Lesson 5 of 5
Post-conversion testing
11 min
A conversion changes the data model under every finance process, even though most screens look familiar. The temptation is to test only what is visibly new. The risk is in what looks unchanged: a custom report that now reads a compatibility view, an interface that wrote to an old table, a month-end job that relied on a classic asset transaction. Post-conversion testing has to cover data, processes, period-end and reporting.
Build the regression list
The first layer happens on the conversion weekend (lesson 1). The others need test cycles in a converted copy of production, usually several times before the real conversion, and once more in the dress rehearsal. Each cycle should end with a full period-end close, because close is where most conversion surprises appear.
| Area | Regression test | Expected result |
|---|---|---|
| G/L | Trial balance for company code 1710 in leading and any non-leading ledgers | Identical to the ECC snapshot for the converted period |
| AP and AR | Open item totals and counts per account; one payment run and one incoming payment | Totals match; clearing works on migrated items |
| Assets | Asset history sheet per area; one depreciation run | Values match; depreciation posts per asset |
| Inventory | Stock value by plant and material; one goods movement | Values match; material ledger updates |
| Controlling | Cost center actuals; one assessment cycle; one order settlement | Actuals match; allocations post to secondary cost accounts |
| Credit | Credit exposure for top customers; one order that should block | Exposure matches; order blocked under check type D |
| Reports | Top 20 finance reports by usage, including custom | Totals agree with ECC for the same period and selections |
A compact finance regression checklist for a conversion.
Find the reports that matter
You cannot regression test every report Contoso has written since 2005. Use usage data: the workload statistics in ECC show which transactions and reports finance users actually run. Take the top reports by use and every report that feeds an external party (tax, consolidation, banks, auditors). For each, run it in ECC and in the converted system for the same period and selections, and compare totals. A custom report that reads old index tables through a compatibility view may be correct but slow; time it as well.
- Does it still run without errors?Yes → next question · No → fix code; it may write to or depend on a removed table
- Do its totals match ECC for the same period and selections?Yes → next question · No → defect; check which table or view it reads
- Does it run in an acceptable time?Yes → pass · No → performance defect; consider a standard app or a view-based rewrite
- Is there a standard app that now does the same job?Yes → consider retiring the custom report after go-live · No → keep and document
- Mock 1First conversion of a production copyReconciliation method proven; large differences found and fixed
- Mock 2Second conversion with fixesRegression list executed; first period-end close on converted data
- Dress rehearsalFull timing runReconciliations at zero; close repeated; reports compared
- Go-live weekendProduction conversionReconciliations signed by controllers; Dan signs finance
- First closeHypercareFirst real month-end on S/4HANA with extra checks and daily triage
A word on timing. Finance migration activities can take hours on a large system, and they sit on the critical path of the conversion weekend. Each mock conversion should record how long each finance activity and each reconciliation took, so Luis can plan the real weekend with realistic durations and float.
What would you do?
In mock 2, all reconciliations are zero, but the first period-end close on converted data fails at foreign currency valuation because a custom program writes to a table that no longer exists. Dress rehearsal is in four weeks. What should the team do?
- Regression scope: migrated data, daily processes, integrated processes, period-end close, reporting and interfaces.
- Every test cycle should end with a full period-end close on converted data.
- Select reports by usage and external dependency; compare totals and timing with ECC.
- Record durations of finance migration activities in each mock for cutover planning.
- Sign-off rests on zero reconciliations, passed regression, a successful close and matching reports.
Check your understanding · not graded
Lesson check
1. Which regression area most often reveals conversion surprises?
2. How should a team choose which custom finance reports to regression test?
3. A custom report runs and totals match ECC, but it now takes 40 minutes. How should it be recorded?
4. Why record timings in each mock conversion?