Priority and customer commitments explain why one order might jump ahead of another.

282
00:13:16,360 --> 00:13:18,800
Material status needs to sit close to that order data

283
00:13:18,800 --> 00:13:20,560
because there's a big difference between material

284
00:13:20,560 --> 00:13:24,120
that purchasing expects next week and material you can issue right now

285
00:13:24,120 --> 00:13:26,720
and the schedule should know whether the order is waiting for receipt,

286
00:13:26,720 --> 00:13:28,840
inspection, kitting, or release.

287
00:13:28,840 --> 00:13:32,600
Otherwise, it might reserve a machine slot for something that can't even start

288
00:13:32,600 --> 00:13:34,640
while another ready order sits in the queue.

289
00:13:34,640 --> 00:13:37,920
The shop floor then turns this model from a plan into a current view

290
00:13:37,920 --> 00:13:42,560
where actual start and finish times tell the scheduler what capacity was really consumed.

291
00:13:42,560 --> 00:13:45,280
Down time reports when a resource became unavailable

292
00:13:45,280 --> 00:13:48,600
and why if possible, scrap changes the remaining quantity,

293
00:13:48,600 --> 00:13:50,280
queue state shows what's waiting

294
00:13:50,280 --> 00:13:53,040
and operator availability adds another layer

295
00:13:53,040 --> 00:13:56,480
when the process depends on someone with a specific skill or approval.

296
00:13:56,480 --> 00:13:59,960
This data doesn't need to come through one giant real-time system.

297
00:13:59,960 --> 00:14:02,560
In some plans, a shift level update is enough,

298
00:14:02,560 --> 00:14:07,040
while in others, a machine stop or completed operation needs faster handling

299
00:14:07,040 --> 00:14:09,200
because the next decision depends on it.

300
00:14:09,200 --> 00:14:11,280
The right timing follows the production problem,

301
00:14:11,280 --> 00:14:14,880
not some generic rule about how advanced the architecture should sound.

302
00:14:14,880 --> 00:14:17,920
Ownership matters just as much as the fields themselves.

303
00:14:17,920 --> 00:14:21,680
ERP typically owns demand, work orders, plant dates,

304
00:14:21,680 --> 00:14:24,560
purchasing signals and most of the material record.

305
00:14:24,560 --> 00:14:27,440
MES owns execution status, labor reporting,

306
00:14:27,440 --> 00:14:29,440
actual quantities and traceability.

307
00:14:29,440 --> 00:14:32,280
Maintenance handles plant downtime and asset condition,

308
00:14:32,280 --> 00:14:34,920
quality controls inspection status and release rules,

309
00:14:34,920 --> 00:14:38,360
and engineering owns routes, standards, process changes

310
00:14:38,360 --> 00:14:42,760
and the technical conditions that decide whether an alternate resource can run apart.

311
00:14:42,760 --> 00:14:45,880
Nobody should assume one team can keep all of this correct alone.

312
00:14:45,880 --> 00:14:48,520
If setup time changes after a process improvement,

313
00:14:48,520 --> 00:14:50,520
someone has to update the planning standard

314
00:14:50,520 --> 00:14:54,120
and if a machine loses a capability after repair or tool change,

315
00:14:54,120 --> 00:14:55,960
that restriction has to reach the schedule.

316
00:14:55,960 --> 00:15:00,200
And if quality blocks a lot, the planning view needs that status before the order hits the machine.

317
00:15:00,200 --> 00:15:02,440
A finite schedule depends on these handoffs,

318
00:15:02,440 --> 00:15:06,840
but clean tables and accurate fields don't automatically give you a workable schedule.

319
00:15:06,840 --> 00:15:11,320
The system also needs to understand how product, operation, resource, material

320
00:15:11,320 --> 00:15:13,960
and production rules connect to each other,

321
00:15:13,960 --> 00:15:17,320
and that's where the model moves beyond records and into relationships.

322
00:15:17,320 --> 00:15:19,880
Constraints are relationships.

323
00:15:19,880 --> 00:15:22,840
Here's the thing, a schedule needs more than good records.

324
00:15:22,840 --> 00:15:25,080
It needs to know how those records depend on each other

325
00:15:25,080 --> 00:15:27,240
when a real order reaches a real machine.

326
00:15:27,240 --> 00:15:31,480
I like to think of a product process resource relationship as a sentence the system can understand.

327
00:15:31,480 --> 00:15:35,160
This product needs this operation, that operation runs on these resources.

328
00:15:35,160 --> 00:15:38,280
It needs this fixture, this skill, this approved material state

329
00:15:38,280 --> 00:15:40,360
and it must follow these process rules.

330
00:15:40,360 --> 00:15:43,720
Without that full sentence, the schedule fills gaps with assumptions.

331
00:15:43,720 --> 00:15:47,240
Take a machine housing as an example where the route may list machining

332
00:15:47,240 --> 00:15:48,680
as the next operation,

333
00:15:48,680 --> 00:15:52,200
and ERP might point to a work center with several CNC machines.

334
00:15:52,200 --> 00:15:54,840
On paper, any machine there looks suitable,

335
00:15:54,840 --> 00:15:57,560
but the housing may need a specific fourth-axis fixture,

336
00:15:57,560 --> 00:15:59,880
a program version approved only on two machines,

337
00:15:59,880 --> 00:16:03,240
and an operator who can run the inspection probe and sign off the first piece.

338
00:16:03,240 --> 00:16:05,960
The operation doesn't just need machining capacity,

339
00:16:05,960 --> 00:16:08,520
it needs a specific combination of conditions.

340
00:16:08,520 --> 00:16:11,240
This is where I find the knowledge graph idea useful.

341
00:16:11,240 --> 00:16:14,520
It's not magic and doesn't need to turn into a huge science project,

342
00:16:14,520 --> 00:16:18,360
because at its simplest, it's a connected map of things and the links between them

343
00:16:18,360 --> 00:16:22,040
where an order links to a product which links to its required operations,

344
00:16:22,040 --> 00:16:26,200
and each operation links to eligible machines, tools, people, materials,

345
00:16:26,200 --> 00:16:28,200
quality checks, and prior operations.

346
00:16:28,200 --> 00:16:32,040
Those links can also carry rules like this tool works for this part

347
00:16:32,040 --> 00:16:33,720
that machine can't run this revision.

348
00:16:33,720 --> 00:16:36,200
This operator can run the process only on the day shift,

349
00:16:36,200 --> 00:16:39,000
or this material lot needs a release before use.

350
00:16:39,000 --> 00:16:41,000
The scheduler can then reason over connections

351
00:16:41,000 --> 00:16:43,720
instead of reading isolated rows from separate tables.

352
00:16:43,720 --> 00:16:46,840
A pile of records might tell you machine 12 is free from 2 o'clock,

353
00:16:46,840 --> 00:16:48,680
or the 481 needs milling,

354
00:16:48,680 --> 00:16:50,440
and the fixture exists somewhere,

355
00:16:50,440 --> 00:16:52,760
but none of that proves the order can start at 2.

356
00:16:52,760 --> 00:16:58,920
The connected model asks harder questions like whether machine 12 is qualified for this operation,

357
00:16:58,920 --> 00:17:01,400
whether the fixture is free for the full run,

358
00:17:01,400 --> 00:17:03,480
whether the material lot matches the part,

359
00:17:03,480 --> 00:17:06,360
whether the prior operation finished an inspection released it,

360
00:17:06,360 --> 00:17:09,240
and whether the operator on that shift can run the setup.

361
00:17:09,240 --> 00:17:10,760
That is factory context.

362
00:17:10,760 --> 00:17:14,760
Alternate routing often exposes the gap between a technical route and a run-able one.

363
00:17:14,760 --> 00:17:17,240
Engineering may know a part can run on two machines,

364
00:17:17,240 --> 00:17:21,080
but that doesn't make both usable today because one machine may need a fixture already in use,

365
00:17:21,080 --> 00:17:23,480
while the other may lack the approved program revision.

366
00:17:23,480 --> 00:17:25,960
The tooling team might not have built the tool package,

367
00:17:25,960 --> 00:17:28,360
or quality, hasn't signed off on that resource.

368
00:17:28,360 --> 00:17:29,560
The machine can cut the part,

369
00:17:29,560 --> 00:17:32,200
but it can't hold the tolerance at the needed volume.

370
00:17:32,200 --> 00:17:33,320
Technically possible,

371
00:17:33,320 --> 00:17:36,040
is not ready approved and practical for this order.

372
00:17:36,040 --> 00:17:38,760
A useful schedule keeps those differences visible.

373
00:17:38,760 --> 00:17:42,600
Otherwise, the system solves overload by moving work onto an alternate machine,

374
00:17:42,600 --> 00:17:44,360
the supervisor can't actually use.

375
00:17:44,360 --> 00:17:46,120
The schedule looks better for 10 minutes,

376
00:17:46,120 --> 00:17:48,440
then the shift spends an hour calling engineering,

377
00:17:48,440 --> 00:17:49,400
looking for tools,

378
00:17:49,400 --> 00:17:51,560
and figuring out whether the move is even allowed,

379
00:17:51,560 --> 00:17:54,040
and that sort of schedule damages trust quickly.

380
00:17:54,040 --> 00:17:56,440
Presidents rules bring another set of links.

381
00:17:56,440 --> 00:17:58,920
Many operations must happen in a strict order,

382
00:17:58,920 --> 00:18:03,320
but the reason for the order matters because Operation 2 physically follows Operation 1,

383
00:18:03,320 --> 00:18:06,200
but can't begin until inspection clears the first.

384
00:18:06,200 --> 00:18:09,160
A part might need to cool before machining continues,

385
00:18:09,160 --> 00:18:11,240
a coding step may need a queue period,

386
00:18:11,240 --> 00:18:14,120
or an assembly might need a test result before final packing.

387
00:18:14,120 --> 00:18:16,760
The system needs to represent those conditions as rules,

388
00:18:16,760 --> 00:18:19,080
not just as assumed gaps between dates.

389
00:18:19,080 --> 00:18:23,160
Consider an order that finishes its first machining step at 10 in the morning.

390
00:18:23,160 --> 00:18:26,680
The next operation may have opened capacity at 10.15,

391
00:18:26,680 --> 00:18:29,000
but if inspection still waits for a measurement,

392
00:18:29,000 --> 00:18:32,920
that capacity is irrelevant because the part hasn't become available yet.

393
00:18:32,920 --> 00:18:36,040
A schedule that ignores that relationship can look perfectly ordered,

394
00:18:36,040 --> 00:18:38,120
with each operation having a start time,

395
00:18:38,120 --> 00:18:40,280
each machine staying within its calendar,

396
00:18:40,280 --> 00:18:42,840
and dates that may even meet the customer request.

397
00:18:42,840 --> 00:18:44,760
But then production tries to execute it,

398
00:18:44,760 --> 00:18:46,600
and discovers the part can't move.

399
00:18:46,600 --> 00:18:49,400
The problem didn't start on the shop floor.

400
00:18:49,400 --> 00:18:51,480
The model left out a dependency.

401
00:18:51,480 --> 00:18:55,960
This is why finite capacity scheduling isn't just a capacity calendar with colored bars.

402
00:18:55,960 --> 00:18:58,600
The scheduler needs a model of what work requires,

403
00:18:58,600 --> 00:19:01,320
what can perform it, what must happen first,

404
00:19:01,320 --> 00:19:03,320
and what conditions block the next step,

405
00:19:03,320 --> 00:19:05,800
because capacity matters, but without context,

406
00:19:05,800 --> 00:19:09,000
it only produces a more detailed version of the same bad promise.

407
00:19:09,960 --> 00:19:13,800
Once those relationships exist, the scheduler can begin its first real task.

408
00:19:13,800 --> 00:19:16,200
Take an order apart, examine each operation,

409
00:19:16,200 --> 00:19:18,200
and decide where and when the work can fit.

410
00:19:18,200 --> 00:19:20,360
From order to scheduled operation.

411
00:19:20,360 --> 00:19:23,560
Here's the shift that matters once the scheduler has the relationships in place.

412
00:19:23,560 --> 00:19:26,600
It stops treating a production order like one solid block of work.

413
00:19:26,600 --> 00:19:28,120
That's a bigger deal than it sounds,

414
00:19:28,120 --> 00:19:31,880
because an order never actually moves through a factory as one chunk.

415
00:19:31,880 --> 00:19:35,000
Each operation carries its own duration, its own resource needs,

416
00:19:35,000 --> 00:19:37,480
and the conditions that decide when the next step can begin,

417
00:19:37,480 --> 00:19:40,200
say you've got a simple order for a 100 machine components.

418
00:19:40,200 --> 00:19:43,320
The raw material gets cut to length first, then moves to milling,

419
00:19:43,320 --> 00:19:46,440
then inspection, then deburring, then assembly.

420
00:19:46,440 --> 00:19:49,960
A broad plan might show one start date and one finish date for the whole order,

421
00:19:49,960 --> 00:19:53,320
but a finite schedule has to place each operation where it can actually happen.

422
00:19:53,320 --> 00:19:54,680
That's a much more detailed question,

423
00:19:54,680 --> 00:19:57,240
and it's where most scheduling breakdowns start.

424
00:19:57,240 --> 00:20:00,920
For the first operation, the scheduler works out how much time the work really needs.

425
00:20:00,920 --> 00:20:01,880
Setup comes first.

426
00:20:01,880 --> 00:20:03,720
An operator may need to load a program,

427
00:20:03,720 --> 00:20:07,800
fit a fixture, prepare tools, verify material, and run a first part before normal production

428
00:20:07,800 --> 00:20:09,960
can even start, then comes run time,

429
00:20:09,960 --> 00:20:12,600
and that's usually where the simple math stops working.

430
00:20:12,600 --> 00:20:14,680
Run time comes from the expected cycle time,

431
00:20:14,680 --> 00:20:16,520
multiplied by the required quantity,

432
00:20:16,520 --> 00:20:18,920
plus any allowances the routing policy calls for.

433
00:20:18,920 --> 00:20:21,720
If each piece takes six minutes and you need a hundred pieces,

434
00:20:21,720 --> 00:20:23,320
that's where the calculation starts,

435
00:20:23,320 --> 00:20:24,520
but it can't stop there,

436
00:20:24,520 --> 00:20:27,400
because setup might eat a full hour before the first part ever runs.

437
00:20:27,400 --> 00:20:29,080
That changes the math completely.

438
00:20:29,080 --> 00:20:32,200
Now, a schedule also has to handle the time between operations.

439
00:20:32,200 --> 00:20:33,960
Some factories use planned queue allowances,

440
00:20:33,960 --> 00:20:35,480
especially for higher level planning.

441
00:20:35,480 --> 00:20:36,520
In practical terms,

442
00:20:36,520 --> 00:20:38,600
that allowance represents the expected weight

443
00:20:38,600 --> 00:20:40,920
before the next work center can pick up the job.

444
00:20:40,920 --> 00:20:43,320
Other plans want the finite schedule to calculate waiting

445
00:20:43,320 --> 00:20:45,000
from the actual loaded queue instead.

446
00:20:45,000 --> 00:20:47,160
Both approaches can work,

447
00:20:47,160 --> 00:20:49,000
but they answer slightly different questions

448
00:20:49,000 --> 00:20:51,000
about what the schedule actually tells you.

449
00:20:51,000 --> 00:20:52,520
Now, don't overlook transfer times.

450
00:20:52,520 --> 00:20:53,800
It's easy to forget,

451
00:20:53,800 --> 00:20:55,080
but it matters.

452
00:20:55,080 --> 00:20:57,400
A part might move straight from one operation to the next,

453
00:20:57,400 --> 00:20:58,760
or it might need a transport run,

454
00:20:58,760 --> 00:21:01,240
a staging area, a cooling period, or a batch handoff.

455
00:21:01,240 --> 00:21:05,000
If the part physically can't arrive at the next resource until later,

456
00:21:05,000 --> 00:21:08,520
the next operation can't start just because a calendar slot looks open,

457
00:21:08,520 --> 00:21:10,920
that's a physical constraint, not a planning choice.

458
00:21:10,920 --> 00:21:14,040
From there, the scheduler searches for eligible resources.

459
00:21:14,040 --> 00:21:15,480
Not every machine in the plant,

460
00:21:15,480 --> 00:21:17,000
just the ones the route allows,

461
00:21:17,000 --> 00:21:19,240
along with whatever constraints are attached to them.

462
00:21:19,240 --> 00:21:20,200
For each one,

463
00:21:20,200 --> 00:21:23,560
it checks the calendar and looks for a slot big enough to hold the work.

464
00:21:23,560 --> 00:21:25,160
Now, here's where it gets interesting.

465
00:21:25,160 --> 00:21:27,080
Sometimes the first open slot is the right answer,

466
00:21:27,080 --> 00:21:27,880
but often it isn't.

467
00:21:27,880 --> 00:21:29,720
A slot might open early on one machine,

468
00:21:29,720 --> 00:21:31,720
but grabbing it could force a longer setup,

469
00:21:31,720 --> 00:21:33,800
or delay another order with tighter conditions.

470
00:21:33,800 --> 00:21:35,640
A different resource might have a later slot

471
00:21:35,640 --> 00:21:37,400
that fits the current sequence better.

472
00:21:37,400 --> 00:21:39,640
The scheduler applies the policy it's been given,

473
00:21:39,640 --> 00:21:41,560
and places the operation into a window

474
00:21:41,560 --> 00:21:44,040
that respects the rules as closely as possible.

475
00:21:44,040 --> 00:21:45,240
When the work gets placed,

476
00:21:45,240 --> 00:21:46,520
the capacity gets reserved,

477
00:21:46,520 --> 00:21:49,400
and that matters because the slot no longer exists for anything else.

478
00:21:49,400 --> 00:21:52,760
If the machining operation runs from 8 in the morning until noon,

479
00:21:52,760 --> 00:21:54,200
including setup and production,

480
00:21:54,200 --> 00:21:56,920
the schedule records that commitment against the machine

481
00:21:56,920 --> 00:21:59,320
and against every other required resource.

482
00:21:59,320 --> 00:22:01,800
Each operation then passes a timing condition forward.

483
00:22:01,800 --> 00:22:04,520
The next step can't begin until the previous operation finishes,

484
00:22:04,520 --> 00:22:06,600
and the transfer or release condition is satisfied.

485
00:22:06,600 --> 00:22:08,360
If the first operation ends late,

486
00:22:08,360 --> 00:22:10,920
the available time for the next operation shifts to,

487
00:22:10,920 --> 00:22:13,560
the schedule carries that effect all the way through the route,

488
00:22:13,560 --> 00:22:16,600
rather than treating each work center like an isolated diary.

489
00:22:16,600 --> 00:22:19,640
This is why finite scheduling can expose a late order

490
00:22:19,640 --> 00:22:21,640
before the work ever reaches the floor.

491
00:22:21,640 --> 00:22:24,840
The system isn't just checking whether each operation fits somewhere.

492
00:22:24,840 --> 00:22:26,440
It follows the order through time,

493
00:22:26,440 --> 00:22:30,760
operation by operation, while every placement consumes real capacity.

494
00:22:30,760 --> 00:22:32,600
The schedule becomes a running commitment,

495
00:22:32,600 --> 00:22:34,600
so how does this actually work in practice?

496
00:22:34,600 --> 00:22:36,520
There are two common ways to run that process.

497
00:22:36,520 --> 00:22:38,600
One starts from the point where work becomes ready

498
00:22:38,600 --> 00:22:40,600
and pushes the order forward through the plant.

499
00:22:40,600 --> 00:22:42,120
The other starts from the promised date

500
00:22:42,120 --> 00:22:43,560
and works backward through the route,

501
00:22:43,560 --> 00:22:45,720
asking when each operation has to start.

502
00:22:45,720 --> 00:22:47,080
Both produce a schedule,

503
00:22:47,080 --> 00:22:48,920
but they show planners very different versions

504
00:22:48,920 --> 00:22:50,920
of the same production reality.

505
00:22:50,920 --> 00:22:53,000
Forward scheduling and backward scheduling,

506
00:22:53,000 --> 00:22:55,160
the first choice is where you start the clock.

507
00:22:55,160 --> 00:22:57,480
Forward scheduling begins when the order can actually start.

508
00:22:57,480 --> 00:23:01,560
When material is ready, the prior operation has released the work,

509
00:23:01,560 --> 00:23:03,880
and an eligible resource has time available.

510
00:23:03,880 --> 00:23:07,000
From that point, the scheduler places each operation in sequence

511
00:23:07,000 --> 00:23:09,720
and lets the finished date emerge from the loaded plan.

512
00:23:09,720 --> 00:23:11,640
That approach answers a practical question.

513
00:23:11,640 --> 00:23:14,520
If we release this order now, when can we actually complete it?

514
00:23:14,520 --> 00:23:16,520
Not the date marketing typed into a field,

515
00:23:16,520 --> 00:23:19,080
but the real date based on actual capacity.

516
00:23:19,080 --> 00:23:20,440
Picture an order waiting for,

517
00:23:20,440 --> 00:23:22,440
machining, inspection and assembly.

518
00:23:22,440 --> 00:23:23,960
Material is ready Monday morning,

519
00:23:23,960 --> 00:23:26,520
but the first suitable machine doesn't have an open slot

520
00:23:26,520 --> 00:23:28,120
until Tuesday afternoon.

521
00:23:28,120 --> 00:23:29,480
Forward scheduling starts there,

522
00:23:29,480 --> 00:23:32,520
not on the date somebody hoped for when the order entered ERP.

523
00:23:32,520 --> 00:23:34,520
Machining finishes when the machine, the setup,

524
00:23:34,520 --> 00:23:36,200
and the runtime say it finishes.

