Stop Reconciling Spreadsheets: Building a Unified HR Lakehouse with Microsoft Fabric

By Mirko Peters · 2026-08-11

Welcome back to the podcast! If you have ever spent hours manually matching rows between a payroll export, a human resources information system, and an IT offboardi…

Welcome back to the podcast! If you have ever spent hours manually matching rows between a payroll export, a human resources information system, and an IT offboarding spreadsheet just to figure out why your headcount numbers don't match, you are not alone. HR and people operations teams everywhere are drowning in disconnected datasets. In this comprehensive companion blog post, we are going to expand on our latest episode and walk through exactly how you can build a unified HR lakehouse using Microsoft Fabric. By the end of this guide, you will see how moving away from siloed spreadsheets and adopting a modern architecture can give your organization a single source of truth for real decision-making.

To dive deeper into the audio discussion and hear us break down these concepts in real-time, make sure to check out the related podcast episode: HR Analytics with Fabric and Dynamics 365 Human Resources.

The problem in one sentence

Siloed HR, payroll, and time data create mismatched numbers, slow decisions, and missed risk signals.

Reference blueprint (Fabric + D365 HR)

Ingest & stage

• Sources: Dynamics 365 HR (employees, jobs, hires/terms, org), payroll/benefits SaaS, time/leave, IT identity/offboarding, engagement/surveys (eNPS), recruiting/ATS.
• Pipelines: Dataflows Gen2 for APIs/flat files; scheduled full + incremental loads.
• Storage: OneLake / Lakehouse layers
  • Bronze: raw snapshots (as-is)
  • Silver: cleaned/standardized (IDs, dates, codes)
  • Gold: conformed HR model for analytics

Model & serve

• Conformed entities: Employee, Job, OrgUnit, Hire, Termination, LeaveBalance, LeaveRequest, TimeEntry, CompensationBand, DiversityAttributes, AccessProvisioning, SurveyResponse.
• Keys & mapping: canonical EmployeeId (crosswalk for HR/payroll/IT GUIDs), PositionId, ManagerId, OrgPath.
• Semantic layer: certified Power BI dataset with row-level security by region/org role.

Consume & act

• Dashboards: HR control panel + role-based views (HRBP, TA, Payroll, DEI, Exec).
• Alerts & automation: Power BI/Power Automate for SLA breaches, leave anomalies, offboarding gaps.
• Predictive: AutoML/Azure ML for attrition/burnout risk and onboarding delay forecasts.

Must-have dashboards & metrics

1) Headcount & movement

• Active HC (FT/PT/contract), net change, joins/exits, internal mobility, vacancy ratio.
• Quality check: HR vs IT accounts (ghost users), payroll sync deltas.

2) Onboarding SLA

• Time-to-hire, time-to-start, provisioning completion (accounts, devices, apps), day-1 readiness.
• Alerts: any step > SLA; missing manager check-ins at day 7/30/60/90.

3) Leave & availability

• Accruals vs balances, planned vs unplanned, overtime/TOIL, leave spikes by team/season.
• Controls: negative balances, payroll mismatch, cost center drift.

4) Attrition & retention

• Voluntary/involuntary rates, tenure bands, regrettable loss, exit reasons.
• Leading indicators: absence surge, low eNPS, stalled promotion velocity, unmanaged overtime.

5) DEI insights

• Representation by level/org/location, hiring funnel drop-offs, pay-band distribution.
• Privacy: small-n suppression & aggregation thresholds.

6) Compliance & audit

• Offboarding SLA (account disablement), mandatory training completion, data access logs.

Gold-layer data model (starter set)

Facts:

• FactHeadcountSnapshot, FactHire, FactTermination, FactOnboardingStep,
  FactLeaveBalance, FactLeaveRequest, FactTimeEntry, FactCompChange, FactSurveyScore

Dims:

• DimEmployeeMasked, DimOrgUnit, DimJob, DimManager, DimGeo, DimDiversityAttr, DimCalendar

Lineage links:

Employee ↔ Job/Org over time (SCD2); Hire → Provisioning → FirstLogin; Leave Accrual ↔ Payroll Posted.

