Onboarding · 3 min
Welcome to Payroll.COMPAIRE.
You're about to run a parallel payroll test. Two systems, tens of thousands of variances, one pay cycle. Before you ingest your first file, let's walk through how this tool turns a three-week review into an afternoon.
Prefer to read? Every step on the left rail covers the same ground — ~3 minutes end-to-end.
Step 2 · The problem
Parallel runs are painful.
Migrating payroll systems means running the old one and the new one side-by-side for three to six cycles. Every cycle produces thousands of variances. Without tooling, reviewers quit by cycle two.
Three things nobody ships without
- Classification. Which variances are real defects vs. rounding, timing, or legitimate differences?
- Explainability. For each variance, the exact rule chain that led to the verdict — auditors insist.
- Reviewer continuity. What you decided in cycle 1 must roll forward into cycles 2, 3, 4.
Step 3 · Personalize
Name your two systems.
Parallel payroll testing always means two systems: the one you're migrating from (Legacy) and the one you're migrating to (New). Tell us what to call yours — we'll use these names throughout the rest of the walkthrough.
Common: Workday, ADP, Paychex, UKG, PeopleSoft, Ceridian, Paycom, Oracle HCM
Common: SAP, Workday, UKG Pro, Paycom, Oracle Fusion, ADP Lifion
Why this matters
Payroll.COMPAIRE is system-agnostic — the engine doesn't care whether you're moving from Legacy System to New System, from ADP to Workday, or between any other pair. Setting the names here just makes the rest of this onboarding match your actual project. You can change them any time from the Admin page later.
Leave blank and we'll just use "Legacy System" and "New System" everywhere.
Step 4 · Taxonomy
Four dimensions. Every variance.
Every variance in the system carries four independent labels. Confusing them is the single biggest failure mode of home-grown tools. You'll see these four everywhere.
Direction matters, independently
Alongside those four, every variance also carries a Direction — TARGET_HIGHER means the new system is paying more (overpayment / tax risk); TARGET_LOWER means it's paying less (DOL / FLSA risk). Auditors filter by this first, every time.
Step 5 · Ingest
Drop files in. See results.
Export your Legacy System pay-calc results and your New System wage-type report as CSVs. Drag both into the File Import page. We handle encoding, column normalization, and mapping — you see a preview, hit Ingest, and watch the classifier run.
What happens in seconds
- Rows pair on (employee_id, canonical_code); orphans become
MISSING_IN_*classifications - Each pair runs through the Rule Network — a YAML decision tree, first-terminal-match wins
- Tolerance, known-differences, calc-method rules, timing-pair detection apply in order
- Every record gets a SHA-256 manifest so auditors can verify bit-for-bit reproducibility
- If this is cycle 2+, Reason Roll Forward carries prior dispositions and defect links automatically
Step 6 · Overview
Read the whole run at a glance.
The Dashboard is your triage surface. KPIs at the top, classification mix on the left, severity strip and over/under diverging bar on the right, Pareto of worst wage types at the bottom. Every chart is filterable and exports to the audit package.
Where reviewers start
Click the red segment on the classification bar → you're in the Variances Browser filtered to MATERIAL_VARIANCE. That's the queue. Sort by abs variance, top-down. Most reviewers clear 80% of material variances in under two hours.
Step 7 · Drill down
Every variance. Every employee.
Open any employee to see a side-by-side pay stub — Legacy System on the left, New System on the right, delta column in the middle. Click any line with a variance to expand the full explainability trace: every rule that evaluated, which matched, and which terminal classification it triggered.
The explainability trace
On that FIT line, expanding reveals: both_zero=F → target_null=F → mapping_unresolved=F → category_mismatch=F → integrity_fail=F → matches_known_diff=F → matches_calc_method=T → CALCULATION_METHOD_DIFFERENCE (CM01 bonus-tax aggregate vs 22% supplemental). That's the whole decision, visible. Auditors love it.
Step 8 · Workflow
Open defects. Resolve them.
When a variance isn't explainable by a known rule, open a defect. One defect can link many variances across many employees. The state machine walks it through fix → validation → resolution, with history and comments preserved across cycles.
Roll forward, automatically
When you ingest cycle 2, every persisting variance carries its cycle-1 disposition, comments, assignee, and linked defect IDs forward. Reviewers typically see a 60–70% time cut from cycle 2 onward.
Step 9 · Readiness
GO or NO-GO. Measured.
Eleven exit criteria, each with a numeric threshold. Live pass/fail against the current pay run. When all eleven are green, the system recommends GO. Override requires a second approver with a documented reason.
In this example, EC05 fails — 3 employees out of 2,000 have material net-pay variance. Click the row to jump straight to those employees and triage.
Step 10 · Ship it
Sealed sign-off.
When all criteria pass, click Generate sealed sign-off PDF. The signed document captures every count, manifest hash, criterion, signer, and timestamp — and seals the whole payload with a SHA-256 fingerprint. Any tamper invalidates the hash.
Pay run PR-2026-04 · Tenant gmt-consulting · Recommendation GO
What lands in the audit ZIP
- The sealed sign-off PDF (above)
- The 9-sheet detailed Excel — every variance, every rollup, every mapping
- The executive PDF — boardroom-ready KPIs + classification mix + exit criteria
- A
manifest.jsonwith every input-file SHA-256 and config version
Step 11 · You're set
Now, go compare.
That's the ritual. Ingest → dashboard → drill down → open defects → exit criteria → sign-off. The first cycle takes a few hours to learn; cycles 2 onward run in under an hour.