525
00:23:36,200 --> 00:23:38,920
Inspection follows when its own capacity allows,

526
00:23:38,920 --> 00:23:40,520
assembly comes after that,

527
00:23:40,520 --> 00:23:41,560
and by the end of the route,

528
00:23:41,560 --> 00:23:45,000
you have a projected completion date based on current conditions.

529
00:23:45,000 --> 00:23:47,400
That date could land later than the customer date,

530
00:23:47,400 --> 00:23:48,440
or earlier.

531
00:23:48,440 --> 00:23:51,080
Either way, it comes from available capacity,

532
00:23:51,080 --> 00:23:53,160
not from a target date typed into a field.

533
00:23:53,160 --> 00:23:54,600
That's a critical distinction.

534
00:23:54,600 --> 00:23:57,640
Backward scheduling takes the same order and starts at the other end,

535
00:23:57,640 --> 00:23:59,320
beginning with the due date and working back

536
00:23:59,320 --> 00:24:00,680
through every required operation.

537
00:24:00,680 --> 00:24:02,840
If assembly has to be done by Friday afternoon,

538
00:24:02,840 --> 00:24:05,320
the scheduler asks when assembly must start,

539
00:24:05,320 --> 00:24:07,000
then when inspection needs to finish,

540
00:24:07,000 --> 00:24:08,200
when machining must finish,

541
00:24:08,200 --> 00:24:10,920
and when material has to be ready for the first operation,

542
00:24:10,920 --> 00:24:13,560
that creates a required start time for each step.

543
00:24:13,560 --> 00:24:16,920
Backward scheduling earns its keep when you need to test a promise.

544
00:24:16,920 --> 00:24:19,640
Sales might ask whether a new order can ship by a certain date,

545
00:24:19,640 --> 00:24:21,000
customer service needs to know whether

546
00:24:21,000 --> 00:24:23,400
an existing commitment still holds after a change,

547
00:24:23,400 --> 00:24:25,160
and planning needs to see how much room remains

548
00:24:25,160 --> 00:24:27,160
before a due date becomes impossible.

549
00:24:27,160 --> 00:24:31,000
The schedule walks backward and checks whether the required slots actually exist.

550
00:24:31,000 --> 00:24:33,640
If enough available capacity shows up in the right sequence,

551
00:24:33,640 --> 00:24:36,280
the date stays feasible under the rules in the model.

552
00:24:36,280 --> 00:24:37,480
If the slots aren't there,

553
00:24:37,480 --> 00:24:41,000
the plan exposes a gap and that order might need an earlier release,

554
00:24:41,000 --> 00:24:43,320
a different route, extra capacity,

555
00:24:43,320 --> 00:24:45,320
or a change customer commitment.

556
00:24:45,320 --> 00:24:47,080
Both methods wrestle with the same constraints,

557
00:24:47,080 --> 00:24:48,840
but they ask different questions.

558
00:24:48,840 --> 00:24:53,320
Backward scheduling asks what has to happen and by when if we intend to meet this date.

559
00:24:53,320 --> 00:24:55,000
Forward scheduling asks.

560
00:24:55,000 --> 00:24:57,720
Given the work that's ready and the capacity we have,

561
00:24:57,720 --> 00:24:59,400
when will this order really finish?

562
00:24:59,400 --> 00:25:01,080
In many plans you need both views.

563
00:25:01,080 --> 00:25:03,960
Use backward scheduling when you're checking whether a requested date

564
00:25:03,960 --> 00:25:05,560
can survive the full route.

565
00:25:05,560 --> 00:25:08,280
It's a promise checking tool that forces the commercial date

566
00:25:08,280 --> 00:25:11,960
to face every required operation and the time each one needs.

567
00:25:11,960 --> 00:25:14,520
Use forward scheduling when you're deciding what work to release

568
00:25:14,520 --> 00:25:16,200
and what production can execute next,

569
00:25:16,200 --> 00:25:18,600
the shop floor doesn't run backward from a due date.

570
00:25:18,600 --> 00:25:20,760
It runs forward from material readiness,

571
00:25:20,760 --> 00:25:22,920
completed work, available resources,

572
00:25:22,920 --> 00:25:24,120
and the actual clock.

573
00:25:24,120 --> 00:25:25,800
A planner might look at the same order

574
00:25:25,800 --> 00:25:27,400
through both lenses on the same day.

575
00:25:27,400 --> 00:25:30,280
Backward scheduling could show that the customer date requires machining

576
00:25:30,280 --> 00:25:31,800
to finish by Wednesday morning,

577
00:25:31,800 --> 00:25:33,640
while forward scheduling could show that

578
00:25:33,640 --> 00:25:35,720
with the current queue and resource limits,

579
00:25:35,720 --> 00:25:38,040
Thursday is the earliest machining gets done.

580
00:25:38,040 --> 00:25:40,920
That difference is not an error between two scheduling methods.

581
00:25:40,920 --> 00:25:44,040
It's the decision gap and that gap is the actual output

582
00:25:44,040 --> 00:25:45,560
worth paying attention to.

583
00:25:45,560 --> 00:25:47,320
If the business still wants to protect that date,

584
00:25:47,320 --> 00:25:48,280
it needs action.

585
00:25:48,280 --> 00:25:50,920
The plant could approve overtime another order could move

586
00:25:50,920 --> 00:25:52,760
or maybe a valid alternate route exists

587
00:25:52,760 --> 00:25:54,120
or maybe the order finishes late

588
00:25:54,120 --> 00:25:55,560
and somebody needs to communicate that

589
00:25:55,560 --> 00:25:58,200
before the customer discovers it through a missed shipment.

590
00:25:58,200 --> 00:26:00,040
Late orders are useful information.

591
00:26:00,040 --> 00:26:01,640
Nobody likes them obviously.

592
00:26:01,640 --> 00:26:04,440
But a late flag from a finite schedule can tell you something

593
00:26:04,440 --> 00:26:06,360
a green plan date cannot.

594
00:26:06,360 --> 00:26:07,880
Current demand and current capacity

595
00:26:07,880 --> 00:26:10,680
don't fit together under the rules you've chosen.

596
00:26:10,680 --> 00:26:13,640
That gives people a chance to act early instead of reacting late.

597
00:26:13,640 --> 00:26:15,480
The wrong response is hiding the conflict

598
00:26:15,480 --> 00:26:17,160
with arbitrary lead time padding.

599
00:26:17,160 --> 00:26:18,920
A plant adds extra days to every route

600
00:26:18,920 --> 00:26:20,520
because schedules keep missing dates

601
00:26:20,520 --> 00:26:22,440
and the padded date starts to look safer.

602
00:26:22,440 --> 00:26:25,560
But nobody knows whether that added time covers a real constraint,

603
00:26:25,560 --> 00:26:28,040
a bad planning standard, a recurring queue,

604
00:26:28,040 --> 00:26:30,840
or just anxiety captured in master data.

605
00:26:30,840 --> 00:26:33,560
Padding can help when it represents a known and controlled buffer,

606
00:26:33,560 --> 00:26:35,560
but it becomes a problem when it covers overload

607
00:26:35,560 --> 00:26:36,600
without explaining it.

608
00:26:36,600 --> 00:26:37,880
You end up with long lead times,

609
00:26:37,880 --> 00:26:40,600
weak promise dates and the same daily firefighting

610
00:26:40,600 --> 00:26:42,120
only earlier in the calendar.

611
00:26:42,120 --> 00:26:44,280
Finite scheduling should expose the real gap,

612
00:26:44,280 --> 00:26:46,520
not bury it under extra calendar days.

613
00:26:46,520 --> 00:26:48,200
The tricky part is that a feasible slot

614
00:26:48,200 --> 00:26:49,720
isn't the same as the right slot.

615
00:26:49,720 --> 00:26:51,560
Capacity tells you what can run,

616
00:26:51,560 --> 00:26:53,720
but the next question is what should run first?

617
00:26:53,720 --> 00:26:55,560
Sequencing logic.

618
00:26:55,560 --> 00:26:56,520
Here's the real challenge.

619
00:26:56,520 --> 00:26:57,800
Once you have more than one order

620
00:26:57,800 --> 00:26:59,560
that fits into the available time,

621
00:26:59,560 --> 00:27:01,560
capacity tells you what could run.

622
00:27:01,560 --> 00:27:04,120
But sequencing decides which order gets the next slot

623
00:27:04,120 --> 00:27:06,520
and that single decision ripples through delivery,

624
00:27:06,520 --> 00:27:08,120
throughput, work in progress

625
00:27:08,120 --> 00:27:10,360
and how much setup time the plant burns through.

626
00:27:10,360 --> 00:27:12,920
The simplest approach is due date sequencing.

627
00:27:12,920 --> 00:27:15,080
You run the order with the earliest due date first,

628
00:27:15,080 --> 00:27:16,920
people like it because it's easy to explain,

629
00:27:16,920 --> 00:27:19,880
and it works when routes and processing times don't very much.

630
00:27:19,880 --> 00:27:20,840
But here's the thing,

631
00:27:20,840 --> 00:27:23,000
an early due date doesn't always mean the order

632
00:27:23,000 --> 00:27:24,200
creates the biggest risk,

633
00:27:24,200 --> 00:27:25,720
because a short job due tomorrow

634
00:27:25,720 --> 00:27:27,800
may sit ahead of a long job due the next day.

635
00:27:27,800 --> 00:27:30,200
If that long job needs most of the remaining capacity,

636
00:27:30,200 --> 00:27:33,160
delaying it could push several later operations into trouble.

637
00:27:33,160 --> 00:27:35,080
Due date sequencing tracks the calendar,

638
00:27:35,080 --> 00:27:37,240
but it may not see the full cost of the queue.

639
00:27:37,240 --> 00:27:39,320
Customer priority takes a different view.

640
00:27:39,320 --> 00:27:40,920
You decide that a service part,

641
00:27:40,920 --> 00:27:42,040
a contract customer,

642
00:27:42,040 --> 00:27:45,000
or a production line facing a shutdown gets preference,

643
00:27:45,000 --> 00:27:46,600
and that can be the right business call.

644
00:27:46,600 --> 00:27:49,080
But every priority order displaces something else,

645
00:27:49,080 --> 00:27:50,760
and if every order is marked urgent,

646
00:27:50,760 --> 00:27:52,120
the label stops helping.

647
00:27:52,120 --> 00:27:53,880
Some plants use shortest processing time,

648
00:27:53,880 --> 00:27:55,000
running smaller jobs first

649
00:27:55,000 --> 00:27:56,600
because short jobs clear the queue fast,

650
00:27:56,600 --> 00:27:57,720
reduce work in progress,

651
00:27:57,720 --> 00:27:59,160
and keep material moving.

652
00:27:59,160 --> 00:28:01,640
The downside is obvious when you stop to think about it.

653
00:28:01,640 --> 00:28:03,160
Long jobs can wait too long,

654
00:28:03,160 --> 00:28:04,920
even when they support an important delivery

655
00:28:04,920 --> 00:28:07,480
or consumes gas capacity later in the route.

656
00:28:07,480 --> 00:28:10,360
Critical ratio tries to bring more context into the choice

657
00:28:10,360 --> 00:28:12,920
by comparing the time remaining until an order is due

658
00:28:12,920 --> 00:28:14,840
with the work still required to finish it.

659
00:28:14,840 --> 00:28:17,480
An order with little time left and a lot of work remaining

660
00:28:17,480 --> 00:28:19,720
gets more attention than an order with plenty of time.

661
00:28:19,720 --> 00:28:21,480
That can help planners spot trouble earlier,

662
00:28:21,480 --> 00:28:23,880
but it still depends on good remaining time estimates,

663
00:28:23,880 --> 00:28:25,240
realistic routing data,

664
00:28:25,240 --> 00:28:27,800
and a schedule that knows which operations can really run.

665
00:28:27,800 --> 00:28:30,520
A need ratio built on weak assumptions is still a need guess.

666
00:28:30,520 --> 00:28:32,520
Campaign sequencing focuses on setups.

667
00:28:32,520 --> 00:28:34,760
If several orders use the same material grade,

668
00:28:34,760 --> 00:28:37,160
color, recipe, fixture, or tool pack,

669
00:28:37,160 --> 00:28:38,600
the plant can group them together,

670
00:28:38,600 --> 00:28:40,920
so the machine runs longer between changes,

671
00:28:40,920 --> 00:28:43,320
and the shift loses less time to cleaning,

672
00:28:43,320 --> 00:28:46,520
tool swaps, parameter checks, and first part approval.

673
00:28:46,520 --> 00:28:48,920
That's often sensible, but it also creates tension.

674
00:28:48,920 --> 00:28:51,640
Suppose an urgent order needs a different setup family.

675
00:28:51,640 --> 00:28:53,160
If you interrupt the campaign,

676
00:28:53,160 --> 00:28:54,920
you spend time changing over,

677
00:28:54,920 --> 00:28:55,800
run the urgent job,

678
00:28:55,800 --> 00:28:57,640
then spend more time changing back.

679
00:28:57,640 --> 00:28:58,920
If you keep the campaign intact,

680
00:28:58,920 --> 00:28:59,960
you protect machine time,

681
00:28:59,960 --> 00:29:01,560
but delay the urgent order.

682
00:29:01,560 --> 00:29:02,600
Neither choice is neutral

683
00:29:02,600 --> 00:29:04,840
and both move cost and risk somewhere else.

684
00:29:04,840 --> 00:29:07,640
The same issue appears with material and process conditions.

685
00:29:07,640 --> 00:29:09,800
A material lot may have a use by date,

686
00:29:09,800 --> 00:29:11,720
so delaying that order creates waste.

687
00:29:11,720 --> 00:29:13,880
A cutting tool may have limited remaining life,

688
00:29:13,880 --> 00:29:16,200
affecting whether two jobs should run back to back.

689
00:29:16,200 --> 00:29:18,040
A quality hold can block one order,

690
00:29:18,040 --> 00:29:19,800
even if its due date looks alarming.

691
00:29:19,800 --> 00:29:23,000
An operator may hold approval for only certain products or processes,

692
00:29:23,000 --> 00:29:25,000
and that person may work only one shift.

693
00:29:25,000 --> 00:29:27,640
These details aren't side notes added after the schedule.

694
00:29:27,640 --> 00:29:29,240
They help decide the sequence.

695
00:29:29,240 --> 00:29:31,480
A scheduling rule should make its purpose clear.

696
00:29:31,480 --> 00:29:33,480
Are you trying to protect customer commitments,

697
00:29:33,480 --> 00:29:35,000
reduce change over losses,

698
00:29:35,000 --> 00:29:36,760
or keep an assembly line supplied?

699
00:29:36,760 --> 00:29:40,040
Are you trying to avoid expired material or reduce over time?

700
00:29:40,040 --> 00:29:42,120
A plant can care about all of those things,

701
00:29:42,120 --> 00:29:45,000
but it cannot maximize every outcome at the same moment.

702
00:29:45,000 --> 00:29:46,680
That's why sequencing is policy expressed

703
00:29:46,680 --> 00:29:48,040
through scheduling logic.

704
00:29:48,040 --> 00:29:50,200
The software can apply the rule consistently,

705
00:29:50,200 --> 00:29:52,600
and test many possible sequences far faster

706
00:29:52,600 --> 00:29:54,600
than a person working from a paper list,

707
00:29:54,600 --> 00:29:56,600
but somebody still needs to decide

708
00:29:56,600 --> 00:29:59,480
what the plant should favor when the rules compete.

709
00:29:59,480 --> 00:30:01,400
I'd be careful when a scheduling tool claims

710
00:30:01,400 --> 00:30:04,280
it has found the best sequence without showing the trade-offs.

711
00:30:04,280 --> 00:30:08,040
Best for what? shorter setups, more on-time orders, less queue time,

712
00:30:08,040 --> 00:30:09,640
fewer late service parts?

713
00:30:09,640 --> 00:30:12,600
The answer comes from production policy and commercial judgment,

714
00:30:12,600 --> 00:30:14,440
not from mass floating above the factory.

715
00:30:14,440 --> 00:30:16,520
In practical terms, a good finite schedule

716
00:30:16,520 --> 00:30:17,960
doesn't hide those choices.

717
00:30:17,960 --> 00:30:19,160
It makes them visible,

718
00:30:19,160 --> 00:30:21,240
telling the planner that running this job first

719
00:30:21,240 --> 00:30:24,120
protects one condition and creates pressure somewhere else.

720
00:30:24,120 --> 00:30:25,960
So let's bring this into one morning shift,

721
00:30:25,960 --> 00:30:28,680
where three orders wait at the same constraint machine

722
00:30:28,680 --> 00:30:31,560
and each sequence choice carries a different consequence.

723
00:30:31,560 --> 00:30:32,920
The morning shift decision.

724
00:30:32,920 --> 00:30:35,960
Picture the start of a morning shift at that CNC machining centre.

725
00:30:35,960 --> 00:30:37,320
The previous job just finished,

726
00:30:37,320 --> 00:30:39,240
and three orders are ready for the next slot.

727
00:30:39,240 --> 00:30:41,800
All three need the bottleneck and all three matter.

728
00:30:41,800 --> 00:30:44,280
The machine has one operator, one current setup,

729
00:30:44,280 --> 00:30:46,520
and no spare hours hiding in the day.

730
00:30:46,520 --> 00:30:48,520
Order A is an urgent repair part.

731
00:30:48,520 --> 00:30:49,640
The customer needs it quickly,

732
00:30:49,640 --> 00:30:51,560
and customer service has already flagged it.

733
00:30:51,560 --> 00:30:53,160
It needs a different fixture and tool pack

734
00:30:53,160 --> 00:30:55,160
from the job that just came off the machine.

735
00:30:55,160 --> 00:30:57,960
Orders B and C are less urgent, at least commercially,

736
00:30:57,960 --> 00:31:01,160
but both belong to the same setup family as the completed job.

737
00:31:01,160 --> 00:31:03,080
Their material is ready, the fixture is installed,

738
00:31:03,080 --> 00:31:04,440
the program is loaded.

739
00:31:04,440 --> 00:31:06,200
If the team runs B and C now,

740
00:31:06,200 --> 00:31:07,800
they can keep the machine in a campaign

741
00:31:07,800 --> 00:31:09,640
that avoids two extra changeovers.

742
00:31:09,640 --> 00:31:11,560
The planner now faces a real decision,

743
00:31:11,560 --> 00:31:12,920
not a sorting exercise.

744
00:31:12,920 --> 00:31:15,160
The first option is simple to explain.

745
00:31:15,160 --> 00:31:17,080
Run the urgent repair part first.

746
00:31:17,080 --> 00:31:18,920
The operator removes the current fixture,

747
00:31:18,920 --> 00:31:20,360
loads the fixture for order A,

748
00:31:20,360 --> 00:31:23,080
changes tools, checks the program and completes the setup.

749
00:31:23,080 --> 00:31:25,160
The urgent part then moves through the bottleneck

750
00:31:25,160 --> 00:31:27,720
and the plant protects the customer commitment as far as it can.

751
00:31:27,720 --> 00:31:29,080
That may be the right call,

752
00:31:29,080 --> 00:31:32,360
but after order A finishes, orders B and C still need to run,

753
00:31:32,360 --> 00:31:33,800
and the machine needs another changeover

754
00:31:33,800 --> 00:31:35,400
to return to their setup family.

755
00:31:35,400 --> 00:31:38,120
The time spent on both changes comes from the same limited day.

756
00:31:38,120 --> 00:31:39,720
If B feeds a later process,

757
00:31:39,720 --> 00:31:41,800
its delay can push work into the next shift

758
00:31:41,800 --> 00:31:44,120
and if C supports a delivery later in the week,

759
00:31:44,120 --> 00:31:47,560
that extra lost time may turn a manageable plan into a late order.

760
00:31:47,560 --> 00:31:50,280
The urgent part gains time, but other orders lose it.

761
00:31:50,280 --> 00:31:52,040
That doesn't mean urgent work should wait.

762
00:31:52,040 --> 00:31:54,680
It means the schedule should show the cost of putting it first

763
00:31:54,680 --> 00:31:56,920
so the planner can decide with open eyes,

764
00:31:56,920 --> 00:31:58,840
rather than discovering the effect.

765
00:31:58,840 --> 00:32:02,520
When somebody asks, why two apparently safe orders suddenly moved?

766
00:32:02,520 --> 00:32:05,480
The second option keeps the campaign intact.

767
00:32:05,480 --> 00:32:07,880
Run B then C while the setup is already in place,

768
00:32:07,880 --> 00:32:09,960
the operator avoids the immediate changeover

769
00:32:09,960 --> 00:32:12,840
so the machine converts more of the shift into production time.

770
00:32:12,840 --> 00:32:15,160
The queue behind the bottleneck may flow more smoothly,

771
00:32:15,160 --> 00:32:16,680
especially if both orders feed work

772
00:32:16,680 --> 00:32:19,720
that already has people and capacity waiting downstream.

773
00:32:19,720 --> 00:32:22,600
From a production view, this can look like the sensible choice.

774
00:32:22,600 --> 00:32:24,360
Still, order A remains in the queue.

775
00:32:24,360 --> 00:32:26,360
Customer service may need to call the customer,

776
00:32:26,360 --> 00:32:28,440
who may accept the revised timing or may not.

777
00:32:28,440 --> 00:32:31,000
A repair part can hold up equipment outside your plant,

778
00:32:31,000 --> 00:32:34,520
and the commercial effect may outweigh the time-save through campaign sequencing.

779
00:32:34,520 --> 00:32:37,720
Production efficiency is not the only measure of a good decision.

780
00:32:37,720 --> 00:32:39,720
The third option looks for an alternate route.

781
00:32:39,720 --> 00:32:43,640
Perhaps another machine can run order A while the bottleneck stays with B and C.

