Operational Excellence KPIs: Which Metrics Matter on the Plant Floor

Disclosure: M365.FM host Mirko Peters works at iFAKT GmbH. This page describes the company factually; paid partnerships and links are labelled as advertising.

By Mirko Peters · 2026-11-11

Operational excellence KPIs show whether a plant serves customers and improves. See which metrics matter on the plant floor, how to define and use them.

Operational excellence KPIs are the few measures that show whether a plant serves its customers well and keeps getting better. This article is for plant managers and operational excellence leads who must choose metrics for the shop floor. It explains which groups of measures matter, how to define the main ones, how to use leading and lagging indicators, a worked example with numbers and the mistakes that turn a scorecard into noise.

Key points

• Choose a few measures in four groups: delivery, quality, flow and equipment, and cost and safety. Tie each one to a goal.
• OEE is availability times performance times quality. Use it with other measures, not alone.
• On-time delivery is on-time units divided by total units. First pass yield is good units divided by units entering the process.
• Results are usually measured with lagging KPIs. Add a few leading measures that the team can change this week.
• A measure needs a clear definition, a target, an owner and a response when it turns red.

Which operational excellence KPIs matter on the plant floor?

There are dozens of manufacturing metrics. MachineMetrics lists over 50 and says that manufacturers can and should choose several to focus on. FourJaw adds that the right metrics depend on the production environment, and that you should define strategic goals first, then choose the metrics that track progress toward them. Both are software vendors, so treat the lists as a menu, not as a rule. Good performance management starts with a short list that you align with the goals of the plant.

A simple structure keeps the choice small. Choose a few measures in four groups.

• Delivery. Does the customer get what was promised, on time? Examples: on-time delivery and schedule attainment.
• Quality. Is the work right the first time? Examples: first pass yield and scrap rate.
• Flow and equipment. Does the work move, and do the machines run? Examples: overall equipment effectiveness and lead time.
• Cost and safety. What does it cost, and is it safe? Examples: unit cost and safety incidents.

The aim is key performance indicators that give actionable signals, not a long report. Safety belongs on every board, even if it is not a classic production measure. It tells people what the plant values.

How do you define each metric?

A measure without a clear definition causes arguments. Write the formula, the data source and the period. The formulas below follow MachineMetrics.

• Overall equipment effectiveness (OEE): availability × performance × quality. Availability is actual production time divided by planned production time. Performance is actual production rate divided by ideal rate. Quality is good parts divided by total parts.
• On-time delivery: on-time units divided by total units.
• First pass yield: good units completed, excluding scrap and rework, divided by total units entering the process.
• Scrap rate: scrap material divided by total material intake.
• Manufacturing lead time: the order delivery date minus the date the order was received.
• Schedule attainment: actual output divided by target output, times 100.
• Cycle time: total parts produced divided by production run time.

Some sources describe the same measure in different ways. FourJaw defines first pass yield as the percentage of products that pass inspection on the first attempt. That is close in meaning, but the exact count differs. So write down your own definition, and use it everywhere.

OEE needs a special warning. MachineMetrics says it should ideally be used alongside other measures. A plant can raise OEE by running products that are not needed. A high OEE with poor delivery is a signal that the plant optimizes the wrong thing.

What are leading and lagging indicators?

A lagging measure reports a result after the fact. On-time delivery for last week and the scrap rate of last month are lagging. They tell you what happened, and they come too late to steer the day. They still matter for customer satisfaction, because the customer feels the result.

The Shingo Institute says on its model page that results are usually measured with lagging KPIs. An overview of the model adds that lagging KPIs can signal that the culture needs attention. So a poor result is often a symptom of how people work, not only of a machine.

A leading measure shows something the team can change now. These are examples of common practice, not terms from a standard:

• Plan adherence for the current shift.
• Number of open stops at the line.
• Share of daily meetings held with the board.
• Age of the oldest open problem.

A good board has both, and it supports continuous improvement because the team sees the effect of each change. The lagging measures show the result, and the leading ones tell the team what to do this morning.

How many measures should you track?

No source gives a fixed number, and the right answer depends on the team. As a working rule, use three to five measures per area, and no more than a team can review in a short daily meeting. If a measure is not discussed and nobody acts on it, remove it.

Use these tests for every candidate:

1. Does it link to a goal? If the plant wants faster delivery, delivery measures come first.
2. Can the team influence it? A number that nobody can change only creates frustration.
3. Is it easy to read at a glance? Show the target next to the actual value.
4. Is the data reliable? If people do not trust the number, they will not use it.
5. Is there a response? Agree on who acts and by when if the value is red.

How do you use the measures every day?

Measures help only if they are part of a routine.

1. Show them where the work is done. A board at the cell, at eye level, not in the office.
2. Review them in a short daily meeting. Look at yesterday's result and today's plan.
3. Ask why, not who. Look at the process before you look for a person.
4. Fix the top problem first. One problem at a time, with an owner and a date.
5. Review the set every month. Remove what does not help, and add one measure when the team is ready.

For the first board, three to five measures are enough. A team that uses five measures well gains more than one that ignores twenty.

Example: reading three measures

This is an illustration with made-up numbers.

A line has 480 minutes of planned production time. It stops for 60 minutes, so it runs for 420 minutes. The ideal rate is one unit per minute. The line makes 357 units, and 340 of them are good.

• Availability: 420 ÷ 480 = 87.5%.
• Performance: 357 ÷ 420 = 85%.
• Quality: 340 ÷ 357 = about 95.2%.
• OEE: 0.875 × 0.85 × 0.952 = about 70.8%.

As a check: 340 good units divided by the 480 units that the ideal rate would give in planned time is also 70.8%.

The biggest loss is performance, at 85%, followed by availability. The team looks at the stops first, then at slow cycles. This is how operational metrics guide the next improvement, and operational performance rises step by step.

Now delivery and quality. The plant ships 400 orders, and 352 are on time.

• On-time delivery: 352 ÷ 400 = 88%.

A process has three steps. First pass yields are 95%, 97% and 98%.

• Rolling throughput yield: 0.95 × 0.97 × 0.98 = about 90.3%.

Each step looks good alone. Together, about 10 of every 100 units need rework or are scrapped. The numbers are assumptions. Use your own definitions and your own data.

Common mistakes

• Too many measures. A scorecard with dozens of numbers hides the few that matter. Pick a short list.
• No definition. If two shifts calculate a measure in different ways, the board starts arguments, not actions.
• OEE alone. Equipment effectiveness does not show whether the right products are made on time.
• Measures without a response. A red number that nobody answers teaches people to ignore the board.
• Targets without understanding. Ask what you are really optimizing for before you set a target. A target alone changes behaviour, not the process.
• Only lagging measures. The team learns about problems when it is too late to act.
• Data that nobody trusts. Check the numbers against what you see at the line.

How software can help

A paper board is a good start. It falls behind when plans change during the day, and when data comes from several systems. iFAKT's Polaris Analyzer provides KPI dashboards across all modules that compare planned and actual values in production. Treat any dashboard as a visual tool for the routine: it needs a definition, a target and a response, and it must be used at the meeting. Real-time data helps only if people trust it.

Next step

Pick one area, choose three to five operational excellence KPIs with a clear definition and a target, and show them at the line. Then hold the first daily review and decide who answers when a measure turns red.