Dataflow patterns (that stop reconciliation hell)

• ID crosswalk: build once; persist as a gold table to join HR/Payroll/IT identities.
• Code normalization: job, band, leave types, cost centers to canonical lists.
• Date conformance: UTC, fiscal calendar, effective-dated SCDs (employment, comp).
• Incremental refresh: only changed rows (hire/term events, balances).
• Quality gates: duplicate employee IDs, orphan managers, negative balances, future-dated terms.

Security, privacy & governance

• PII minimization: exclude DOB, SSN; use masked DimEmployee.
• RLS: by HRBP portfolio, region, or org tree; sensitivity labels on gold/semantic.
• Small-n protection: suppress DEI slices < threshold; show “insufficient data”.
• Data contracts: SLAs for freshness, schema change policy, owner + steward per table.
• Audit: log data refreshes, access to sensitive visuals, and all export events.

Alerts & automations (copy/paste starters)

• Onboarding breach: step > SLA ➜ notify manager + IT; auto-create ticket.
• Leave anomaly: sudden spike > X% vs baseline ➜ alert HRBP; review staffing risk.
• Offboarding gap: termed but active M365 account ➜ immediate disable + confirmation.
• Headcount mismatch: HR active ≠ Payroll paid ➜ reconciliation task to Payroll Ops.
• Overtime fatigue: OT > threshold for 3 weeks ➜ manager nudge + wellness resources.

Predictive add-ons (when foundation is stable)

• Attrition risk model: features = tenure, comp delta vs band, manager span, absence & OT, survey trend, promotion wait.
• Onboarding delay forecast: predict bottleneck steps by role/site/vendor.
• Burnout early warning: combine OT, weekend work, low eNPS, leave deferral.

Principles: explainable features, fairness checks, opt-in pilots, human review before action.

14-day quick start

Days 1–3 Connect sources (D365 HR, payroll export/API, time/leave).
Days 4–6 Build ID crosswalk; silver cleans; gold conformed entities.
Days 7–9 Publish certified semantic model with RLS; headcount + onboarding dashboards.
Days 10–11 Wire alerts: onboarding SLA, offboarding gap.
Days 12–14 Reconcile with Payroll/IT; document data contracts; exec demo & next-step backlog.

Common pitfalls (and fixes)

• Mismatched IDs: create & maintain the crosswalk (don’t rely on names/emails).
• Spreadsheet drift: lock certified datasets; discourage ad-hoc extracts.
• PII sprawl: mask employees in dims; apply labels + export controls.
• Over-modeling day 1: ship HC + onboarding SLAs first; iterate to leave/DEI/attrition.
• Alert fatigue: start with 3–4 high-value rules; add severities & cooldowns.

FAQs

Q: Can we do this without Dynamics 365 HR?
A: Yes—use Dataflows Gen2 for any HRIS/payroll/ATS exports or APIs; the lakehouse pattern stays the same.

Q: How often should data refresh?
A: Headcount/DEI daily; onboarding/leave every 1–4 hours; compliance/offboarding near real-time.

Q: How do we protect sensitive DEI data?
A: Aggregate at safe levels, apply small-n suppression, and enforce RLS + sensitivity labels.

Q: Do we need data scientists for predictions?
A: Start with AutoML and interpretable features; graduate to Azure ML as needs grow.

Conclusion

Building a unified HR lakehouse in Microsoft Fabric is the definitive cure for spreadsheet reconciliation fatigue. By systematically moving your data through the Bronze, Silver, and Gold layers, you eliminate data silos, safeguard sensitive employee information, and unlock reliable operational dashboards. To hear a thorough discussion on implementing these techniques and to get even more insider tips, make sure to listen to our accompanying episode HR Analytics with Fabric and Dynamics 365 Human Resources. Transform your HR analytics today, ditch the endless manual cross-checks, and start driving strategic organizational decisions with absolute confidence!