782
00:32:43,640 --> 00:32:44,760
That sounds attractive,

783
00:32:44,760 --> 00:32:47,800
because it appears to protect both the urgent order and the campaign,

784
00:32:47,800 --> 00:32:49,720
but before anyone treats it as a free answer,

785
00:32:49,720 --> 00:32:51,160
the plant needs to test it.

786
00:32:51,160 --> 00:32:54,040
Can the alternate machine hold the required tolerance?

787
00:32:54,040 --> 00:32:56,920
Is the correct fixture free and physically compatible?

788
00:32:56,920 --> 00:32:59,400
Does the operator hold the right process approval?

789
00:32:59,400 --> 00:33:00,760
Is the program current?

790
00:33:00,760 --> 00:33:02,600
Does quality allow that part to run there

791
00:33:02,600 --> 00:33:05,080
or does it need an engineering or quality review first?

792
00:33:05,080 --> 00:33:06,200
The alternate route may work,

793
00:33:06,200 --> 00:33:08,280
but it may also turn into a fast-looking decision

794
00:33:08,280 --> 00:33:10,840
that creates rework, delay, or a non-conforming part.

795
00:33:10,840 --> 00:33:12,760
A routing field that names two machines

796
00:33:12,760 --> 00:33:15,000
cannot settle those questions by itself.

797
00:33:15,000 --> 00:33:17,400
This is where finite capacity scheduling earns its place.

798
00:33:17,400 --> 00:33:19,560
It doesn't tell every plant to protect campaigns

799
00:33:19,560 --> 00:33:21,320
or to chase the earliest due date

800
00:33:21,320 --> 00:33:23,400
or to move urgent work to another resource.

801
00:33:23,400 --> 00:33:25,800
It takes each option and tests the conditions around it.

802
00:33:26,520 --> 00:33:27,560
Run order A first,

803
00:33:27,560 --> 00:33:29,480
and the schedule can show the added changeovers

804
00:33:29,480 --> 00:33:30,760
and later Q effects.

805
00:33:30,760 --> 00:33:32,280
Keep B and C together.

806
00:33:32,280 --> 00:33:34,760
And it can show the exposure on the repair part date.

807
00:33:34,760 --> 00:33:36,440
Move A to another machine,

808
00:33:36,440 --> 00:33:38,760
and it can test whether the needed machine time,

809
00:33:38,760 --> 00:33:41,320
tooling, labor, and approval exist together.

810
00:33:41,320 --> 00:33:43,080
The planner still owns the call.

811
00:33:43,080 --> 00:33:46,040
A good schedule gives people a shared frame for that call.

812
00:33:46,040 --> 00:33:48,760
The supervisor can challenge whether the option fits the shift.

813
00:33:48,760 --> 00:33:51,560
Customer service can explain the customer impact and quality,

814
00:33:51,560 --> 00:33:54,600
and engineering can confirm whether the alternate route is real.

815
00:33:55,560 --> 00:33:57,800
Nobody needs to pretend that one dispatch rule

816
00:33:57,800 --> 00:34:00,520
fits every product, customer, and production day.

817
00:34:00,520 --> 00:34:01,960
There is one part of this decision

818
00:34:01,960 --> 00:34:04,040
that plans often reduced to a rough average

819
00:34:04,040 --> 00:34:06,680
or leave out entirely, the changeover itself,

820
00:34:06,680 --> 00:34:09,720
that lost production time leads its own place in the schedule.

821
00:34:09,720 --> 00:34:11,560
Changeovers are scheduled work.

822
00:34:11,560 --> 00:34:13,080
Here's a truth that sounds obvious,

823
00:34:13,080 --> 00:34:15,560
but gets buried under noise more often than it should.

824
00:34:15,560 --> 00:34:18,360
A changeover isn't dead time between two production orders.

825
00:34:18,360 --> 00:34:20,520
It's real work that burns the same resource hours

826
00:34:20,520 --> 00:34:21,400
that production does.

827
00:34:21,400 --> 00:34:24,040
When an operator stops the machine,

828
00:34:24,040 --> 00:34:26,200
pulls tools, cleanser tanks, swaps fixtures,

829
00:34:26,200 --> 00:34:28,360
loads a new program, dials in parameters,

830
00:34:28,360 --> 00:34:31,400
and runs a first piece check to prove the next part meets spec

831
00:34:31,400 --> 00:34:33,800
that machine isn't making anything that ships.

832
00:34:33,800 --> 00:34:36,440
If the schedule doesn't account for that time honestly,

833
00:34:36,440 --> 00:34:39,720
it quietly spins the same hour twice in two different places.

834
00:34:39,720 --> 00:34:42,040
I know that sounds basic, but here's the real challenge.

835
00:34:42,040 --> 00:34:44,040
Most plans treat setup as a rough allowance

836
00:34:44,040 --> 00:34:45,720
around the edges of a work center,

837
00:34:45,720 --> 00:34:47,960
maybe a single average number in the routing,

838
00:34:47,960 --> 00:34:48,920
or even nothing at all,

839
00:34:48,920 --> 00:34:51,240
because someone assumed the supervisor would just handle it.

840
00:34:52,520 --> 00:34:54,520
And that schedule looks great on paper

841
00:34:54,520 --> 00:34:56,520
until the shift tries to switch from one product

842
00:34:56,520 --> 00:34:58,840
to another and the whole thing comes apart.

843
00:34:58,840 --> 00:35:02,600
A finite schedule treats setup as a proper operation on the resource,

844
00:35:02,600 --> 00:35:04,760
slotting it in before the production run,

845
00:35:04,760 --> 00:35:06,760
locking the machine during the changeover,

846
00:35:06,760 --> 00:35:08,760
and including every person, tool, and check

847
00:35:08,760 --> 00:35:10,200
that setup actually needs.

848
00:35:10,200 --> 00:35:12,200
The result is a view of capacity that matches

849
00:35:12,200 --> 00:35:13,640
what the plant can deliver.

850
00:35:13,640 --> 00:35:14,840
When the data supports it,

851
00:35:14,840 --> 00:35:17,480
there are two kinds of setup time worth separating.

852
00:35:17,480 --> 00:35:19,800
The first sequence independent setup

853
00:35:19,800 --> 00:35:21,320
is work you do for an order,

854
00:35:21,320 --> 00:35:23,240
no matter what ran before it,

855
00:35:23,240 --> 00:35:25,400
loading an order specific program,

856
00:35:25,400 --> 00:35:28,280
verifying material, or mounting a unique fixture.

857
00:35:28,280 --> 00:35:30,920
The second sequence dependent change over time

858
00:35:30,920 --> 00:35:32,600
depends on what happened before.

859
00:35:32,600 --> 00:35:34,440
Moving from one product family to another

860
00:35:34,440 --> 00:35:37,320
can take way longer than moving between two similar parts

861
00:35:37,320 --> 00:35:40,440
and that difference drives most of the value in good sequencing.

862
00:35:40,440 --> 00:35:42,040
Think about a paint line.

863
00:35:42,040 --> 00:35:44,680
Switching from one shade to another might be a quick adjustment

864
00:35:44,680 --> 00:35:47,800
or it could need a full clean-out to avoid contamination.

865
00:35:47,800 --> 00:35:50,920
In machining, one transition keeps the same tool set and fixture

866
00:35:50,920 --> 00:35:53,800
while the next demands a complete tear down and fresh setup.

867
00:35:53,800 --> 00:35:56,120
And in food chemicals, plastics, or coatings,

868
00:35:56,120 --> 00:35:58,360
a recipe change may require cleaning and validation

869
00:35:58,360 --> 00:36:00,120
that don't care what the schedule thinks,

870
00:36:00,120 --> 00:36:02,440
that same pattern shows up in family matrices.

871
00:36:02,440 --> 00:36:04,360
A family matrix records the transition time

872
00:36:04,360 --> 00:36:07,320
between setup families, color, alloy, tool set,

873
00:36:07,320 --> 00:36:09,560
recipe, cleaning class, or fixture type.

874
00:36:09,560 --> 00:36:10,760
Instead of telling the scheduler

875
00:36:10,760 --> 00:36:13,160
that every change takes one average number,

876
00:36:13,160 --> 00:36:15,880
the matrix says moving from family A to family B

877
00:36:15,880 --> 00:36:16,920
takes this amount,

878
00:36:16,920 --> 00:36:18,920
but moving from family A to family D

879
00:36:18,920 --> 00:36:20,760
might take three times longer.

880
00:36:20,760 --> 00:36:22,680
That matches how plants actually work.

881
00:36:22,680 --> 00:36:25,320
In practical terms, you can start with just the transitions

882
00:36:25,320 --> 00:36:27,000
that create the most friction.

883
00:36:27,000 --> 00:36:29,480
The ones the production team already complains about.

884
00:36:29,480 --> 00:36:32,040
You don't need to build an enormous matrix on day one.

885
00:36:32,040 --> 00:36:34,600
In most factories, a small number of high impact transitions

886
00:36:34,600 --> 00:36:36,120
cause nearly all the trouble.

887
00:36:36,120 --> 00:36:36,920
Start there.

888
00:36:36,920 --> 00:36:39,000
If a particular alloy change requires cleaning

889
00:36:39,000 --> 00:36:41,480
or a fixture swap regularly blocks the bottleneck,

890
00:36:41,480 --> 00:36:43,800
model that transition before trying to describe

891
00:36:43,800 --> 00:36:45,400
every possible combination.

892
00:36:45,400 --> 00:36:47,480
And don't forget the time after the physical setup.

893
00:36:47,480 --> 00:36:50,520
A machine can look ready while the first part still waits for inspection.

894
00:36:50,840 --> 00:36:52,680
Some processes need a warm up cycle,

895
00:36:52,680 --> 00:36:55,000
others need a test piece, an initial measurement,

896
00:36:55,000 --> 00:36:58,680
a quality sign-off, or a process check before normal production can begin.

897
00:36:58,680 --> 00:37:01,320
If that activity holds the machine or needs a person

898
00:37:01,320 --> 00:37:02,920
who isn't immediately available,

899
00:37:02,920 --> 00:37:04,440
it belongs in the schedule.

900
00:37:04,440 --> 00:37:05,960
Production teams already know this,

901
00:37:05,960 --> 00:37:09,720
but the risk is when the planning model calls the machine available,

902
00:37:09,720 --> 00:37:11,800
the second the operator presses start.

903
00:37:11,800 --> 00:37:13,960
Average setup times cause their own problem,

904
00:37:13,960 --> 00:37:15,720
an average can work for broad planning,

905
00:37:15,720 --> 00:37:17,800
but it hides the transition that really matters.

906
00:37:17,800 --> 00:37:20,200
Say a work center usually changes over in 40 minutes

907
00:37:20,200 --> 00:37:23,480
but one product switch needs two hours because of cleaning and approval.

908
00:37:23,480 --> 00:37:25,960
If the schedule uses 40 minutes for every change,

909
00:37:25,960 --> 00:37:28,760
it will repeatedly promise time the plant cannot deliver.

910
00:37:28,760 --> 00:37:30,200
The average looks reasonable,

911
00:37:30,200 --> 00:37:31,720
but the sequence fails anyway.

912
00:37:31,720 --> 00:37:34,600
And that's a problem no amount of wishful thinking can fix.

913
00:37:34,600 --> 00:37:37,800
This is why setup reduction and finite scheduling belong together,

914
00:37:37,800 --> 00:37:39,240
but are not the same thing.

915
00:37:39,240 --> 00:37:41,960
Set-up reduction changes the process through better fixtures,

916
00:37:41,960 --> 00:37:44,200
standard tool packs, pre-staged material,

917
00:37:44,200 --> 00:37:45,800
or improved work instructions.

918
00:37:45,800 --> 00:37:48,840
Finite scheduling works with the setup time that exists today

919
00:37:48,840 --> 00:37:50,520
and places orders accordingly.

920
00:37:50,520 --> 00:37:52,600
One improves capacity, the other stops the plant

921
00:37:52,600 --> 00:37:54,920
from pretending that capacity already exists.

922
00:37:54,920 --> 00:37:58,040
A scheduler doesn't need perfect change over data to help.

923
00:37:58,040 --> 00:38:00,840
It needs enough truth to stop hiding the biggest losses,

924
00:38:00,840 --> 00:38:04,520
so the planner can see when an urgent order breaks a campaign,

925
00:38:04,520 --> 00:38:06,520
how much time that interruption eats,

926
00:38:06,520 --> 00:38:08,600
and which later orders come under pressure.

927
00:38:08,600 --> 00:38:12,760
Even a perfectly sequenced machine cannot start without the right material.

928
00:38:12,760 --> 00:38:14,280
The schedule may reserve the slot,

929
00:38:14,280 --> 00:38:15,560
the operator may be ready,

930
00:38:15,560 --> 00:38:17,400
and the fixture may already be mounted.

931
00:38:17,400 --> 00:38:20,520
But if the components themselves remain unavailable for release,

932
00:38:20,520 --> 00:38:21,560
none of that matters.

933
00:38:21,560 --> 00:38:24,280
Material availability and release control.

934
00:38:24,280 --> 00:38:27,320
Material planning can tell you a component should be available next week.

935
00:38:27,320 --> 00:38:28,200
That's useful,

936
00:38:28,200 --> 00:38:32,120
but it doesn't tell the operator whether that component is actually sitting at the point of use,

937
00:38:32,120 --> 00:38:33,720
inspected, identified,

938
00:38:33,720 --> 00:38:36,200
and ready to consume when the machine slot opens.

939
00:38:36,200 --> 00:38:37,720
Those are completely different states.

940
00:38:37,720 --> 00:38:40,280
MRP calculates demand from bills of material,

941
00:38:40,280 --> 00:38:42,520
current stock, supplier lead times,

942
00:38:42,520 --> 00:38:43,560
and planned orders,

943
00:38:43,560 --> 00:38:45,400
sending purchase signals early

944
00:38:45,400 --> 00:38:47,080
and showing where shortages may appear.

945
00:38:47,080 --> 00:38:51,400
That work matters because a finite schedule cannot run parts without material,

946
00:38:51,400 --> 00:38:53,640
but planned availability is a planning signal,

947
00:38:53,640 --> 00:38:56,360
while physical availability is an execution condition.

948
00:38:56,360 --> 00:38:58,760
Picture an order scheduled on the bottleneck at 8 in the morning.

949
00:38:58,760 --> 00:39:03,000
The ERP record may show enough stock because the receipt posted overnight,

950
00:39:03,000 --> 00:39:05,480
but the material might still be sitting in receiving,

951
00:39:05,480 --> 00:39:07,080
quality might need to inspect it,

952
00:39:07,080 --> 00:39:09,000
the warehouse might not have picked the kit,

953
00:39:09,000 --> 00:39:10,520
or the material might be on the line,

954
00:39:10,520 --> 00:39:13,080
but lacks the traceability record required for release.

955
00:39:13,080 --> 00:39:15,960
The machine slot exists, but the job still cannot start.

956
00:39:15,960 --> 00:39:18,840
Partial kits create the same problem in a less obvious way.

957
00:39:18,840 --> 00:39:20,680
You may have most of the items for an assembly,

958
00:39:20,680 --> 00:39:22,280
but one boarding component is late.

959
00:39:22,280 --> 00:39:25,080
Starting the order could burn labor and floor space,

960
00:39:25,080 --> 00:39:27,720
then leave unfinished work waiting for a single missing part.

961
00:39:27,720 --> 00:39:28,920
And in some cases that makes sense,

962
00:39:28,920 --> 00:39:32,120
in others, it creates work and progress that hides the shortage,

963
00:39:32,120 --> 00:39:33,320
instead of fixing it.

964
00:39:33,320 --> 00:39:35,080
So the rule needs to be explicit.

965
00:39:35,080 --> 00:39:37,720
Not just a policy, but a condition the schedule can check.

966
00:39:37,720 --> 00:39:39,000
The schedule needs the rule.

967
00:39:39,000 --> 00:39:40,600
Some products can start with a partial kit,

968
00:39:40,600 --> 00:39:42,840
because the missing item comes in late on the route,

969
00:39:42,840 --> 00:39:44,680
while others need every controlled component

970
00:39:44,680 --> 00:39:47,160
ready before the first operation begins.

971
00:39:47,160 --> 00:39:49,400
A finite schedule should respect that difference,

972
00:39:49,400 --> 00:39:53,000
rather than treating all material status as one green or red field.

973
00:39:53,000 --> 00:39:54,760
Lot restrictions add another layer.

974
00:39:54,760 --> 00:39:57,240
A customer may require a specific material lot,

975
00:39:57,240 --> 00:40:00,040
a regulated process may require approved traceability,

976
00:40:00,040 --> 00:40:01,880
or a chemical coating or food ingredient

977
00:40:01,880 --> 00:40:03,640
may have a limited use window.

978
00:40:03,640 --> 00:40:05,320
You may have stock in the system,

979
00:40:05,320 --> 00:40:08,840
but not stock that the order can legally or technically use.

980
00:40:08,840 --> 00:40:10,520
Substitutes need the same care.

981
00:40:10,520 --> 00:40:12,760
A planner may spot an equivalent material code

982
00:40:12,760 --> 00:40:14,360
and assume it solves the shortage,

983
00:40:14,360 --> 00:40:17,640
but engineering may need to approve the substitute in quality

984
00:40:17,640 --> 00:40:19,000
may need to release it,

985
00:40:19,000 --> 00:40:21,560
and the root, work instruction, test plan,

986
00:40:21,560 --> 00:40:24,120
or customer agreement may change as a result.

987
00:40:24,120 --> 00:40:26,360
Until that approval exists,

988
00:40:26,360 --> 00:40:29,320
the substitute is a possible option, not available material.

989
00:40:29,320 --> 00:40:31,800
These are the kinds of details that separate a real schedule

990
00:40:31,800 --> 00:40:33,000
from a theoretical one.

991
00:40:33,000 --> 00:40:35,080
This is why material readiness needs more detail

992
00:40:35,080 --> 00:40:36,200
than an inventory balance.

993
00:40:36,200 --> 00:40:37,880
For a given order,

994
00:40:37,880 --> 00:40:39,160
the scheduling model needs to know

995
00:40:39,160 --> 00:40:41,080
whether the required quantity exists,

996
00:40:41,080 --> 00:40:42,840
whether the correct lot is allowed,

997
00:40:42,840 --> 00:40:44,520
whether inspection has released it,

998
00:40:44,520 --> 00:40:46,920
and whether the material can reach the operation in time.

999
00:40:46,920 --> 00:40:48,840
That sounds like a lot of detail,

1000
00:40:48,840 --> 00:40:50,520
but the alternative is worse.

1001
00:40:50,520 --> 00:40:52,760
You reserve constrained capacity for work

1002
00:40:52,760 --> 00:40:55,000
that stops before the first part runs.

1003
00:40:55,000 --> 00:40:56,600
Many plans respond to uncertainty

1004
00:40:56,600 --> 00:40:58,520
by releasing every order early.

1005
00:40:58,520 --> 00:40:59,720
The reasoning makes sense.

1006
00:40:59,720 --> 00:41:01,160
If all work sits on the floor,

1007
00:41:01,160 --> 00:41:03,880
the team can pick the next ready job when something goes wrong.

1008
00:41:03,880 --> 00:41:06,120
But early release often creates a larger queue

1009
00:41:06,120 --> 00:41:07,000
around the bottleneck,

1010
00:41:07,000 --> 00:41:08,680
more material handling, more searching,

1011
00:41:08,680 --> 00:41:11,160
and more arguments about which order really matters.

1012
00:41:11,160 --> 00:41:14,120
The floor becomes a storage area for planning uncertainty.

1013
00:41:14,120 --> 00:41:16,120
If this sounds familiar, you're not alone.

1014
00:41:16,120 --> 00:41:17,640
Most plans hit this wall.

1015
00:41:17,640 --> 00:41:20,040
A long queue also hides actual priority.

1016
00:41:20,040 --> 00:41:22,280
Every order looks urgent once it's been released,

1017
00:41:22,280 --> 00:41:23,880
staged, and waiting for days.

1018
00:41:23,880 --> 00:41:26,200
So supervisors spend time sorting paperwork,

1019
00:41:26,200 --> 00:41:28,040
checking material, calling planners,

1020
00:41:28,040 --> 00:41:29,480
and protecting local flow.

1021
00:41:29,480 --> 00:41:31,880
While the central plan loses contact with the work,

1022
00:41:31,880 --> 00:41:33,160
people can actually execute.

1023
00:41:33,160 --> 00:41:35,960
Release control takes a different approach,

1024
00:41:35,960 --> 00:41:37,880
instead of pushing every order into production

1025
00:41:37,880 --> 00:41:39,480
as soon as ERP creates it.

1026
00:41:39,480 --> 00:41:42,280
The plan releases work when it meets readiness conditions,

1027
00:41:42,280 --> 00:41:44,280
and when the constrained resource can use it

1028
00:41:44,280 --> 00:41:46,040
within the near term schedule.

1029
00:41:46,040 --> 00:41:48,200
That means material is complete and released,

1030
00:41:48,200 --> 00:41:49,160
the root is current,

1031
00:41:49,160 --> 00:41:50,760
and the bottleneck has capacity

1032
00:41:50,760 --> 00:41:52,120
within the agreed release window.

1033
00:41:52,120 --> 00:41:53,560
The point isn't to starve production,

1034
00:41:53,560 --> 00:41:56,280
it's to feed the bottleneck with work that can actually run.

1035
00:41:56,280 --> 00:41:59,400
When the bottleneck receives a manageable queue of ready orders,

1036
00:41:59,400 --> 00:42:01,240
the scheduler can sequence those orders

1037
00:42:01,240 --> 00:42:02,520
by due dates,

1038
00:42:02,520 --> 00:42:03,800
setup conditions,

1039
00:42:03,800 --> 00:42:05,640
and downstream needs.

