Lesson 4 of 6
Comparison testing
6 min
In a system conversion, the strongest evidence that nothing was lost is a comparison: the same numbers, from the same data, before and after. If the trial balance, open receivables and stock value in ECC equal those in the converted system, the business can trust the conversion. If they do not, the difference must be explained before anyone signs.
Comparison testing has two parts. Data comparison checks that what was converted matches what existed. Process comparison (sometimes called a parallel run) checks that the same business transaction, run in both systems, produces the same result. Process owners define both, because only they know which outputs matter.
What to compare
| Area | Comparison | Typical tolerance |
|---|---|---|
| General ledger | Trial balance by company code, ledger and account, in each currency | Zero difference |
| Receivables and payables | Open items count and amount by customer or supplier and reconciliation account | Zero difference |
| Inventory | Stock quantity by material, plant and storage location; stock value by valuation area | Quantity zero; value zero, or documented rounding |
| Assets | Acquisition value, accumulated depreciation and net book value by asset class | Zero difference |
| Controlling | Cost center and profit center balances for the period | Zero difference |
| Open logistics documents | Open sales orders, deliveries and purchase orders: count and value | Zero difference |
| Master data | Counts of customers, suppliers (business partners), materials and their key fields | Zero, with documented exclusions |
A starting comparison set for a conversion. Tolerances are set before the first mock, never after.
Most tolerances should be zero. Balances do not drift by themselves. Where a non-zero tolerance is justified, for example rounding in a revalued currency, write down why, how much and who approved it before any comparison runs. A tolerance set after seeing the difference is not a tolerance; it is a waiver.
- Is the difference zero or within the tolerance agreed before the mock?Yes → record the result · No → continue
- Is it explained by the report, selection or key date used?Yes → fix the comparison method and rerun · No → continue
- Is it a data or conversion error?Yes → fix at source and re-verify in the next mock · No → continue
- Is it a legitimate, understood difference the owner accepts?Yes → add to the accepted-differences register · No → keep investigating; it blocks sign-off
When to take the snapshots
- Before downtimePre-conversion snapshot in ECCBusiness closes the period as planned; extract comparison reports with the same selection as later runs.
- Technical conversionSystem convertedIT runs the conversion; finance data migration steps run and are monitored.
- After conversionPost-conversion snapshotSame reports, same selections, run in the converted system.
- Same dayCompare and explainReconcile line by line; classify each difference; fix, accept or investigate.
- Before go decisionOwner signs comparisonResults and accepted differences go into the sign-off record.
Do this in every mock conversion, not just at cutover. Mock conversions run on copies of production, so they find the data issues that only real data reveals. By the final mock, the comparison should be routine and fast, because cutover is the worst time to discover how long it takes.
Process comparison
For process comparison, pick a small set of high-risk transactions and run them in both systems with the same inputs: a copy of ECC and the converted test system, from the same production copy. Compare the outputs that matter. For Contoso that includes the price and tax on a sales order for Great Lakes Fleet Services, the journal entry for a goods receipt of CMC-RTR-200 from Lakeside Components, a billing document for CMC-ALT-100, and a depreciation run for one asset class.
- Same input data and dates in both systems.
- Compare business outputs (amounts, accounts, quantities, statuses), not screen layouts.
- Expect some structural differences in S/4HANA, such as journal entry line items carrying more attributes; compare amounts and accounts, not document structure.
- Record each comparison with both document numbers as evidence.
What would you do?
In Contoso's second mock conversion, the trial balance matches exactly, but open receivables are 42,000 USD lower after conversion than before. IT says the difference is "probably timing." What should Dan do?
- Data comparison checks converted data; process comparison checks that transactions give the same results.
- Compare trial balance, open items, stock, assets, controlling balances, open documents and master data counts.
- Set tolerances before the first mock; most should be zero.
- Use the same reports and selections before and after, in every mock conversion.
- Every difference must be explained and fixed or formally accepted.
Check your understanding · not graded
Lesson check
1. What is the usual tolerance for a trial balance comparison?
2. When should tolerances be set?
3. What is the most common cause of false differences?
4. What should process comparison compare?