1040
00:42:05,640 --> 00:42:07,560
Material control can focus on the jobs

1041
00:42:07,560 --> 00:42:09,160
that will actually run soon,

1042
00:42:09,160 --> 00:42:11,960
rather than building kits for work that may wait for weeks,

1043
00:42:11,960 --> 00:42:14,440
and planners get a clearer view of what's missing,

1044
00:42:14,440 --> 00:42:16,360
and what decision can remove the block.

1045
00:42:16,360 --> 00:42:18,440
This is where finite capacity scheduling starts

1046
00:42:18,440 --> 00:42:20,280
to become part of production control.

1047
00:42:20,280 --> 00:42:23,160
It doesn't sit off to the side as a weekly planning calculation.

1048
00:42:23,160 --> 00:42:25,240
It shapes which work reaches the floor,

1049
00:42:25,240 --> 00:42:26,680
when material gets staged,

1050
00:42:26,680 --> 00:42:28,680
and when exceptions need attention.

1051
00:42:28,680 --> 00:42:30,280
The schedule becomes a working agreement

1052
00:42:30,280 --> 00:42:31,160
between planning,

1053
00:42:31,160 --> 00:42:33,320
material control, production, and quality.

1054
00:42:33,320 --> 00:42:35,080
And even when the machine calendar looks open

1055
00:42:35,080 --> 00:42:36,520
and the material is fully ready,

1056
00:42:36,520 --> 00:42:38,120
capacity can still disappear.

1057
00:42:38,120 --> 00:42:40,280
People bring their own limits, skills,

1058
00:42:40,280 --> 00:42:42,160
shift patterns, and real world constraints

1059
00:42:42,160 --> 00:42:43,640
into the same schedule.

1060
00:42:43,640 --> 00:42:45,120
That's the subject for next time,

1061
00:42:45,120 --> 00:42:47,960
but for now, understand that the schedule isn't just a plan.

1062
00:42:47,960 --> 00:42:50,840
It's a conversation with the whole operation.

1063
00:42:50,840 --> 00:42:53,120
Labor, skills, and shift calendars.

1064
00:42:53,120 --> 00:42:55,480
Here's a scenario that trips up a lot of schedules.

1065
00:42:55,480 --> 00:42:57,280
Your machine calendar shows an open slot,

1066
00:42:57,280 --> 00:42:59,480
but real productive capacity sits at zero

1067
00:42:59,480 --> 00:43:02,080
because nobody qualified to run the operation is on shift,

1068
00:43:02,080 --> 00:43:04,360
or the person who can run it is stuck on another task.

1069
00:43:04,360 --> 00:43:06,120
The schedule never accounted for.

1070
00:43:06,120 --> 00:43:08,160
The machine exists, material might be ready,

1071
00:43:08,160 --> 00:43:09,360
and the order looks urgent,

1072
00:43:09,360 --> 00:43:12,080
but the work still cannot start safely or correctly.

1073
00:43:12,080 --> 00:43:13,440
That right there is one of the gaps

1074
00:43:13,440 --> 00:43:15,840
between an equipment schedule and a production schedule.

1075
00:43:15,840 --> 00:43:18,840
Take a CNC machine that can technically run both shifts.

1076
00:43:18,840 --> 00:43:21,040
On the day shift, an experienced operator sets it up,

1077
00:43:21,040 --> 00:43:23,920
runs the approved program, handles first piece checks,

1078
00:43:23,920 --> 00:43:26,920
and reacts when a tool or process issue appears.

1079
00:43:26,920 --> 00:43:29,560
On the evening shift, the machine can run repeat work,

1080
00:43:29,560 --> 00:43:31,680
but a new product setup needs a senior operator

1081
00:43:31,680 --> 00:43:34,160
or a setup technician who only works days.

1082
00:43:34,160 --> 00:43:36,760
Two different capacity profiles, plain and simple.

1083
00:43:36,760 --> 00:43:38,920
The calendar might show 16 machine hours,

1084
00:43:38,920 --> 00:43:41,160
but for a certain product, the plant may only have eight

1085
00:43:41,160 --> 00:43:43,560
workable hours because the required skill is present

1086
00:43:43,560 --> 00:43:44,760
for one shift.

1087
00:43:44,760 --> 00:43:46,720
Load that order across the full day,

1088
00:43:46,720 --> 00:43:50,280
and you have created a conflict before production even sees it.

1089
00:43:50,280 --> 00:43:53,120
That is the kind of mismatch that feels invisible in the system,

1090
00:43:53,120 --> 00:43:55,160
but stops work cold on the floor.

1091
00:43:55,160 --> 00:43:57,600
Shift patterns matter here, and they are never simple.

1092
00:43:57,600 --> 00:44:00,480
A plant may work two shifts, three shifts, weekends,

1093
00:44:00,480 --> 00:44:02,880
or a pattern that flexes with demand.

1094
00:44:02,880 --> 00:44:05,080
Within those shifts, people take breaks,

1095
00:44:05,080 --> 00:44:07,360
attend handovers, join safety meetings,

1096
00:44:07,360 --> 00:44:10,680
support quality checks, and sometimes cover another work center.

1097
00:44:10,680 --> 00:44:12,480
None of that makes the plant poorly run.

1098
00:44:12,480 --> 00:44:14,160
It's simply how work happens.

1099
00:44:14,160 --> 00:44:16,560
Overtime adds capacity, but it's not a blank line

1100
00:44:16,560 --> 00:44:17,800
in a planning system.

1101
00:44:17,800 --> 00:44:20,600
There are rules on who can work it, how much is allowed,

1102
00:44:20,600 --> 00:44:22,560
and whether the right skill is available.

1103
00:44:22,560 --> 00:44:25,440
Someone may agree to stay late today, but not tomorrow.

1104
00:44:25,440 --> 00:44:27,480
Another person may already have worked a long shift

1105
00:44:27,480 --> 00:44:30,000
and pushing more work onto them can create safety,

1106
00:44:30,000 --> 00:44:31,560
quality and retention problems

1107
00:44:31,560 --> 00:44:34,200
that never show up in a simple hour calculation.

1108
00:44:34,200 --> 00:44:36,160
Absence changes the picture fast.

1109
00:44:36,160 --> 00:44:39,160
One operator calling in sick can affect a whole resource group

1110
00:44:39,160 --> 00:44:41,320
if that person holds a rare approval

1111
00:44:41,320 --> 00:44:44,360
or carries the practical knowledge for a difficult setup.

1112
00:44:44,360 --> 00:44:46,360
Training creates a related limit.

1113
00:44:46,360 --> 00:44:48,960
A newer operator may run the machine under supervision,

1114
00:44:48,960 --> 00:44:50,440
run only certain products,

1115
00:44:50,440 --> 00:44:53,600
or need a senior colleague nearby for the first few jobs.

1116
00:44:53,600 --> 00:44:56,680
The schedule should treat those limits as facts, not weakness.

1117
00:44:56,680 --> 00:44:59,440
And in practical terms, that means building them into the model.

1118
00:44:59,440 --> 00:45:02,160
That is where a skill matrix becomes useful.

1119
00:45:02,160 --> 00:45:04,160
It records who can perform which work,

1120
00:45:04,160 --> 00:45:05,720
on which machine or process,

1121
00:45:05,720 --> 00:45:07,560
and sometimes at what level of approval.

1122
00:45:07,560 --> 00:45:10,240
It may show that one operator can run standard milling jobs,

1123
00:45:10,240 --> 00:45:11,920
another can perform high tolerance work,

1124
00:45:11,920 --> 00:45:13,760
and a third can complete the inspection step

1125
00:45:13,760 --> 00:45:15,520
that releases the part to assembly.

1126
00:45:15,520 --> 00:45:18,080
This gives the scheduler another set of conditions.

1127
00:45:18,080 --> 00:45:20,680
An operation may require a machine and a skill,

1128
00:45:20,680 --> 00:45:23,160
and the order can only occupy a time slot

1129
00:45:23,160 --> 00:45:24,840
when both exist together.

1130
00:45:24,840 --> 00:45:26,520
If the operator works from six in the morning

1131
00:45:26,520 --> 00:45:27,560
until two in the afternoon,

1132
00:45:27,560 --> 00:45:30,000
the machine slot at four o'clock may not count for that job,

1133
00:45:30,000 --> 00:45:32,520
even though the machine itself remains open.

1134
00:45:32,520 --> 00:45:34,080
Now here's where it gets practical.

1135
00:45:34,080 --> 00:45:36,400
Some operations need more than one person.

1136
00:45:36,400 --> 00:45:38,600
A complex setup may require the machine operator

1137
00:45:38,600 --> 00:45:40,440
and a setup technician at the same time.

1138
00:45:40,440 --> 00:45:42,960
A heavier awkward component may need a material handler

1139
00:45:42,960 --> 00:45:43,880
during loading,

1140
00:45:43,880 --> 00:45:46,360
and a first article may need a quality specialist available

1141
00:45:46,360 --> 00:45:48,840
to review the part before the run continues.

1142
00:45:48,840 --> 00:45:51,240
These pairing rules often live in people's heads.

1143
00:45:51,240 --> 00:45:53,360
The supervisor knows that the job cannot start

1144
00:45:53,360 --> 00:45:55,760
until the technician returns from another area.

1145
00:45:55,760 --> 00:45:58,400
A planner knows that quality gets busy at certain times,

1146
00:45:58,400 --> 00:46:00,160
and the operator knows that a particular part

1147
00:46:00,160 --> 00:46:01,920
takes two people to load safely,

1148
00:46:01,920 --> 00:46:04,440
despite what the routing time suggests.

1149
00:46:04,440 --> 00:46:06,080
Finite scheduling becomes more credible

1150
00:46:06,080 --> 00:46:08,720
when these rules move into the model in a practical form.

1151
00:46:08,720 --> 00:46:11,600
Not every human detail belongs in an automated schedule.

1152
00:46:11,600 --> 00:46:13,240
You do not need to turn the workforce

1153
00:46:13,240 --> 00:46:15,480
into a spreadsheet of every small action.

1154
00:46:15,480 --> 00:46:17,960
But if a skill, role, or approval repeatedly

1155
00:46:17,960 --> 00:46:19,800
decides whether work can run,

1156
00:46:19,800 --> 00:46:21,560
it belongs in the planning logic.

1157
00:46:21,560 --> 00:46:23,200
People are not virtual computer resources,

1158
00:46:23,200 --> 00:46:25,880
that you cannot move an experienced operator from one work center

1159
00:46:25,880 --> 00:46:28,520
to another with a mouse click and assume the output follows.

1160
00:46:28,520 --> 00:46:30,040
They need time to walk there,

1161
00:46:30,040 --> 00:46:33,120
understand the current state of the job, take over safely,

1162
00:46:33,120 --> 00:46:36,320
and sometimes learn a process they have not run in months.

1163
00:46:36,320 --> 00:46:38,440
Even where a person holds the formal skill,

1164
00:46:38,440 --> 00:46:42,000
the supervisor may know they are already caring too much work.

1165
00:46:42,000 --> 00:46:43,320
That judgment still matters.

1166
00:46:43,320 --> 00:46:45,280
A finite schedule should support the supervisor,

1167
00:46:45,280 --> 00:46:47,720
not issue a fantasy instruction from an office system.

1168
00:46:47,720 --> 00:46:50,120
It can show that the bottleneck needs a qualified person

1169
00:46:50,120 --> 00:46:51,400
at a specific time,

1170
00:46:51,400 --> 00:46:54,800
and reveal that a staffing gap turns an apparently open machine

1171
00:46:54,800 --> 00:46:56,600
into unavailable capacity.

1172
00:46:56,600 --> 00:46:59,040
Then production can decide whether to change the sequence,

1173
00:46:59,040 --> 00:47:02,160
move labor, approve overtime, or accept the date risk.

1174
00:47:02,160 --> 00:47:04,240
People are only one part of the resource picture,

1175
00:47:04,240 --> 00:47:06,520
and the next layer includes the physical items

1176
00:47:06,520 --> 00:47:08,480
and planned interruptions that can block work

1177
00:47:08,480 --> 00:47:11,800
even when the machine and operator are both ready.

1178
00:47:11,800 --> 00:47:14,400
Tools, fixtures, maintenance, and quality gates.

1179
00:47:14,400 --> 00:47:16,920
A machine can be free, an operator can be ready,

1180
00:47:16,920 --> 00:47:18,800
and the order can still have nowhere to go

1181
00:47:18,800 --> 00:47:22,400
because machines are rarely the only finite resource in a process.

1182
00:47:22,400 --> 00:47:24,400
One fixture set may exist only once,

1183
00:47:24,400 --> 00:47:26,240
a mold may already sit in another press,

1184
00:47:26,240 --> 00:47:28,000
a gauge may wait for calibration,

1185
00:47:28,000 --> 00:47:30,560
and a test rig can support several product lines.

1186
00:47:30,560 --> 00:47:32,640
A shared crane may become the real constraint

1187
00:47:32,640 --> 00:47:34,800
long before the machine calendar fills up.

1188
00:47:34,800 --> 00:47:37,680
The schedule needs to see those resources as occupied time,

1189
00:47:37,680 --> 00:47:39,840
not as a note, in a work instruction.

1190
00:47:39,840 --> 00:47:43,040
Take a machining operation that needs a specific fixture.

1191
00:47:43,040 --> 00:47:46,320
The CNC machine may have an open slot from nine until noon,

1192
00:47:46,320 --> 00:47:48,640
and the right operator may work that shift,

1193
00:47:48,640 --> 00:47:51,680
but if the fixture remains on another machine until 11,

1194
00:47:51,680 --> 00:47:53,760
that operation cannot begin at nine.

1195
00:47:53,760 --> 00:47:57,080
That machine is available, but the full production resource is not.

1196
00:47:57,080 --> 00:47:59,120
The same applies to cutting tools.

1197
00:47:59,120 --> 00:48:01,280
A tool package might suit several jobs,

1198
00:48:01,280 --> 00:48:04,400
but one set can only sit in one machine at a time,

1199
00:48:04,400 --> 00:48:06,520
and tool life can matter too.

1200
00:48:06,520 --> 00:48:08,320
If a tool has limited life left,

1201
00:48:08,320 --> 00:48:10,560
the plant may need to replace it during the run,

1202
00:48:10,560 --> 00:48:13,480
prepare another tool set or reserve time for the change.

1203
00:48:13,480 --> 00:48:15,440
If the schedule ignores that condition,

1204
00:48:15,440 --> 00:48:18,160
it can look fine until the operator reaches the machine

1205
00:48:18,160 --> 00:48:20,560
and discovers that the required tooling is elsewhere,

1206
00:48:20,560 --> 00:48:22,160
expired or not ready.

1207
00:48:22,160 --> 00:48:24,880
Molds and test rigs create a similar issue.

1208
00:48:24,880 --> 00:48:27,400
In molding, a press without the required mold

1209
00:48:27,400 --> 00:48:28,640
cannot produce the part.

1210
00:48:28,640 --> 00:48:30,720
In assembly and test, a finished unit may wait

1211
00:48:30,720 --> 00:48:33,080
because the only test rig is busy with another product,

1212
00:48:33,080 --> 00:48:35,200
and the routing may show assembly capacity

1213
00:48:35,200 --> 00:48:38,000
while the actual flow stops at the shared test asset.

1214
00:48:38,000 --> 00:48:40,960
Shared cranes can create some of the most confusing failures.

1215
00:48:40,960 --> 00:48:43,440
Several machines may run independently most of the time,

1216
00:48:43,440 --> 00:48:46,360
but each one needs the same crane for loading, unloading,

1217
00:48:46,360 --> 00:48:47,520
or a tool change.

1218
00:48:47,520 --> 00:48:49,600
If two jobs need that crane at the same time,

1219
00:48:49,600 --> 00:48:52,800
one job waits, and a schedule that only looks at machine capacity

1220
00:48:52,800 --> 00:48:55,440
sees no conflict while the shift sees two teams

1221
00:48:55,440 --> 00:48:57,280
waiting for the same hook.

1222
00:48:57,280 --> 00:48:59,640
This is where resource groups become useful.

1223
00:48:59,640 --> 00:49:02,200
They describe the set of things that must exist together

1224
00:49:02,200 --> 00:49:03,600
for an operation to run.

1225
00:49:03,600 --> 00:49:07,000
It may include one machine, one fixture, one qualified person,

1226
00:49:07,000 --> 00:49:08,840
and a gauge for first piece approval,

1227
00:49:08,840 --> 00:49:10,560
and the scheduler needs to find a time

1228
00:49:10,560 --> 00:49:13,920
when that full group is available at once, not eventually.

1229
00:49:13,920 --> 00:49:16,760
Consider a complex operation where a machine needs a fixture,

1230
00:49:16,760 --> 00:49:19,800
a trained operator, and a quality inspector for the first article.

1231
00:49:19,800 --> 00:49:21,360
The machine may have four, three hours,

1232
00:49:21,360 --> 00:49:22,640
the fixture may be free all day,

1233
00:49:22,640 --> 00:49:24,640
the operator may be available in the morning,

1234
00:49:24,640 --> 00:49:27,400
and quality may only have capacity in the afternoon.

1235
00:49:27,400 --> 00:49:30,120
So the actual start time follows the last required condition,

1236
00:49:30,120 --> 00:49:32,200
not the first open calendar slot.

1237
00:49:32,200 --> 00:49:34,240
That can feel like extra complexity,

1238
00:49:34,240 --> 00:49:36,880
but production already works this way.

1239
00:49:36,880 --> 00:49:40,880
The difference is whether the model sees it before the shift begins

1240
00:49:40,880 --> 00:49:43,440
or whether people discover it through phone calls,

1241
00:49:43,440 --> 00:49:45,800
workarounds, and a machine standing still

1242
00:49:45,800 --> 00:49:48,000
within order beside it.

1243
00:49:48,000 --> 00:49:50,920
Maintenance introduces another form of finite capacity.

1244
00:49:50,920 --> 00:49:53,000
Preventive maintenance needs planned windows

1245
00:49:53,000 --> 00:49:54,320
in the resource calendar.

1246
00:49:54,320 --> 00:49:56,120
If a machine needs a service check,

1247
00:49:56,120 --> 00:49:59,160
lubrication inspection or scheduled replacement of a ware part,

1248
00:49:59,160 --> 00:50:01,000
that time belongs in the plan.

1249
00:50:01,000 --> 00:50:02,640
Treating maintenance is something that happens

1250
00:50:02,640 --> 00:50:04,680
after production finishes, often means

1251
00:50:04,680 --> 00:50:07,760
it never finds a safe place to happen until the asset fails.

1252
00:50:07,760 --> 00:50:09,920
A finite schedule should reserve maintenance time

1253
00:50:09,920 --> 00:50:11,520
the same way it reserves production time,

1254
00:50:11,520 --> 00:50:14,440
but condition-based downtime needs a slightly different approach.

1255
00:50:14,440 --> 00:50:16,800
Maintenance may receive a warning from an inspection,

1256
00:50:16,800 --> 00:50:20,400
a vibration trend, an alarm pattern, or an operator report,

1257
00:50:20,400 --> 00:50:23,800
and nobody can always predict the exact duration of the intervention.

1258
00:50:23,800 --> 00:50:26,000
The scheduler should not pretend it can,

1259
00:50:26,000 --> 00:50:28,160
but it can mark the resource as restricted,

1260
00:50:28,160 --> 00:50:30,640
reduce available capacity or present choices,

1261
00:50:30,640 --> 00:50:32,880
run the order and accept equipment risk,

1262
00:50:32,880 --> 00:50:35,160
stop the machine now, move selected work,

1263
00:50:35,160 --> 00:50:37,640
or delay the maintenance within an agreed limit.

1264
00:50:37,640 --> 00:50:40,040
Those choices belong with maintenance and production,

1265
00:50:40,040 --> 00:50:43,320
not with the calendar pretending the asset has no condition.

1266
00:50:43,320 --> 00:50:45,600
Quality gates also consume capacity.

1267
00:50:45,600 --> 00:50:47,600
A gauge needs a current calibration status

1268
00:50:47,600 --> 00:50:49,240
before it can support a measurement.

1269
00:50:49,240 --> 00:50:50,840
A test rig may need verification,

1270
00:50:50,840 --> 00:50:53,760
and an inspector can only review so many first pieces, samples,

1271
00:50:53,760 --> 00:50:55,360
or release decisions in a shift.

1272
00:50:55,360 --> 00:50:57,160
If inspection capacity becomes overloaded,

1273
00:50:57,160 --> 00:50:59,280
completed parts can stack up, waiting for release,

1274
00:50:59,280 --> 00:51:01,040
while the next operation runs short of work.

1275
00:51:01,040 --> 00:51:03,200
It changes flow, not just quality.

1276
00:51:03,200 --> 00:51:05,840
A schedule can show every machine free and still fail

1277
00:51:05,840 --> 00:51:08,600
because the required fixture, crane, test rig,

1278
00:51:08,600 --> 00:51:10,400
inspector, or calibrated gauge

1279
00:51:10,400 --> 00:51:12,760
cannot join the operation at the right time.

1280
00:51:12,760 --> 00:51:15,520
Finite capacity scheduling works when it models the resources

1281
00:51:15,520 --> 00:51:17,600
that repeatedly decide whether work starts,

1282
00:51:17,600 --> 00:51:19,800
not only the resources with the biggest motors.

1283
00:51:19,800 --> 00:51:22,480
Once work does start, the schedule meets another source

1284
00:51:22,480 --> 00:51:23,640
of uncertainty.

1285
00:51:23,640 --> 00:51:25,600
Plant, duration, and actual performance

1286
00:51:25,600 --> 00:51:28,200
are rarely identical on the factory floor.

1287
00:51:28,200 --> 00:51:30,640
Standards, actuals, and schedule credibility.

1288
00:51:30,640 --> 00:51:32,920
Every schedule starts with a planning assumption.

1289
00:51:32,920 --> 00:51:35,840
Arrouting might say an operation takes six minutes per,

1290
00:51:35,840 --> 00:51:37,720
part plus 40 minutes for setup,

1291
00:51:37,720 --> 00:51:39,800
and those numbers let the scheduler put something

1292
00:51:39,800 --> 00:51:40,720
on the calendar.

1293
00:51:40,720 --> 00:51:41,680
But here's the thing,

1294
00:51:41,680 --> 00:51:43,840
they aren't permanent facts about the process.

1295
00:51:43,840 --> 00:51:46,760
They describe what the plant expects under specific conditions

1296
00:51:46,760 --> 00:51:48,480
and conditions change all the time.

1297
00:51:48,480 --> 00:51:51,000
A worn tool might slow the machine down.

1298
00:51:51,000 --> 00:51:53,880
A new operator needs extra time on a product they don't run often.

1299
00:51:53,880 --> 00:51:56,360
Material can vary enough to throw off cycle time

1300
00:51:56,360 --> 00:51:58,240
and the first part might pass one day,

1301
00:51:58,240 --> 00:52:00,640
then need adjustment and another inspection the next.

1302
00:52:00,640 --> 00:52:02,920
None of that is unusual, but it's already more variation

1303
00:52:02,920 --> 00:52:05,080
than the planning assumption captured.

1304
00:52:05,080 --> 00:52:07,080
Now, standard cycle time still matters.

1305
00:52:07,080 --> 00:52:10,040
Without it, you can't plan capacity in a repeatable way,

1306
00:52:10,040 --> 00:52:13,320
can't compare performance, can't spot when the process has shifted.

1307
00:52:13,320 --> 00:52:15,800
But you have to treat that standard as a controlled estimate,

1308
00:52:15,800 --> 00:52:17,720
not a promise carved in stone.

1309
00:52:17,720 --> 00:52:20,720
Actual cycle time tells you what the operation consumed on the floor,

1310
00:52:20,720 --> 00:52:22,280
and that's where the real number lives.

1311
00:52:22,280 --> 00:52:24,240
That difference between standard and actual

1312
00:52:24,240 --> 00:52:26,280
can come from normal variation, sure.

1313
00:52:26,280 --> 00:52:28,520
It can also come from scrap, rework, short stops,

1314
00:52:28,520 --> 00:52:30,360
tool adjustments, material handling delays,

1315
00:52:30,360 --> 00:52:32,400
or machine that pauses enough to disrupt flow

1316
00:52:32,400 --> 00:52:33,960
without anyone calling it a breakdown.

1317
00:52:33,960 --> 00:52:36,680
Those microstops eat into capacity across a shift,

1318
00:52:36,680 --> 00:52:39,240
even when the machine never registers as down.

1319
00:52:39,240 --> 00:52:41,280
Scrap changes the schedule in a different way.

1320
00:52:41,280 --> 00:52:44,560
If an order needs 100 good parts and 5 fail inspection,

1321
00:52:44,560 --> 00:52:47,360
the remaining work doesn't follow the original plan anymore.

1322
00:52:47,360 --> 00:52:49,480
The plant may need to run extra quantity,

1323
00:52:49,480 --> 00:52:52,480
use rework capacity, or adjust the material balance.

1324
00:52:52,480 --> 00:52:54,480
A downstream operation might wait for good parts

1325
00:52:54,480 --> 00:52:57,400
while the machine still shows it completed the planned quantity.

1326
00:52:57,400 --> 00:53:00,240
Completion and usable output aren't always the same thing.

1327
00:53:00,240 --> 00:53:01,800
Rework needs the same attention,

1328
00:53:01,800 --> 00:53:04,120
a part that returns to machining or inspection

1329
00:53:04,120 --> 00:53:06,040
consumes capacity all over again

1330
00:53:06,040 --> 00:53:08,520
and competes with fresh orders for the same resource.

1331
00:53:08,520 --> 00:53:10,920
If the schedule pretends rework doesn't exist

1332
00:53:10,920 --> 00:53:12,840
because it's outside the normal route,

1333
00:53:12,840 --> 00:53:16,120
the plan gets more optimistic every time quality finds a problem.

1334
00:53:16,120 --> 00:53:18,840
That optimism shows up later as a misdelivery.

1335
00:53:18,840 --> 00:53:22,080
So this is where shop floor actuals earn their place in scheduling.

1336
00:53:22,080 --> 00:53:24,720
A manufacturing execution system, or MES,

1337
00:53:24,720 --> 00:53:27,800
can record actual starts, finishes, quantities, scrap,

1338
00:53:27,800 --> 00:53:29,760
downtime reasons, and labor activity,

1339
00:53:29,760 --> 00:53:31,200
depending on how the plant runs.

1340
00:53:31,200 --> 00:53:34,200
That information shouldn't just feed a report at month end.

1341
00:53:34,200 --> 00:53:35,920
It needs to inform the planning view

1342
00:53:35,920 --> 00:53:38,880
while there's still time to make a better decision right now.

1343
00:53:38,880 --> 00:53:40,640
But I'd be careful with automatic reactions.

1344
00:53:40,640 --> 00:53:42,880
Take overall equipment effectiveness or OE,

1345
00:53:42,880 --> 00:53:45,200
it brings together availability, performance, and quality

1346
00:53:45,200 --> 00:53:46,000
into one number.

1347
00:53:46,000 --> 00:53:48,600
It helps you see whether a resource loses time from stops,

1348
00:53:48,600 --> 00:53:50,880
slower running, or rejected output.

1349
00:53:50,880 --> 00:53:52,720
It can reveal patterns worth attention,

1350
00:53:52,720 --> 00:53:54,400
but it doesn't automatically tell the scheduler

1351
00:53:54,400 --> 00:53:57,480
to cut every rooting time by the current OE percentage.

1352
00:53:57,480 --> 00:54:00,480
An OE drop might come from a single unusual issue,

1353
00:54:00,480 --> 00:54:03,280
a tool failure, a new job that needed extra setup,

1354
00:54:03,280 --> 00:54:04,840
an operator training a colleague,

1355
00:54:04,840 --> 00:54:07,280
applying that one result to all future schedules

1356
00:54:07,280 --> 00:54:09,960
turns a temporary event into a permanent capacity cut.

1357
00:54:09,960 --> 00:54:11,320
The opposite problem happens too.

1358
00:54:11,320 --> 00:54:13,720
A few good shifts show faster than standard performance,

1359
00:54:13,720 --> 00:54:15,720
but that might rely on an experienced operator,

1360
00:54:15,720 --> 00:54:18,320
easy material, or a sequence that avoided tough setups.

1361
00:54:18,320 --> 00:54:21,160
Changing the rooting standard that quickly creates a plan

1362
00:54:21,160 --> 00:54:23,000
that only works on its best days.

1363
00:54:23,000 --> 00:54:24,960
Planning needs judgment about the pattern.

1364
00:54:24,960 --> 00:54:27,920
Temporary bad performance should trigger an operational response.

1365
00:54:27,920 --> 00:54:30,480
Check the cause, see if the problem remains active,

1366
00:54:30,480 --> 00:54:33,880
adjust the near term schedule if capacity has genuinely changed.

1367
00:54:33,880 --> 00:54:36,440
A change process baseline needs a different response.

1368
00:54:36,440 --> 00:54:39,320
Engineering and production can review repeated actuals,

1369
00:54:39,320 --> 00:54:41,440
confirm the method or condition has shifted

1370
00:54:41,440 --> 00:54:42,840
and update the standard.

1371
00:54:42,840 --> 00:54:44,840
That distinction protects schedule credibility.

1372
00:54:44,840 --> 00:54:46,360
If actuals never feedback,

1373
00:54:46,360 --> 00:54:47,920
the schedule keeps repeating assumptions

1374
00:54:47,920 --> 00:54:49,680
production has already disproved.

1375
00:54:49,680 --> 00:54:51,760
If every variation changes the standard,

1376
00:54:51,760 --> 00:54:54,720
the plan becomes unstable and nobody knows which number to trust.

1377
00:54:54,720 --> 00:54:56,480
You need a middle ground that uses the data

1378
00:54:56,480 --> 00:54:58,160
without reacting to every blip.

1379
00:54:58,160 --> 00:55:00,040
The feedback loop needs clear roles.

1380
00:55:00,040 --> 00:55:01,760
M.S. supplies execution facts.

1381
00:55:01,760 --> 00:55:03,920
Engineering owns the root method and standard time

1382
00:55:03,920 --> 00:55:05,720
once a change has been verified.

1383
00:55:05,720 --> 00:55:07,320
Planning uses the approved standards

1384
00:55:07,320 --> 00:55:09,400
and current execution state to place work.

1385
00:55:09,400 --> 00:55:12,520
Production adds the context that raw time stamps can't capture.

1386
00:55:12,520 --> 00:55:14,080
A delay code doesn't always explain

1387
00:55:14,080 --> 00:55:16,160
whether the issue will repeat tomorrow.

1388
00:55:16,160 --> 00:55:18,520
That loop doesn't need to turn into slow committee work.

1389
00:55:18,520 --> 00:55:20,720
It just needs a regular way to ask a simple question.

1390
00:55:20,720 --> 00:55:23,040
Did this event change today's schedule only?

1391
00:55:23,040 --> 00:55:25,880
Or did it change our expected capacity going forward?

1392
00:55:25,880 --> 00:55:27,960
Once actual production reaches the schedule,

1393
00:55:27,960 --> 00:55:30,200
another event always waits in the background.

1394
00:55:30,200 --> 00:55:33,440
The machine that stops when the plan has no spare time left.

1395
00:55:33,440 --> 00:55:34,720
The machine breakdown.

1396
00:55:34,720 --> 00:55:36,280
Now put that schedule under pressure.

1397
00:55:36,280 --> 00:55:38,640
The bottleneck machine is halfway through a loaded shift

1398
00:55:38,640 --> 00:55:40,720
running an order that feeds assembly later that day

1399
00:55:40,720 --> 00:55:43,160
with the next two orders already waiting in sequence.

1400
00:55:43,160 --> 00:55:45,760
Material is staged, the operator knows the plan

1401
00:55:45,760 --> 00:55:48,720
and downstream teams expect parts at specific times.

1402
00:55:48,720 --> 00:55:50,840
Then the machine stops.

1403
00:55:50,840 --> 00:55:53,680
Maybe the spindle alarms, a hydraulic fault appears

1404
00:55:53,680 --> 00:55:55,800
or a tool changer jams.

1405
00:55:55,800 --> 00:55:57,120
The course doesn't matter yet.

1406
00:55:57,120 --> 00:55:58,400
The machine cannot continue

1407
00:55:58,400 --> 00:56:00,640
and an operation that looked safely placed now

1408
00:56:00,640 --> 00:56:02,240
has an unknown finish time.

1409
00:56:02,240 --> 00:56:03,960
The first impact is easy to see.

1410
00:56:03,960 --> 00:56:05,840
The current order remains unfinished.

1411
00:56:05,840 --> 00:56:07,480
But the effect moves quickly.

1412
00:56:07,480 --> 00:56:10,160
Parts needed by the next operation don't arrive.

1413
00:56:10,160 --> 00:56:12,360
Assembly may have people ready, components prepared

1414
00:56:12,360 --> 00:56:13,480
and a work slot reserved.

1415
00:56:13,480 --> 00:56:16,240
Inspection may expect a batch that never reaches the queue.

1416
00:56:16,240 --> 00:56:18,240
Material handlers may wait with the next kit

1417
00:56:18,240 --> 00:56:21,200
while the bottleneck holds both machine and schedule in place.

1418
00:56:21,200 --> 00:56:23,000
A breakdown changes more than one order.

1419
00:56:23,000 --> 00:56:26,120
Suppose the order on the machine needs another two hours of run time.

1420
00:56:26,120 --> 00:56:28,720
The maintenance technician arrives and starts diagnosis.

1421
00:56:28,720 --> 00:56:31,120
At that point, the plan can't honestly assume

1422
00:56:31,120 --> 00:56:34,040
the machine will return after a fixed number of minutes

1423
00:56:34,040 --> 00:56:35,640
just because the schedule needs it to.

1424
00:56:35,640 --> 00:56:36,440
You know it won't.

1425
00:56:36,440 --> 00:56:38,120
Repair duration is uncertain.

1426
00:56:38,120 --> 00:56:39,680
A fault may clear after a reset

1427
00:56:39,680 --> 00:56:42,360
or it may require a spare part from the maintenance store

1428
00:56:42,360 --> 00:56:45,400
or it may expose damage that needs outside support.

1429
00:56:45,400 --> 00:56:48,280
Production needs a planning view before all that becomes certain

1430
00:56:48,280 --> 00:56:51,160
but that planning view should show the assumption it uses.

1431
00:56:51,160 --> 00:56:52,840
So think about three repair scenarios,

1432
00:56:52,840 --> 00:56:55,200
a short interruption if the fault clears quickly,

1433
00:56:55,200 --> 00:56:58,320
a longer outage if the repair takes the rest of the shift

1434
00:56:58,320 --> 00:57:02,200
or a worst case where the resource stays unavailable into the next day.

1435
00:57:02,200 --> 00:57:04,400
The point isn't to create drama around every alarm.

1436
00:57:04,400 --> 00:57:08,160
It's to stop treating uncertain downtime as a promise that's already been kept.

1437
00:57:08,160 --> 00:57:10,560
People need to decide while the facts are still moving.

1438
00:57:10,560 --> 00:57:13,680
Can maintenance restore the machine safely to complete the order?

1439
00:57:13,680 --> 00:57:16,240
Can part of the work move to an approved alternate resource

1440
00:57:16,240 --> 00:57:18,840
that has capacity, tooling and a qualified operator?

1441
00:57:18,840 --> 00:57:20,280
Could the plan split the batch,

1442
00:57:20,280 --> 00:57:23,120
send the completed parts ahead while the remainder waits?

1443
00:57:23,120 --> 00:57:24,480
Each question carries a cost.

1444
00:57:24,480 --> 00:57:27,680
Re-routing may protect the delivery but add setup time elsewhere.

1445
00:57:27,680 --> 00:57:29,360
Over time may recover capacity,

1446
00:57:29,360 --> 00:57:30,960
but only if the right people can work it

1447
00:57:30,960 --> 00:57:32,960
and the machine returns in usable condition.

1448
00:57:32,960 --> 00:57:35,040
Subcontracting may help with a later operation,

1449
00:57:35,040 --> 00:57:38,800
but it introduces transport time, quality checks and commercial approval.

1450
00:57:38,800 --> 00:57:41,280
Sometimes the clearest action is a customer call

1451
00:57:41,280 --> 00:57:42,640
that doesn't mean the plant has given up.

1452
00:57:42,640 --> 00:57:46,320
It means customer service can speak from a realistic production position

1453
00:57:46,320 --> 00:57:50,240
instead of repeating the original date after the conditions that supported it are gone.

1454
00:57:50,240 --> 00:57:53,040
If a service part will miss the agreed shipment time,

1455
00:57:53,040 --> 00:57:56,480
an early conversation creates options that a late surprise removes.

1456
00:57:56,480 --> 00:57:57,920
The same applies to priority.

1457
00:57:57,920 --> 00:58:01,200
Maybe the interrupted order matters less than the next job in the queue,

1458
00:58:01,200 --> 00:58:04,400
maybe completing the current work first protects traceability,

1459
00:58:04,400 --> 00:58:07,440
avoids scrap or keeps a later assembly cell supplied,

1460
00:58:07,440 --> 00:58:09,760
or perhaps the unfinished order can wait

1461
00:58:09,760 --> 00:58:12,880
because another order carries a more serious customer risk.

1462
00:58:12,880 --> 00:58:15,360
The original schedule can't answer those questions anymore.

1463
00:58:15,360 --> 00:58:17,040
It answered a different question.

1464
00:58:17,040 --> 00:58:19,920
Given the capacity and constraints known when the plan ran,

1465
00:58:19,920 --> 00:58:21,520
where should the work go?

1466
00:58:21,520 --> 00:58:23,520
The breakdown has changed available capacity,

1467
00:58:23,520 --> 00:58:25,040
the state of the current operation,

1468
00:58:25,040 --> 00:58:27,440
and possibly the priority of the next decision.

1469
00:58:27,440 --> 00:58:30,240
Keeping the old schedule on the board doesn't preserve control.

1470
00:58:30,240 --> 00:58:33,680
It only preserves an old assumption that no longer applies.

1471
00:58:33,680 --> 00:58:36,320
This is where finite scheduling becomes useful in the moment,

1472
00:58:36,320 --> 00:58:38,000
not just before the shift starts.

1473
00:58:38,000 --> 00:58:41,280
The planner needs to know which orders now miss their planned finish,

1474
00:58:41,280 --> 00:58:43,360
what downstream work loses supply,

1475
00:58:43,360 --> 00:58:45,360
which alternatives remain feasible,

1476
00:58:45,360 --> 00:58:47,520
and what each alternative displaces.

1477
00:58:47,520 --> 00:58:50,880
Maintenance needs room to state what it knows and what it doesn't know.

1478
00:58:50,880 --> 00:58:54,080
Production needs a plan that recognizes the work already in motion.

1479
00:58:54,080 --> 00:58:55,760
No schedule removes the disruption,

1480
00:58:55,760 --> 00:58:57,840
but a schedule can make the choices visible

1481
00:58:57,840 --> 00:59:01,680
before the disruption turns into a row of missed dates and phone calls.

1482
00:59:01,680 --> 00:59:04,800
The next step is to recalculate the plan without treating every order

1483
00:59:04,800 --> 00:59:06,400
as if it can move freely,

1484
00:59:06,400 --> 00:59:08,400
rescheduling without creating chaos.

1485
00:59:08,400 --> 00:59:10,480
Here's a problem most schedulers don't talk about.

1486
00:59:10,480 --> 00:59:11,760
When a bottleneck stops,

1487
00:59:11,760 --> 00:59:13,920
you don't need to tear down the entire factory plan,

1488
00:59:13,920 --> 00:59:16,400
the reschedule should start from the event and nothing more.

1489
00:59:16,400 --> 00:59:17,760
It might be a machine breakdown,

1490
00:59:17,760 --> 00:59:21,120
a late delivery leaving an order short or material equality hold,

1491
00:59:21,120 --> 00:59:22,800
blocking a batch after inspection,

1492
00:59:22,800 --> 00:59:24,960
or a priority change from the commercial side.

1493
00:59:24,960 --> 00:59:26,960
In each case, something in the model shifted,

1494
00:59:26,960 --> 00:59:29,760
and the scheduler needs to test the effect from that point forward,

1495
00:59:29,760 --> 00:59:31,600
not rebuild everything from scratch.

1496
00:59:31,600 --> 00:59:32,560
That sounds simple,

1497
00:59:32,560 --> 00:59:33,920
but the real danger appears

1498
00:59:33,920 --> 00:59:36,720
when every little exception triggers a complete reshuffle.

1499
00:59:36,720 --> 00:59:38,480
Picture a machine losing four hours.

1500
00:59:38,480 --> 00:59:41,040
The scheduling system responds by moving every single order

1501
00:59:41,040 --> 00:59:42,320
across the next two weeks.

1502
00:59:42,320 --> 00:59:44,320
Dates change, sequences change,

1503
00:59:44,320 --> 00:59:45,920
material teams get a new list,

1504
00:59:45,920 --> 00:59:50,400
and supervisors start the shift with one plan only to receive another before lunch.

1505
00:59:50,400 --> 00:59:51,520
Then a second event hits,

1506
00:59:51,520 --> 00:59:53,360
and the whole plan moves yet again.

1507
00:59:53,360 --> 00:59:55,120
You know what happens next.

1508
00:59:55,120 --> 00:59:56,640
People stop trusting it.

1509
00:59:56,640 --> 00:59:59,680
A schedule that dances around all the time might be technically accurate,

1510
00:59:59,680 --> 01:00:02,240
but it doesn't give production a stable basis for action.

1511
01:00:02,240 --> 01:00:04,720
The floor needs enough consistency to stage material,

1512
01:00:04,720 --> 01:00:06,160
prepare tools, assign work,

1513
01:00:06,160 --> 01:00:07,680
and start the shift with some confidence,

1514
01:00:07,680 --> 01:00:09,840
which is exactly why freezes zones matter.

1515
01:00:09,840 --> 01:00:11,920
A free zone is a period close to execution

1516
01:00:11,920 --> 01:00:14,720
where the scheduler limits or blocks automatic changes.

1517
01:00:14,720 --> 01:00:16,080
Work already running stays put

1518
01:00:16,080 --> 01:00:18,000
unless somebody deliberately intervenes.

1519
01:00:18,000 --> 01:00:19,760
Work starting soon also gets protection

1520
01:00:19,760 --> 01:00:22,240
because the team may have already staged material

1521
01:00:22,240 --> 01:00:23,680
prepared tooling, plan labor,

1522
01:00:23,680 --> 01:00:25,200
or started setup around it.

1523
01:00:25,200 --> 01:00:27,440
The exact length depends on your process.

1524
01:00:27,440 --> 01:00:29,280
A fast-moving line might need a short window

1525
01:00:29,280 --> 01:00:31,600
since conditions change quickly and lots move fast,

1526
01:00:31,600 --> 01:00:33,920
while a complex batch process might need a longer one

1527
01:00:33,920 --> 01:00:35,120
because shifting the sequence,

1528
01:00:35,120 --> 01:00:36,080
wastes preparation,

1529
01:00:36,080 --> 01:00:38,160
or creates real safety and quality risk.

1530
01:00:38,160 --> 01:00:39,920
There's no one-size-fits-all setting.

1531
01:00:39,920 --> 01:00:42,800
The policy has to match how production actually runs.

1532
01:00:42,800 --> 01:00:45,840
Outside the free zone, the scheduler has more room to shift work.

1533
01:00:45,840 --> 01:00:47,520
That's the planning horizon where rescheduling

1534
01:00:47,520 --> 01:00:49,360
can still improve the outcome.

1535
01:00:49,360 --> 01:00:51,200
If a breakdown affects tomorrow's output,

1536
01:00:51,200 --> 01:00:52,960
the system might bump later orders,

1537
01:00:52,960 --> 01:00:54,400
check alternative slots,

1538
01:00:54,400 --> 01:00:56,320
and flag where dates come under pressure.

1539
01:00:56,320 --> 01:00:58,640
And here's why that matters as we look further out.

1540
01:00:58,640 --> 01:01:01,040
The schedule often ends up less precise anyway.

1541
01:01:01,040 --> 01:01:02,960
Demand changes, suppliers move dates,

1542
01:01:02,960 --> 01:01:04,640
engineering revises a route.

1543
01:01:04,640 --> 01:01:06,720
So it makes little sense to treat every future slot

1544
01:01:06,720 --> 01:01:07,760
as a hard commitment.

1545
01:01:07,760 --> 01:01:09,120
The schedule needs enough detail

1546
01:01:09,120 --> 01:01:10,800
near execution to support the shift

1547
01:01:10,800 --> 01:01:13,360
and enough flexibility later to absorb new information

1548
01:01:13,360 --> 01:01:14,960
that balance controls nervousness.

1549
01:01:14,960 --> 01:01:17,920
Nervousness means the plan reacts to aggressively

1550
01:01:17,920 --> 01:01:19,120
to small changes.

1551
01:01:19,120 --> 01:01:21,680
One order finishes a little late and 10 others move.

1552
01:01:21,680 --> 01:01:24,720
A delivery date changes in the system re-orders a whole campaign.

1553
01:01:24,720 --> 01:01:27,280
The schedule keeps finding technically feasible answers,

1554
01:01:27,280 --> 01:01:30,160
but the people executing it spend their day chasing the latest version.

1555
01:01:30,160 --> 01:01:31,120
You don't want that.

1556
01:01:31,120 --> 01:01:33,440
A better approach uses rules about when an event

1557
01:01:33,440 --> 01:01:34,800
deserves a reschedule,

1558
01:01:34,800 --> 01:01:36,480
how much of the horizon can move

1559
01:01:36,480 --> 01:01:38,560
and what changes need human approval.

1560
01:01:38,560 --> 01:01:41,120
A short delay might only affect the current resource

1561
01:01:41,120 --> 01:01:42,240
and the immediate queue.

1562
01:01:42,240 --> 01:01:44,720
A major outage might justify a broader recalculation

1563
01:01:44,720 --> 01:01:46,160
because the lost capacity changes

1564
01:01:46,160 --> 01:01:47,360
several customer commitments.

1565
01:01:47,360 --> 01:01:49,200
The response should fit the event.

1566
01:01:49,200 --> 01:01:51,360
Planets can also protect certain decisions

1567
01:01:51,360 --> 01:01:52,720
from automatic movement.

1568
01:01:52,720 --> 01:01:54,320
An order might sit close to shipment,

1569
01:01:54,320 --> 01:01:56,720
a campaign might have already consumed setup time

1570
01:01:56,720 --> 01:01:58,800
or a customer promise might need direct review

1571
01:01:58,800 --> 01:02:00,080
before anything changes.

1572
01:02:00,080 --> 01:02:02,080
These aren't excuses to ignore the math.

1573
01:02:02,080 --> 01:02:04,080
Their business conditions the math needs to respect.

1574
01:02:04,080 --> 01:02:06,320
Now we get to the human review.

1575
01:02:06,320 --> 01:02:09,440
The scheduling engine can calculate feasible options quickly.

1576
01:02:09,440 --> 01:02:11,840
It can show the effect of holding the current sequence,

1577
01:02:11,840 --> 01:02:13,280
moving selected orders,

1578
01:02:13,280 --> 01:02:14,800
using extra capacity,

1579
01:02:14,800 --> 01:02:17,440
or accepting lateness on lower priority work

1580
01:02:17,440 --> 01:02:20,080
that saves the planner from rebuilding the schedule by hand

1581
01:02:20,080 --> 01:02:21,600
each time an exception hits.

1582
01:02:21,600 --> 01:02:23,840
But the scheduler doesn't own the decision alone.

1583
01:02:23,840 --> 01:02:26,080
The planner brings context that rarely fits

1584
01:02:26,080 --> 01:02:27,280
cleanly into any model.

1585
01:02:27,280 --> 01:02:29,600
A customer may accept a one-day delay on one order,

1586
01:02:29,600 --> 01:02:30,400
but not another.

1587
01:02:30,400 --> 01:02:32,240
A supervisor may know a suggested change

1588
01:02:32,240 --> 01:02:34,240
creates trouble during the next shift handover.

1589
01:02:34,240 --> 01:02:35,440
Maintenance may have information

1590
01:02:35,440 --> 01:02:37,760
that changes the confidence around the repair estimate.

1591
01:02:37,760 --> 01:02:39,200
So the system should propose,

1592
01:02:39,200 --> 01:02:40,720
the planner should decide.

1593
01:02:40,720 --> 01:02:42,560
That decision also needs a clear record.

1594
01:02:42,560 --> 01:02:44,800
If the planner overrides the proposed sequence,

1595
01:02:44,800 --> 01:02:46,080
people should know why.

1596
01:02:46,080 --> 01:02:48,080
Maybe the plant protected a service order,

1597
01:02:48,080 --> 01:02:49,600
avoided a high-risk setup,

1598
01:02:49,600 --> 01:02:51,040
or chose to keep work stable

1599
01:02:51,040 --> 01:02:52,640
and accept a later date rather than

1600
01:02:52,640 --> 01:02:55,040
create confusion across several departments.

1601
01:02:55,040 --> 01:02:57,760
Over time, those decisions show where the model needs better rules

1602
01:02:57,760 --> 01:02:59,680
and where human judgment should remain visible.

1603
01:02:59,680 --> 01:03:00,560
To make that work,

1604
01:03:00,560 --> 01:03:02,880
the planner needs facts from more than one system.

1605
01:03:02,880 --> 01:03:04,560
The machine event starts on the shop floor.

1606
01:03:04,560 --> 01:03:06,560
Material status lives somewhere else.

1607
01:03:06,560 --> 01:03:08,080
Quality release, order priority,

1608
01:03:08,080 --> 01:03:09,440
and current work completion

1609
01:03:09,440 --> 01:03:12,480
each exist in different parts of the IT and OT environment.

1610
01:03:12,480 --> 01:03:15,040
The architecture from shop floor to schedule.

1611
01:03:15,040 --> 01:03:16,880
Let's cut through the hype for a second.

1612
01:03:16,880 --> 01:03:18,960
A schedule can only respond to a disruption

1613
01:03:18,960 --> 01:03:21,040
if it receives a usable signal from the place

1614
01:03:21,040 --> 01:03:22,400
where the disruption happened.

1615
01:03:22,400 --> 01:03:24,080
That starts close to the equipment.

1616
01:03:24,080 --> 01:03:25,600
A machine reports a state,

1617
01:03:25,600 --> 01:03:28,240
running, stopped, idle, or in setup,

1618
01:03:28,240 --> 01:03:30,560
while the operator records what order is running,

1619
01:03:30,560 --> 01:03:32,160
how many good parts have completed,

1620
01:03:32,160 --> 01:03:33,760
and whether the operation can continue.

1621
01:03:34,640 --> 01:03:37,440
When a stop occurs, the downtime reason matters too.

1622
01:03:37,440 --> 01:03:39,840
Because a planned tool change needs a different response

1623
01:03:39,840 --> 01:03:40,960
from a machine fault.

1624
01:03:40,960 --> 01:03:44,000
Raw machine state alone doesn't tell the full story.

1625
01:03:44,000 --> 01:03:46,800
A machine may show as running while it produces a test part,

1626
01:03:46,800 --> 01:03:47,840
waits for approval,

1627
01:03:47,840 --> 01:03:50,480
or processes a batch that can't release downstream.

1628
01:03:50,480 --> 01:03:53,200
The operator and the execution process at meaning.

1629
01:03:53,200 --> 01:03:55,200
They connect a signal from an asset

1630
01:03:55,200 --> 01:03:57,760
to a real production order, a real operation,

1631
01:03:57,760 --> 01:03:59,280
and the current state of that work.

1632
01:03:59,280 --> 01:04:01,920
That's usually where the manufacturing execution system,

1633
01:04:01,920 --> 01:04:05,360
MES, comes in, the MES manages execution on the shop floor.

1634
01:04:05,360 --> 01:04:08,560
It dispatches work to a resource, records actual start and finish,

1635
01:04:08,560 --> 01:04:10,800
tracks quantities, captures labor activity,

1636
01:04:10,800 --> 01:04:13,760
and enforces steps like inspection or electronic sign-off.

1637
01:04:13,760 --> 01:04:16,720
In plans that need traceability, it also connects material lots,

1638
01:04:16,720 --> 01:04:18,880
serial numbers, process parameters,

1639
01:04:18,880 --> 01:04:20,720
and quality results to the work order.

1640
01:04:20,720 --> 01:04:23,920
Think of the MES as the system that knows what people are doing right now.

1641
01:04:23,920 --> 01:04:25,920
It should know whether an operation has started,

1642
01:04:25,920 --> 01:04:27,920
finished, produced, usable quantity,

1643
01:04:27,920 --> 01:04:29,920
or hit a hold that blocks the next step.

1644
01:04:29,920 --> 01:04:31,760
A scheduler needs that execution state

1645
01:04:31,760 --> 01:04:34,720
because it can't sensibly move work that's already started,

1646
01:04:34,720 --> 01:04:36,480
and it shouldn't plan a later operation

1647
01:04:36,480 --> 01:04:39,440
as available when the prior one still waits for inspection.

1648
01:04:39,440 --> 01:04:41,520
Labor reporting belongs in the same picture.

1649
01:04:41,520 --> 01:04:43,680
An operator may report time against an order,

1650
01:04:43,680 --> 01:04:46,400
a setup, a quality check, or a downtime event.

1651
01:04:46,400 --> 01:04:49,040
That data helps the plant understand actual resource use,

1652
01:04:49,040 --> 01:04:51,600
but it also helps planning see whether the current schedule

1653
01:04:51,600 --> 01:04:53,280
still resembles the work underway.

1654
01:04:53,280 --> 01:04:56,720
The point isn't to turn every human action into a data entry task.

1655
01:04:56,720 --> 01:04:59,360
It's to capture the events that change a planning decision.

1656
01:04:59,360 --> 01:05:00,720
Above that execution layer,

1657
01:05:00,720 --> 01:05:03,920
sits the enterprise resource planning system, ERP.

1658
01:05:03,920 --> 01:05:06,720
ERP holds the commercial and transactional side of the work.

1659
01:05:06,720 --> 01:05:08,960
It knows customer demand, sales orders,

1660
01:05:08,960 --> 01:05:12,000
work orders, bills of material, purchase orders,

1661
01:05:12,000 --> 01:05:14,720
inventory transactions, and requested delivery dates.

1662
01:05:14,720 --> 01:05:16,880
It may also carry cost context that matters

1663
01:05:16,880 --> 01:05:19,840
when a planner considers overtime, subcontracting,

1664
01:05:19,840 --> 01:05:21,840
or a less efficient alternate route.

1665
01:05:21,840 --> 01:05:24,880
ERP answers questions the shop floor can't answer alone.

1666
01:05:24,880 --> 01:05:26,080
Why does this order exist?

1667
01:05:26,080 --> 01:05:27,440
What quantity does the customer need,

1668
01:05:27,440 --> 01:05:29,040
which components should be available?

1669
01:05:29,040 --> 01:05:30,720
When did the business promise delivery

1670
01:05:30,720 --> 01:05:33,280
and which purchase receipt affects the order?

1671
01:05:33,280 --> 01:05:34,720
Those facts belong in the schedule,

1672
01:05:34,720 --> 01:05:37,600
but ERP doesn't need to become the real time control system

1673
01:05:37,600 --> 01:05:39,120
for every machine and operator.

1674
01:05:39,120 --> 01:05:41,520
The scheduler sits between demand and execution.

1675
01:05:41,520 --> 01:05:43,920
It takes work orders and due dates from ERP,

1676
01:05:43,920 --> 01:05:46,400
actual progress and constraints from miss and the floor

1677
01:05:46,400 --> 01:05:47,840
then applies the production model.

1678
01:05:47,840 --> 01:05:50,640
That model includes the route, eligible resources,

1679
01:05:50,640 --> 01:05:51,920
available calendars,

1680
01:05:51,920 --> 01:05:53,760
and rules about what can run where and when.

1681
01:05:53,760 --> 01:05:55,840
It turns business demand into a sequence

1682
01:05:55,840 --> 01:05:58,480
that respects the limits the plant has chosen to model.

1683
01:05:58,480 --> 01:06:00,720
The finite calendar is more than a list of hours.

1684
01:06:00,720 --> 01:06:03,200
It describes when a resource can do work when it can't

1685
01:06:03,200 --> 01:06:06,080
and sometimes what kind of work it can perform during that time.

1686
01:06:06,080 --> 01:06:08,240
The scheduling engine then looks for feasible slots

1687
01:06:08,240 --> 01:06:10,720
while respecting operation order, resource limits,

1688
01:06:10,720 --> 01:06:11,840
and policy rules.

1689
01:06:11,840 --> 01:06:14,720
Depending on the problem, it may use fixed dispatch rules,

1690
01:06:14,720 --> 01:06:17,120
search methods that find workable answers quickly

1691
01:06:17,120 --> 01:06:20,240
or formal optimization logic that compares competing outcomes.

1692
01:06:20,240 --> 01:06:23,600
None of those methods removes the need for clear inputs.

1693
01:06:23,600 --> 01:06:25,760
If the engine receives an order as complete

1694
01:06:25,760 --> 01:06:27,520
when it's actually blocked in quality,

1695
01:06:27,520 --> 01:06:29,520
it will plan against a false state.

1696
01:06:29,520 --> 01:06:32,400
If a machine appears available while maintenance has restricted it,

1697
01:06:32,400 --> 01:06:34,960
the schedule will load time that production can't use.

1698
01:06:34,960 --> 01:06:37,200
The architecture needs a reliable path for events

1699
01:06:37,200 --> 01:06:39,040
and a clear owner for each fact.

1700
01:06:39,040 --> 01:06:41,120
Timing matters as much as data content.

1701
01:06:41,120 --> 01:06:42,880
Some information can move in batches,

1702
01:06:42,880 --> 01:06:45,760
demand changes, new work orders, revised due dates,

1703
01:06:45,760 --> 01:06:47,360
and broad material planning updates

1704
01:06:47,360 --> 01:06:49,680
may only need a scheduled exchange a few times a day

1705
01:06:49,680 --> 01:06:50,960
depending on the plant.

1706
01:06:50,960 --> 01:06:52,720
That's often enough for medium term planning

1707
01:06:52,720 --> 01:06:54,720
because the decision doesn't change every minute.

1708
01:06:54,720 --> 01:06:56,400
Other events need a faster path,

1709
01:06:56,400 --> 01:06:58,400
a breakdown at a constrained resource,

1710
01:06:58,400 --> 01:07:01,200
an operation completion that releases work downstream

1711
01:07:01,200 --> 01:07:04,480
or a quality hold that stops a batch can affect today's schedule.

1712
01:07:04,480 --> 01:07:07,680
Those events might flow through near real-time integration,

1713
01:07:07,680 --> 01:07:09,520
where the scheduler receives them soon enough

1714
01:07:09,520 --> 01:07:12,320
to assess the effect while people can still act.

1715
01:07:12,320 --> 01:07:14,480
Near real-time doesn't mean every sensor value

1716
01:07:14,480 --> 01:07:16,160
needs to enter the scheduling engine.

1717
01:07:16,160 --> 01:07:18,480
Most machine signals don't change the plan.

1718
01:07:18,480 --> 01:07:20,960
A scheduler needs meaningful state changes,

1719
01:07:20,960 --> 01:07:22,480
confirmed production progress,

1720
01:07:22,480 --> 01:07:25,120
and exceptions that alter capacity or readiness.

1721
01:07:25,120 --> 01:07:27,600
Sending every signal into planning creates noise,

1722
01:07:27,600 --> 01:07:29,200
cost, and a very busy system

1723
01:07:29,200 --> 01:07:31,920
that still can't tell a planner whether an order can run.

1724
01:07:31,920 --> 01:07:34,160
Some events stay manual, and that's sensible.

1725
01:07:34,160 --> 01:07:36,720
A supervisor may know a machine can finish the current batch

1726
01:07:36,720 --> 01:07:38,240
despite a fault warning.

1727
01:07:38,240 --> 01:07:40,320
Customer service may receive a priority request

1728
01:07:40,320 --> 01:07:43,040
that needs commercial review before it changes the queue.

1729
01:07:43,040 --> 01:07:45,360
Maintenance may estimate an outage with uncertainty

1730
01:07:45,360 --> 01:07:47,840
and update that estimate as diagnosis continues.

1731
01:07:47,840 --> 01:07:50,160
Those decisions need controlled exception handling.

1732
01:07:50,160 --> 01:07:51,600
A planner should see the event,

1733
01:07:51,600 --> 01:07:53,920
the assumption behind it, and the schedule effect.

1734
01:07:53,920 --> 01:07:56,160
Then the plan can approve a change, reject it,

1735
01:07:56,160 --> 01:07:58,000
or wait for more information.

1736
01:07:58,000 --> 01:08:00,080
Integration that actually works end-to-end

1737
01:08:00,080 --> 01:08:02,640
doesn't mean removing people from the process.

1738
01:08:02,640 --> 01:08:05,280
It means connecting the dots between IT and OT

1739
01:08:05,280 --> 01:08:07,280
without losing who owns the decision.

1740
01:08:07,280 --> 01:08:09,040
That architecture gives us the right place

1741
01:08:09,040 --> 01:08:10,400
to discuss Microsoft.

1742
01:08:10,400 --> 01:08:11,920
Not as the scheduler by default,

1743
01:08:11,920 --> 01:08:14,400
but as part of the data, integration, analysis,

1744
01:08:14,400 --> 01:08:16,400
and decision layer around the production schedule.

1745
01:08:16,400 --> 01:08:18,640
Where Microsoft fits?

1746
01:08:18,640 --> 01:08:19,600
Let me put it this way.

1747
01:08:19,600 --> 01:08:21,680
Microsoft can help with the scheduling decision,

1748
01:08:21,680 --> 01:08:23,440
but it won't replace the production logic

1749
01:08:23,440 --> 01:08:25,040
that actually knows your plant.

1750
01:08:25,040 --> 01:08:27,040
You still need that layer and you need to own it.

1751
01:08:27,040 --> 01:08:29,920
Start at the shop floor, because that's where the real work happens.

1752
01:08:29,920 --> 01:08:31,520
Azure gives you a connection layer,

1753
01:08:31,520 --> 01:08:34,480
where equipment, edge systems, or IoT gateways

1754
01:08:34,480 --> 01:08:37,120
send operational events into the rest of your architecture.

1755
01:08:37,120 --> 01:08:39,280
A plant might send a confirmed stop

1756
01:08:39,280 --> 01:08:40,320
at a constrained machine,

1757
01:08:40,320 --> 01:08:42,160
a machine state change, a completed quantity,

1758
01:08:42,160 --> 01:08:44,400
or an alarm that maintenance needs to check.

1759
01:08:44,400 --> 01:08:46,480
I said useful events, and that word matters.

1760
01:08:46,480 --> 01:08:47,600
You're not just collecting noise.

1761
01:08:47,600 --> 01:08:49,760
You don't need to stream every PLC signal

1762
01:08:49,760 --> 01:08:52,320
into a planning process just because the technology can do it.

1763
01:08:52,320 --> 01:08:53,600
The scheduling team needs events

1764
01:08:53,600 --> 01:08:56,240
that change capacity, work status, or readiness.

1765
01:08:56,240 --> 01:08:59,120
Azure and an IoT layer can pull those events out of OT

1766
01:08:59,120 --> 01:09:00,240
in a controlled way,

1767
01:09:00,240 --> 01:09:02,320
but you have to respect that the machine network

1768
01:09:02,320 --> 01:09:04,720
and the production process have very different needs

1769
01:09:04,720 --> 01:09:06,160
than your enterprise systems.

1770
01:09:06,160 --> 01:09:07,680
That's not a technology problem.

1771
01:09:07,680 --> 01:09:09,120
It's an architecture problem.

1772
01:09:09,120 --> 01:09:10,800
That brings us to the next piece.

1773
01:09:10,800 --> 01:09:12,480
You need a place to bring facts together

1774
01:09:12,480 --> 01:09:14,160
without letting every department own

1775
01:09:14,160 --> 01:09:15,600
its own version of production.

1776
01:09:15,600 --> 01:09:17,120
That's where Microsoft fabric comes in.

1777
01:09:17,120 --> 01:09:18,720
It pulls ERP records,

1778
01:09:18,720 --> 01:09:21,760
MES events, maintenance data, quality results,

1779
01:09:21,760 --> 01:09:25,040
and planning history into a shared data hub everyone can trust.

1780
01:09:25,040 --> 01:09:26,640
But here's the key distinction.

1781
01:09:26,640 --> 01:09:28,160
Fabric doesn't run the machine

1782
01:09:28,160 --> 01:09:29,920
and it doesn't create the schedule.

1783
01:09:29,920 --> 01:09:31,760
Its job is to build a common foundation

1784
01:09:31,760 --> 01:09:33,760
around the decisions people already make.

1785
01:09:33,760 --> 01:09:35,200
Picture a planner trying to figure out

1786
01:09:35,200 --> 01:09:36,720
why an order finished late.

1787
01:09:36,720 --> 01:09:38,880
The answer sits across several systems.

1788
01:09:38,880 --> 01:09:41,600
ERP holds the promised date and customer priority.

1789
01:09:41,600 --> 01:09:44,800
MS holds actual operation times and reported quantities.

1790
01:09:44,800 --> 01:09:46,480
Maintenance recorded an outage,

1791
01:09:46,480 --> 01:09:48,640
quality locked or hold or a failed inspection.

1792
01:09:48,640 --> 01:09:51,680
Fabric pulls those records into one model

1793
01:09:51,680 --> 01:09:53,360
where people can trace the chain of events

1794
01:09:53,360 --> 01:09:55,520
instead of arguing from separate reports.

1795
01:09:55,520 --> 01:09:57,360
That gives you a clear view of the plan

1796
01:09:57,360 --> 01:09:58,800
versus what actually happened.

1797
01:09:58,800 --> 01:10:01,120
Once you have that foundation, power BI works well,

1798
01:10:01,120 --> 01:10:02,480
it can show schedule adherence,

1799
01:10:02,480 --> 01:10:03,760
Q growth at a bottleneck,

1800
01:10:03,760 --> 01:10:05,600
recurring reasons for late completion

1801
01:10:05,600 --> 01:10:08,480
and the effect of changes people made during a disruption.

1802
01:10:08,480 --> 01:10:11,120
It also helps compare scenarios after the fact.

1803
01:10:11,120 --> 01:10:12,960
For example, did a decision to protect

1804
01:10:12,960 --> 01:10:16,080
a high priority order create avoidable delays elsewhere

1805
01:10:16,080 --> 01:10:18,320
or did it prevent a larger customer problem?

1806
01:10:18,320 --> 01:10:19,920
A dashboard doesn't schedule production

1807
01:10:19,920 --> 01:10:22,640
but it exposes patterns that suggest changing the plan.

1808
01:10:22,640 --> 01:10:26,080
If a bottleneck repeatedly loses time during certain transitions,

1809
01:10:26,080 --> 01:10:28,320
the plant should question the setup model.

1810
01:10:28,320 --> 01:10:30,960
When planners override the same proposal repeatedly,

1811
01:10:30,960 --> 01:10:33,600
the rules probably don't match how production actually works.

1812
01:10:33,600 --> 01:10:35,680
If late orders trace back to material release

1813
01:10:35,680 --> 01:10:37,120
rather than machine capacity,

1814
01:10:37,120 --> 01:10:39,600
the response belongs upstream of the scheduler.

1815
01:10:39,600 --> 01:10:41,360
That's visibility with the purpose.

1816
01:10:41,360 --> 01:10:42,720
It leads to a better decision,

1817
01:10:42,720 --> 01:10:45,360
not another report confirming everyone had a tough week.

1818
01:10:45,360 --> 01:10:48,160
Power Platform supports the actions around exceptions.

1819
01:10:48,160 --> 01:10:50,240
Say maintenance posts are likely outage

1820
01:10:50,240 --> 01:10:53,120
and the new schedule puts several customer dates at risk.

1821
01:10:53,120 --> 01:10:55,520
A controlled workflow can send the affected decision

1822
01:10:55,520 --> 01:10:56,880
to the right people.

1823
01:10:56,880 --> 01:10:58,720
The planner reviews the schedule impact,

1824
01:10:58,720 --> 01:11:00,480
production checks, shift feasibility,

1825
01:11:00,480 --> 01:11:02,960
quality approves or rejects an alternate route

1826
01:11:02,960 --> 01:11:05,440
and customer service receives a confirmed position

1827
01:11:05,440 --> 01:11:07,440
before communicating a revised date.

1828
01:11:07,440 --> 01:11:10,240
The workflow needs clear ownership.

1829
01:11:10,240 --> 01:11:11,840
That's the part where people skip.

1830
01:11:11,840 --> 01:11:14,160
A Power app gives a supervisor or planner

1831
01:11:14,160 --> 01:11:16,320
a simple way to record an exception,

1832
01:11:16,320 --> 01:11:18,960
choose a reason code and request a decision

1833
01:11:18,960 --> 01:11:22,160
without working through email chains and unofficial spreadsheets.

1834
01:11:22,160 --> 01:11:25,120
Power automate routes that request keeps an approval record

1835
01:11:25,120 --> 01:11:26,960
and notifies the people who need to act.

1836
01:11:26,960 --> 01:11:28,880
The workflow ensures a priority change

1837
01:11:28,880 --> 01:11:30,480
or routing exception doesn't disappear

1838
01:11:30,480 --> 01:11:31,920
into somebody's inbox.

1839
01:11:31,920 --> 01:11:34,080
It's not about automating every conversation.

1840
01:11:34,080 --> 01:11:37,360
Copilot and Azure AI support a different part of the work.

1841
01:11:37,360 --> 01:11:39,360
They let people ask questions in plain language

1842
01:11:39,360 --> 01:11:40,720
across approved data.

1843
01:11:40,720 --> 01:11:43,200
A planner might ask why a specific order moved

1844
01:11:43,200 --> 01:11:45,120
which constrained blocked an alternate resource

1845
01:11:45,120 --> 01:11:47,360
or which orders face risk if a machine stays down

1846
01:11:47,360 --> 01:11:48,480
for the rest of the shift.

1847
01:11:48,480 --> 01:11:50,240
The AI layer retrieves context,

1848
01:11:50,240 --> 01:11:52,560
explains the schedule logic in readable language

1849
01:11:52,560 --> 01:11:55,840
and drafts a summary for a production meeting or customer update.

1850
01:11:55,840 --> 01:11:58,240
That saves time, but here's the boundary.

1851
01:11:58,240 --> 01:12:01,600
Generative AI should not replace deterministic scheduling logic.

1852
01:12:01,600 --> 01:12:04,640
It can explain a recommendation and summarize options

1853
01:12:04,640 --> 01:12:06,800
and it can retrieve current facts from ERP,

1854
01:12:06,800 --> 01:12:09,120
MES, maintenance and quality data.

1855
01:12:09,120 --> 01:12:11,200
What it should not do is invent a production sequence

1856
01:12:11,200 --> 01:12:13,920
from text alone and present it as physically feasible.

1857
01:12:13,920 --> 01:12:15,440
For the actual scheduling decision,

1858
01:12:15,440 --> 01:12:17,440
the system still needs explicit constraints,

1859
01:12:17,440 --> 01:12:20,000
calendars, routes, capacities and policy rules

1860
01:12:20,000 --> 01:12:22,400
and optimization engine tests those conditions.

1861
01:12:22,400 --> 01:12:24,080
Copilot sits beside that engine

1862
01:12:24,080 --> 01:12:26,000
and helps people understand the result.

1863
01:12:26,000 --> 01:12:27,920
Ask the same question every time.

1864
01:12:27,920 --> 01:12:30,080
What does the AI actually know?

1865
01:12:30,080 --> 01:12:31,840
If it cannot see that a fixture is occupied

1866
01:12:31,840 --> 01:12:33,040
and operator lacks approval,

1867
01:12:33,040 --> 01:12:34,800
a material that remains on hold

1868
01:12:34,800 --> 01:12:36,880
or a machine has a plant maintenance window,

1869
01:12:36,880 --> 01:12:38,640
then it cannot give a trustworthy answer

1870
01:12:38,640 --> 01:12:39,680
about what should run next.

1871
01:12:40,480 --> 01:12:43,520
Good language does not repair missing production context.

1872
01:12:43,520 --> 01:12:45,840
It only makes the gaps sound more convincing.

1873
01:12:45,840 --> 01:12:47,680
So here's where Microsoft fits.

1874
01:12:47,680 --> 01:12:50,160
It connects the dots between IT and OT,

1875
01:12:50,160 --> 01:12:52,400
governs data, supports analysis

1876
01:12:52,400 --> 01:12:54,400
and helps people work through exceptions.

1877
01:12:54,400 --> 01:12:55,840
It doesn't know your valid routes,

1878
01:12:55,840 --> 01:12:57,200
your real change over rules

1879
01:12:57,200 --> 01:13:00,000
or the production policy behind the next order on the machine

1880
01:13:00,000 --> 01:13:01,360
that knowledge still belongs to you.

1881
01:13:01,360 --> 01:13:04,080
What Microsoft does not model for you,

1882
01:13:04,080 --> 01:13:05,840
no generic platform knows the rules

1883
01:13:05,840 --> 01:13:08,320
that make your factory run safely and predictably.

1884
01:13:08,320 --> 01:13:09,200
Consider this.

1885
01:13:09,200 --> 01:13:11,280
A part can technically run on two machines,

1886
01:13:11,280 --> 01:13:13,760
but only one holds the approved fixture this week,

1887
01:13:13,760 --> 01:13:16,720
only one evening shift operator holds the needed approval

1888
01:13:16,720 --> 01:13:18,640
and the second route needs a quality review

1889
01:13:18,640 --> 01:13:20,000
before anyone can use it.

1890
01:13:20,000 --> 01:13:21,840
Those are not standard software settings,

1891
01:13:21,840 --> 01:13:23,040
that's production knowledge

1892
01:13:23,040 --> 01:13:24,560
and it lives in your plant.

1893
01:13:24,560 --> 01:13:26,640
The same is true for setup relationships.

1894
01:13:26,640 --> 01:13:28,640
A system doesn't know which product transition needs

1895
01:13:28,640 --> 01:13:29,760
a quick tool change,

1896
01:13:29,760 --> 01:13:31,200
which needs a long clean-out

1897
01:13:31,200 --> 01:13:32,800
or which sequence production avoids

1898
01:13:32,800 --> 01:13:35,200
because it caused quality issues in the past.

1899
01:13:35,200 --> 01:13:36,880
Somebody needs to model those conditions,

1900
01:13:36,880 --> 01:13:37,840
keep them current

1901
01:13:37,840 --> 01:13:40,640
and agree who can change them when the process changes.

1902
01:13:40,640 --> 01:13:42,800
Production policy also belongs to the plant,

1903
01:13:42,800 --> 01:13:44,480
things like whether the scheduler should protect

1904
01:13:44,480 --> 01:13:46,080
the earliest customer date,

1905
01:13:46,080 --> 01:13:48,080
keep a campaign together for throughput,

1906
01:13:48,080 --> 01:13:50,160
reserve capacity for service parts,

1907
01:13:50,160 --> 01:13:51,120
split an order,

1908
01:13:51,120 --> 01:13:52,880
or decide when an urgent request

1909
01:13:52,880 --> 01:13:54,960
justifies disrupting a stable sequence.

1910
01:13:54,960 --> 01:13:57,680
Software can apply those choices consistently,

1911
01:13:57,680 --> 01:13:58,960
but it cannot choose them

1912
01:13:58,960 --> 01:14:00,720
without being told what good looks like,

1913
01:14:00,720 --> 01:14:03,120
that scheduling logic can live in different places.

1914
01:14:03,120 --> 01:14:05,760
Some manufacturers use functions built into the ERP,

1915
01:14:05,760 --> 01:14:08,960
others use the MES to dispatch work close to the shop floor,

1916
01:14:08,960 --> 01:14:10,720
some use a separate APS system,

1917
01:14:10,720 --> 01:14:13,200
and others use specialized optimization software

1918
01:14:13,200 --> 01:14:14,800
for a narrow but difficult problem

1919
01:14:14,800 --> 01:14:16,080
like a complex bottleneck

1920
01:14:16,080 --> 01:14:18,000
or a sequence with many setup limits.

1921
01:14:18,000 --> 01:14:19,840
The location matters less than the fit.

1922
01:14:19,840 --> 01:14:21,440
An APS system is built for planning

1923
01:14:21,440 --> 01:14:23,120
and scheduling under constraints.

1924
01:14:23,120 --> 01:14:24,480
It takes production orders,

1925
01:14:24,480 --> 01:14:25,920
breaks them into operations,

1926
01:14:25,920 --> 01:14:27,280
checks resource calendars,

1927
01:14:27,280 --> 01:14:28,800
accounts for setup rules,

1928
01:14:28,800 --> 01:14:30,160
and places work in a sequence

1929
01:14:30,160 --> 01:14:31,760
that follows your model limits.

1930
01:14:31,760 --> 01:14:34,640
It usually sits between ERP demand planning

1931
01:14:34,640 --> 01:14:35,920
and mass execution,

1932
01:14:35,920 --> 01:14:38,080
though the exact split depends on the plant

1933
01:14:38,080 --> 01:14:39,840
and the systems already in place.

1934
01:14:39,840 --> 01:14:41,760
Think of APS as the scheduling brain.

1935
01:14:41,760 --> 01:14:43,200
It doesn't own every fact.

1936
01:14:43,200 --> 01:14:45,120
ERP owns orders, demand, purchasing,

1937
01:14:45,120 --> 01:14:46,400
and core master data.

1938
01:14:46,400 --> 01:14:49,520
MES owns execution status and traceability,

1939
01:14:49,520 --> 01:14:50,960
maintenance owns planned downtime

1940
01:14:50,960 --> 01:14:52,320
and equipment restrictions.

1941
01:14:52,320 --> 01:14:54,320
Quality owns root approvals and holds.

1942
01:14:54,320 --> 01:14:56,320
The scheduling engine uses agreed inputs

1943
01:14:56,320 --> 01:14:58,080
from each area and produces a plan

1944
01:14:58,080 --> 01:14:59,360
for review and execution,

1945
01:14:59,360 --> 01:15:02,240
but that only works when integration has clear contracts.

1946
01:15:02,240 --> 01:15:04,560
By contract, I mean more than a technical interface.

1947
01:15:04,560 --> 01:15:06,000
It defines what an event means,

1948
01:15:06,000 --> 01:15:07,360
who sends it when it arrives,

1949
01:15:07,360 --> 01:15:08,640
who owns its accuracy,

1950
01:15:08,640 --> 01:15:10,400
and what the scheduler should do with it.

1951
01:15:10,400 --> 01:15:12,080
If MES reports an operation complete,

1952
01:15:12,080 --> 01:15:14,240
does that mean the whole quantity passed inspection,

1953
01:15:14,240 --> 01:15:16,240
or only that the machine finished its run?

1954
01:15:16,240 --> 01:15:18,640
If maintenance flags a resource as unavailable,

1955
01:15:18,640 --> 01:15:20,400
does the event include a return estimate

1956
01:15:20,400 --> 01:15:21,680
or only a stop signal?

1957
01:15:21,680 --> 01:15:23,680
Small differences change the schedule.

1958
01:15:23,680 --> 01:15:25,680
Master data needs the same discipline.

1959
01:15:25,680 --> 01:15:28,160
Roots, work centers, calendars, approved alternates,

1960
01:15:28,160 --> 01:15:29,760
setup families, and resource rules

1961
01:15:29,760 --> 01:15:31,680
need an owner and a controlled way to change.

1962
01:15:31,680 --> 01:15:34,000
An integration can move bad data very quickly,

1963
01:15:34,000 --> 01:15:35,040
which is efficient.

1964
01:15:35,040 --> 01:15:37,440
The same way a fast conveyor can move wrong parts

1965
01:15:37,440 --> 01:15:38,640
to the next station.

1966
01:15:38,640 --> 01:15:41,440
Test cases matter because production rules contain exceptions.

1967
01:15:41,440 --> 01:15:43,120
You need to test a routine order,

1968
01:15:43,120 --> 01:15:44,560
a late material receipt,

1969
01:15:44,560 --> 01:15:45,600
a quality hold,

1970
01:15:45,600 --> 01:15:46,960
a planned maintenance slot,

1971
01:15:46,960 --> 01:15:48,000
a partial completion,

1972
01:15:48,000 --> 01:15:50,720
and an urgent order that conflicts with the campaign.

1973
01:15:50,720 --> 01:15:53,200
The question isn't whether data reaches the scheduler.

1974
01:15:53,200 --> 01:15:55,200
It's whether the resulting schedule behaves

1975
01:15:55,200 --> 01:15:56,800
in a way production can accept.

1976
01:15:56,800 --> 01:15:58,080
Here's the reality.

1977
01:15:58,080 --> 01:15:59,920
A lake house can store every constraint

1978
01:15:59,920 --> 01:16:02,000
in the plant and still schedule nothing.

1979
01:16:02,000 --> 01:16:04,080
It can hold machine events, order history,

1980
01:16:04,080 --> 01:16:06,560
setup times, labor records, and quality data.

1981
01:16:06,560 --> 01:16:09,360
That gives you a strong data foundation for analysis,

1982
01:16:09,360 --> 01:16:12,080
but somebody still needs to define the scheduling problem.

1983
01:16:12,080 --> 01:16:14,880
Model the constraints, select the decision logic,

1984
01:16:14,880 --> 01:16:17,440
and connect the output to the people who own the shift.

1985
01:16:17,440 --> 01:16:20,240
So when you compare scheduling tools, skip the product name.

1986
01:16:20,240 --> 01:16:22,480
Ask where the finite scheduling logic lives today,

1987
01:16:22,480 --> 01:16:24,080
what decisions it needs to make,

1988
01:16:24,080 --> 01:16:25,520
which facts it can trust,

1989
01:16:25,520 --> 01:16:27,840
and where people need to review an exception.

1990
01:16:27,840 --> 01:16:30,720
Once that's clear, you can look at the methods the engine might use

1991
01:16:30,720 --> 01:16:32,880
from direct rules through fast search methods

1992
01:16:32,880 --> 01:16:34,880
to formal mathematical optimization.

1993
01:16:34,880 --> 01:16:36,880
That's where the real architecture work begins.

1994
01:16:36,880 --> 01:16:39,040
Rules, heuristics, and optimization.

1995
01:16:39,040 --> 01:16:41,680
So once you find where the scheduling logic lives,

1996
01:16:41,680 --> 01:16:45,680
the real question becomes how it picks one feasible sequence over another.

1997
01:16:45,680 --> 01:16:49,200
I keep seeing people lump all these approaches together under optimization,

1998
01:16:49,200 --> 01:16:50,800
but they're not the same thing at all.

1999
01:16:50,800 --> 01:16:52,320
The simplest approach uses rules.

2000
01:16:52,320 --> 01:16:55,120
A rule might say run the order with the earliest due date first

2001
01:16:55,120 --> 01:16:57,120
to protect orders for a named customer

2002
01:16:57,120 --> 01:17:00,000
or run the shortest job first to clear the queue.

2003
01:17:00,000 --> 01:17:01,760
A plant can also adopt a campaign rule

2004
01:17:01,760 --> 01:17:04,720
to keep similar products together and avoid costly changeovers.

2005
01:17:04,720 --> 01:17:06,080
Rules are easy to explain,

2006
01:17:06,080 --> 01:17:08,240
and that matters because a planner, supervisor,

2007
01:17:08,240 --> 01:17:11,280
or operator can see why the system chose the next order.

2008
01:17:11,280 --> 01:17:14,320
A schedule nobody can explain won't survive a rough shift.

2009
01:17:14,320 --> 01:17:16,720
A rule also gives you a consistent starting point,

2010
01:17:16,720 --> 01:17:19,760
instead of relying on whoever happened to build that morning sequence.

2011
01:17:19,760 --> 01:17:20,800
Here's the problem, though.

2012
01:17:20,800 --> 01:17:23,840
A single rule becomes too simple when constraints collide.

2013
01:17:23,840 --> 01:17:25,440
Take earliest due date.

2014
01:17:25,440 --> 01:17:27,040
It protects delivery performance

2015
01:17:27,040 --> 01:17:28,800
until it forces repeated changeovers

2016
01:17:28,800 --> 01:17:30,800
and burns hours at the bottleneck.

2017
01:17:30,800 --> 01:17:33,040
Shortest processing time reduces the queue

2018
01:17:33,040 --> 01:17:35,200
while a large, urgent order keeps slipping.

2019
01:17:35,200 --> 01:17:37,600
Campaign sequencing improves flow on one resource

2020
01:17:37,600 --> 01:17:39,840
but delays in order assembly needs today.

2021
01:17:39,840 --> 01:17:41,680
The rules did exactly what you told it to do,

2022
01:17:41,680 --> 01:17:43,520
but the plant wanted more than one outcome.

2023
01:17:43,520 --> 01:17:44,800
That brings us to heuristics.

2024
01:17:44,800 --> 01:17:46,640
A heuristic is a fast search method

2025
01:17:46,640 --> 01:17:48,400
that tests possible schedules,

2026
01:17:48,400 --> 01:17:50,000
follows practical shortcuts,

2027
01:17:50,000 --> 01:17:52,960
and looks for an answer that works well enough within the time you have.

2028
01:17:52,960 --> 01:17:55,120
It does not promise the mathematically best schedule

2029
01:17:55,120 --> 01:17:56,400
out of every combination

2030
01:17:56,400 --> 01:17:58,720
and for most real factories that is a sensible trade.

2031
01:17:58,720 --> 01:18:00,080
The number of possible sequences

2032
01:18:00,080 --> 01:18:02,160
explodes once you include multiple machines.

2033
01:18:02,160 --> 01:18:03,920
Alternate routes, setup relationships,

2034
01:18:03,920 --> 01:18:05,840
release dates, and resource limits.

2035
01:18:05,840 --> 01:18:08,160
When you need a revised schedule during the shift,

2036
01:18:08,160 --> 01:18:10,800
waiting hours for a perfect answer does not help.

2037
01:18:10,800 --> 01:18:14,080
A heuristic can give you a workable plan in seconds or minutes

2038
01:18:14,080 --> 01:18:15,600
that let you review the trade-offs.

2039
01:18:15,600 --> 01:18:19,120
Think of heuristics as structured judgment at machine speed.

2040
01:18:19,120 --> 01:18:21,440
The heuristic can start with due date priority

2041
01:18:21,440 --> 01:18:24,560
then improve the sequence by reducing unnecessary changeovers,

2042
01:18:24,560 --> 01:18:27,600
filling usable gaps, or moving work to an alternate resource

2043
01:18:27,600 --> 01:18:28,880
where the rules allow it.

2044
01:18:28,880 --> 01:18:31,680
Each method differs, but the aim stays the same.

2045
01:18:31,680 --> 01:18:35,120
Find a feasible plan quickly enough to support an actual decision.

2046
01:18:35,120 --> 01:18:38,240
In practice, that speed is often the difference between a schedule

2047
01:18:38,240 --> 01:18:41,040
that gets used and one that sits in a spreadsheet

2048
01:18:41,040 --> 01:18:42,720
while the floor runs on intuition.

2049
01:18:42,720 --> 01:18:45,600
Now formal optimization sits at the other end of the spectrum.

2050
01:18:45,600 --> 01:18:47,520
Optimization defines an objective,

2051
01:18:47,520 --> 01:18:48,880
or several objectives,

2052
01:18:48,880 --> 01:18:51,760
and applies constraints as hard rules or agreed limits.

2053
01:18:51,760 --> 01:18:54,480
The engine searches for the schedule that best meets that objective

2054
01:18:54,480 --> 01:18:56,640
within the production conditions you've modeled.

2055
01:18:56,640 --> 01:18:57,840
And that wording matters.

2056
01:18:57,840 --> 01:19:00,880
Optimization does not discover what your business should care about.

2057
01:19:00,880 --> 01:19:02,000
Someone has to define it.

2058
01:19:02,000 --> 01:19:04,080
I've seen teams spend months building a model

2059
01:19:04,080 --> 01:19:07,600
only to realize they never agreed on what good actually means.

2060
01:19:07,600 --> 01:19:09,600
Let's cut through the hype for a second.

2061
01:19:09,600 --> 01:19:11,840
You might want to reduce late orders,

2062
01:19:11,840 --> 01:19:13,440
reduce total lateness,

2063
01:19:13,440 --> 01:19:15,760
rather than treating every late order the same,

2064
01:19:15,760 --> 01:19:16,960
limit over time,

2065
01:19:16,960 --> 01:19:19,280
preserve throughput at a constrained resource,

2066
01:19:19,280 --> 01:19:20,880
reduce work in progress,

2067
01:19:20,880 --> 01:19:23,600
or avoid changeovers that create waste and risk.

2068
01:19:23,600 --> 01:19:25,040
Those goals can conflict,

2069
01:19:25,040 --> 01:19:26,320
push hard to meet every due date,

2070
01:19:26,320 --> 01:19:27,920
and you'll create more setups and over time

2071
01:19:27,920 --> 01:19:30,480
while minimizing setups makes urgent orders wait longer.

2072
01:19:30,480 --> 01:19:32,400
Release more work to keep every machine busy

2073
01:19:32,400 --> 01:19:33,760
and work in progress grows,

2074
01:19:33,760 --> 01:19:35,680
lead times become unpredictable.

2075
01:19:35,680 --> 01:19:37,200
Focus only on throughput,

2076
01:19:37,200 --> 01:19:39,440
and you risk producing the wrong orders early

2077
01:19:39,440 --> 01:19:41,840
while customer critical work waits behind them.

2078
01:19:41,840 --> 01:19:43,520
There is no neutral schedule.

2079
01:19:43,520 --> 01:19:45,600
The scheduling method expresses a policy,

2080
01:19:45,600 --> 01:19:46,800
whether you write it down,

2081
01:19:46,800 --> 01:19:48,640
or leave it hidden inside a spreadsheet

2082
01:19:48,640 --> 01:19:49,840
in someone's experience.

2083
01:19:49,840 --> 01:19:52,640
Formal optimization helps when trade-offs get too complex

2084
01:19:52,640 --> 01:19:54,000
for a simple dispatch rule,

2085
01:19:54,000 --> 01:19:55,920
because it makes those trade-offs explicit

2086
01:19:55,920 --> 01:19:58,000
and test them in a disciplined way,

2087
01:19:58,000 --> 01:20:01,040
but it only works if the model actually reflects the floor reality.

2088
01:20:01,040 --> 01:20:03,120
That's why I'm careful with the phrase "best schedule."

2089
01:20:03,120 --> 01:20:06,640
The best answer only exists relative to your objective,

2090
01:20:06,640 --> 01:20:07,680
constraints,

2091
01:20:07,680 --> 01:20:10,400
and the information available when the engine runs.

2092
01:20:10,400 --> 01:20:13,120
If setup times a wrong material status is stale,

2093
01:20:13,120 --> 01:20:15,200
or commercial priority hasn't been agreed,

2094
01:20:15,200 --> 01:20:18,400
a mathematically elegant result can still produce a poor decision.

2095
01:20:18,400 --> 01:20:20,720
False precision creates false confidence.

2096
01:20:20,720 --> 01:20:22,720
A schedule with exact timestamps

2097
01:20:22,720 --> 01:20:24,080
can look authoritative,

2098
01:20:24,080 --> 01:20:26,640
even when its inputs contain rough assumptions.

2099
01:20:26,640 --> 01:20:29,120
Planners need to see why the engine chose a sequence

2100
01:20:29,120 --> 01:20:30,960
which assumptions drive the outcome,

2101
01:20:30,960 --> 01:20:34,160
and what changes if a priority or capacity condition moves.

2102
01:20:34,160 --> 01:20:35,920
So here's the takeaway.

2103
01:20:35,920 --> 01:20:37,840
Use rules when a clear policy works.

2104
01:20:37,840 --> 01:20:40,400
Use heuristics when speed and practicality matter,

2105
01:20:40,400 --> 01:20:44,000
and use optimization when competing outcomes need a systematic test.

2106
01:20:44,000 --> 01:20:45,600
Either way, keep the decision visible.

2107
01:20:45,600 --> 01:20:47,680
And that distinction matters when generative AI

2108
01:20:47,680 --> 01:20:48,800
enters the picture.

2109
01:20:48,800 --> 01:20:50,960
A language model can help people understand a plan

2110
01:20:50,960 --> 01:20:52,480
and work through exceptions.

2111
01:20:52,480 --> 01:20:54,560
But it should not quietly take over the part

2112
01:20:54,560 --> 01:20:56,720
that needs explicit production logic.

2113
01:20:56,720 --> 01:20:59,040
Industrial AI and scheduling decisions.

2114
01:20:59,040 --> 01:21:00,880
Industrial AI can support scheduling,

2115
01:21:00,880 --> 01:21:02,640
but it needs a clear job description.

2116
01:21:02,640 --> 01:21:05,920
Generative AI is most useful when a planner needs to ask questions

2117
01:21:05,920 --> 01:21:07,760
across approved production data,

2118
01:21:07,760 --> 01:21:09,120
especially during a disruption,

2119
01:21:09,120 --> 01:21:10,800
when facts are scattered across systems

2120
01:21:10,800 --> 01:21:13,520
and nobody has time to piece together a manual briefing.

2121
01:21:13,520 --> 01:21:15,600
A planner can ask why an order moved,

2122
01:21:15,600 --> 01:21:17,840
which constraint blocked its next operation,

2123
01:21:17,840 --> 01:21:19,440
or which customer dates face risk

2124
01:21:19,440 --> 01:21:22,160
if a resource remains unavailable until the next shift.

2125
01:21:22,160 --> 01:21:24,240
The AI can retrieve the relevant information,

2126
01:21:24,240 --> 01:21:25,760
explain it in plain language,

2127
01:21:25,760 --> 01:21:27,760
and point back to the underlying records.

2128
01:21:27,760 --> 01:21:29,840
That saves time, but it does not create capacity.

2129
01:21:29,840 --> 01:21:32,240
It can also take a messy pile of event loads

2130
01:21:32,240 --> 01:21:34,400
and turn them into a coherent disruption summary.

2131
01:21:34,400 --> 01:21:36,480
Say maintenance reports are fault,

2132
01:21:36,480 --> 01:21:38,000
quality holds a batch,

2133
01:21:38,000 --> 01:21:41,200
and material control updates the status of an incoming component,

2134
01:21:41,200 --> 01:21:43,360
and AI assistant can collect those updates

2135
01:21:43,360 --> 01:21:45,360
and draft a brief situation report

2136
01:21:45,360 --> 01:21:46,880
for the planner and supervisor.

2137
01:21:46,880 --> 01:21:49,360
It can draft a customer service message too,

2138
01:21:49,360 --> 01:21:50,880
not the final decision,

2139
01:21:50,880 --> 01:21:52,640
and it shouldn't invent promises,

2140
01:21:52,640 --> 01:21:55,600
but it can prepare a clear note with the affected order,

2141
01:21:55,600 --> 01:21:57,200
current production position,

2142
01:21:57,200 --> 01:21:59,360
assumptions behind the recovery plan,

2143
01:21:59,360 --> 01:22:01,840
and the date risk that still needs approval.

2144
01:22:01,840 --> 01:22:03,920
That is a much better use of language AI

2145
01:22:03,920 --> 01:22:05,520
than asking it to guess the next sequence

2146
01:22:05,520 --> 01:22:06,880
from a few lines of text.

2147
01:22:06,880 --> 01:22:08,720
Predictive models play a different role.

2148
01:22:08,720 --> 01:22:10,960
They look at patterns in past and current data

2149
01:22:10,960 --> 01:22:12,640
to estimate a future condition,

2150
01:22:12,640 --> 01:22:15,920
maybe a higher risk of machine failure for maintenance signals,

2151
01:22:15,920 --> 01:22:17,120
drifting cycle times,

2152
01:22:17,120 --> 01:22:18,960
the chance of a late material receipt,

2153
01:22:18,960 --> 01:22:20,880
or a quality risk that can hold an order

2154
01:22:20,880 --> 01:22:22,240
before the next operation.

2155
01:22:22,240 --> 01:22:25,520
Those predictions can change a scheduling decision.

2156
01:22:25,520 --> 01:22:27,600
If a resource carries a high failure risk,

2157
01:22:27,600 --> 01:22:30,480
you may avoid loading a customer critical order onto it.

2158
01:22:30,480 --> 01:22:33,120
If a process consistently takes longer than expected,

2159
01:22:33,120 --> 01:22:35,680
the near term schedule needs more realistic duration.

2160
01:22:35,680 --> 01:22:37,200
If material risk increases,

2161
01:22:37,200 --> 01:22:39,600
the scheduler can avoid reserving a scarce slot

2162
01:22:39,600 --> 01:22:41,520
for an order that probably won't be ready,

2163
01:22:41,520 --> 01:22:43,760
but predictions still carry uncertainty,

2164
01:22:43,760 --> 01:22:46,080
and the people using them need to understand that.

2165
01:22:46,080 --> 01:22:47,840
A maintenance risk score does not mean

2166
01:22:47,840 --> 01:22:49,520
the machine will fail at two o'clock,

2167
01:22:49,520 --> 01:22:51,520
and a late material estimate does not mean

2168
01:22:51,520 --> 01:22:53,200
the supplier missed the delivery.

2169
01:22:53,200 --> 01:22:55,200
It means the plan should consider the risk,

2170
01:22:55,200 --> 01:22:57,520
instead of treating the original assumption as certain.

2171
01:22:57,520 --> 01:23:00,560
I've seen plants treat a prediction as gospel,

2172
01:23:00,560 --> 01:23:03,120
and that's how you end up rescheduling around ghosts.

2173
01:23:03,120 --> 01:23:05,600
Scheduling and optimization handle a different task.

2174
01:23:05,600 --> 01:23:07,520
They take the constraints the plant has defined

2175
01:23:07,520 --> 01:23:09,280
and calculate a feasible sequence,

2176
01:23:09,280 --> 01:23:11,120
including actual resource limits,

2177
01:23:11,120 --> 01:23:13,600
operation order, approved alternatives,

2178
01:23:13,600 --> 01:23:15,840
available time, material status,

2179
01:23:15,840 --> 01:23:18,080
and business rules that shape priority.

2180
01:23:18,080 --> 01:23:20,960
The scheduler answers a structured question,

2181
01:23:20,960 --> 01:23:22,240
given what we know right now,

2182
01:23:22,240 --> 01:23:24,080
where can this work run, when can it run,

2183
01:23:24,080 --> 01:23:26,320
and what must move if we choose that option?

2184
01:23:26,320 --> 01:23:29,040
AI can support that process from several angles,

2185
01:23:29,040 --> 01:23:30,640
helping explain a result,

2186
01:23:30,640 --> 01:23:32,800
surfacing patterns from past overrides,

2187
01:23:32,800 --> 01:23:34,720
warning that a condition may change,

2188
01:23:34,720 --> 01:23:36,560
or finding the facts behind an exception.

2189
01:23:36,560 --> 01:23:39,360
But the scheduling engine needs explicit rules

2190
01:23:39,360 --> 01:23:41,920
and constraints because feasibility is not something

2191
01:23:41,920 --> 01:23:43,600
a language model should improvise.

2192
01:23:43,600 --> 01:23:45,200
The roles need to stay clear.

2193
01:23:45,200 --> 01:23:47,280
The AI assistant supports the conversation,

2194
01:23:47,280 --> 01:23:49,440
the predictive model estimates risk,

2195
01:23:49,440 --> 01:23:52,000
the scheduling or optimization engine tests,

2196
01:23:52,000 --> 01:23:53,280
feasible options,

2197
01:23:53,280 --> 01:23:56,640
and the planner remains accountable for the production choice,

2198
01:23:56,640 --> 01:23:58,720
because that choice carries customer,

2199
01:23:58,720 --> 01:24:01,040
safety, quality, labor,

2200
01:24:01,040 --> 01:24:02,800
and commercial consequences

2201
01:24:02,800 --> 01:24:04,720
that no model fully owns.

2202
01:24:04,720 --> 01:24:06,160
Ask yourself one question.

2203
01:24:06,160 --> 01:24:08,720
What data and constraints does the AI actually know?

2204
01:24:08,720 --> 01:24:11,360
Can it see current work completion not just planned?

2205
01:24:11,360 --> 01:24:14,000
Does it know which alternate route has approval today?

2206
01:24:14,000 --> 01:24:17,760
Can it distinguish material on hand from material released for the order?

2207
01:24:17,760 --> 01:24:19,200
Does it know the production policy

2208
01:24:19,200 --> 01:24:22,240
when an urgent service order collides with a setup campaign?

2209
01:24:22,240 --> 01:24:24,640
If the answer is no, the AI should say so.

2210
01:24:24,640 --> 01:24:26,160
That is not a failure of AI.

2211
01:24:26,160 --> 01:24:29,360
It is a boundary around what the system can responsibly recommend.

2212
01:24:29,360 --> 01:24:31,920
A confident answer without the relevant production context

2213
01:24:31,920 --> 01:24:34,160
is just a faster way to spread a bad assumption.

2214
01:24:34,160 --> 01:24:37,120
So don't start with an AI program across the whole plant.

2215
01:24:37,120 --> 01:24:40,080
Pick one planning decision where people already lose time,

2216
01:24:40,080 --> 01:24:42,320
where the constraints are known well enough to model,

2217
01:24:42,320 --> 01:24:44,480
and where a better answer can still change

2218
01:24:44,480 --> 01:24:46,000
what production does next.

2219
01:24:46,000 --> 01:24:48,720
Starts more, validate the logic, and build from there.

2220
01:24:48,720 --> 01:24:50,400
Start with one planning problem.

2221
01:24:50,400 --> 01:24:52,960
When you start digging into finite capacity scheduling,

2222
01:24:52,960 --> 01:24:55,040
don't try to model the whole factory at once.

2223
01:24:55,040 --> 01:24:57,360
Pick one decision, the one that hurts most right now,

2224
01:24:57,360 --> 01:24:58,960
where a better answer could actually change

2225
01:24:58,960 --> 01:25:00,480
what happens this shift or tomorrow.

2226
01:25:00,480 --> 01:25:02,800
Maybe it's sequencing work at a single bottleneck

2227
01:25:02,800 --> 01:25:04,880
or figuring out which late orders to run first

2228
01:25:04,880 --> 01:25:07,600
when multiple customers need the same constrained resource.

2229
01:25:07,600 --> 01:25:09,280
Or maybe the problem is labor capacity

2230
01:25:09,280 --> 01:25:12,000
where your schedule keeps assuming people are available when they're not.

2231
01:25:12,000 --> 01:25:14,560
Keep that first scope tight, even if it feels small.

2232
01:25:14,560 --> 01:25:16,800
A plant-wide model sounds like the right approach,

2233
01:25:16,800 --> 01:25:18,320
but here's what actually happens.

2234
01:25:18,320 --> 01:25:21,200
It drags in every route, every calendar, every exception,

2235
01:25:21,200 --> 01:25:24,000
and every old argument about how work really flows.

2236
01:25:24,000 --> 01:25:25,840
The project turns into a master data debate

2237
01:25:25,840 --> 01:25:28,000
before anyone has seen a schedule they can actually use.

2238
01:25:28,000 --> 01:25:30,880
So narrow it down to a product family with repeatable work,

2239
01:25:30,880 --> 01:25:32,560
a limited group of resources,

2240
01:25:32,560 --> 01:25:35,600
and a decision that planners and supervisors already make every day,

2241
01:25:35,600 --> 01:25:38,240
ideally one with a visible cost when they get it wrong.

2242
01:25:38,240 --> 01:25:41,280
Say the pain lives at one machining cell.

2243
01:25:41,280 --> 01:25:43,120
That cell feeds several assembly orders,

2244
01:25:43,120 --> 01:25:44,480
carries a growing queue,

2245
01:25:44,480 --> 01:25:47,440
and the planner spends part of each day changing the sequence

2246
01:25:47,440 --> 01:25:49,840
because production calls in with another exception.

2247
01:25:49,840 --> 01:25:51,360
You don't need to model every press,

2248
01:25:51,360 --> 01:25:53,920
every warehouse move, or the final packing step on day one.

2249
01:25:53,920 --> 01:25:56,800
You just need to test whether a schedule can give that planner

2250
01:25:56,800 --> 01:25:59,200
a better answer about what should run next.

2251
01:25:59,200 --> 01:26:00,880
Be precise about the decision.

2252
01:26:00,880 --> 01:26:03,120
I want to improve planning is too vague.

2253
01:26:03,120 --> 01:26:05,360
A workable first question might sound like this,

2254
01:26:05,360 --> 01:26:07,040
given the orders ready at this bottleneck,

2255
01:26:07,040 --> 01:26:09,520
which sequence protects the most urgent commitments

2256
01:26:09,520 --> 01:26:12,160
while keeping avoidable changeovers under control.

2257
01:26:12,160 --> 01:26:13,840
That gives people something to test.

2258
01:26:13,840 --> 01:26:15,200
You also need to name the users,

2259
01:26:15,200 --> 01:26:16,640
usually that includes the planner,

2260
01:26:16,640 --> 01:26:17,760
who owns the sequence,

2261
01:26:17,760 --> 01:26:19,840
the supervisor who needs to execute it,

2262
01:26:19,840 --> 01:26:21,600
and probably maintenance, quality,

2263
01:26:21,600 --> 01:26:24,880
or material control when their constraints affect the choice.

2264
01:26:24,880 --> 01:26:27,040
If the schedule only works for the project team,

2265
01:26:27,040 --> 01:26:29,200
it won't survive contact with the morning shift.

2266
01:26:29,200 --> 01:26:31,680
Set the planning horizon before you build the model.

2267
01:26:31,680 --> 01:26:32,880
For a busy bottleneck,

2268
01:26:32,880 --> 01:26:34,960
the useful horizon might be the current shift

2269
01:26:34,960 --> 01:26:36,320
plus the next few days.

2270
01:26:36,320 --> 01:26:38,960
For a slower process with long runs and longer setup work,

2271
01:26:38,960 --> 01:26:40,320
you'll need to look further out.

2272
01:26:40,320 --> 01:26:43,600
The right horizon depends on when a decision still changes the outcome,

2273
01:26:43,600 --> 01:26:45,760
not on how many months your system can display.

2274
01:26:45,760 --> 01:26:46,960
Response time matters too.

2275
01:26:46,960 --> 01:26:50,000
If planners need an answer during a disruption,

2276
01:26:50,000 --> 01:26:52,640
a schedule that takes half a day to recalculate won't help.

2277
01:26:52,640 --> 01:26:54,960
If the use case is a weekly capacity review,

2278
01:26:54,960 --> 01:26:57,520
speed matters less than confidence in the assumptions.

2279
01:26:57,520 --> 01:26:59,520
Define how fast people need an answer,

2280
01:26:59,520 --> 01:27:02,960
then choose your logic and integration timing around that.

2281
01:27:02,960 --> 01:27:05,200
The pilot needs measures, but don't drown it in them.

2282
01:27:05,200 --> 01:27:07,280
Track schedule adherence.

2283
01:27:07,280 --> 01:27:09,600
Did work start and finish close to where you planned?

2284
01:27:09,600 --> 01:27:11,680
Track Q time at the resource under study.