Lesson 6 of 6
The data sign-off evidence pack
9 min
Listen to this lesson
Everything in this course ends in one artifact: the data sign-off evidence pack. It is what Dan signs before the margin dashboard goes to the CFO, what Grace shows the steering group, what the auditors ask for, and what the finance agent team relies on before an agent reads the model. Without it, the data layer is trusted on faith.
Contents
| Section | Contents | Owner |
|---|---|---|
| 1. Scope and lineage | Sources, replication objects, data products (with versions), views, models, consumers; one lineage diagram | Noor |
| 2. Strategy and tolerances | Signed strategy from lesson 1 | Priya, Dan |
| 3. Replication results | Initial load reconciliation; delta tests (insert, update, delete); stop-and-resume test; flow run IDs and timestamps | Priya |
| 4. Reconciliation results | Counts and totals by dimension per hand-off for the closed period; key-level diffs and their resolutions | Priya, Wei |
| 5. Model results | Rule checks with known cases; model-to-ledger reconciliation; access tests | Priya, Dan |
| 6. Regression baseline | Baseline values, suite version, triggers and calendar | Noor |
| 7. Defects and open items | All defects with keys, cause, fix and retest; open items with owner and date | Priya |
| 8. Operations | Monitoring owners, daily checks, alert routes, KBA and release-note watch | Noor |
| 9. Decision | Accept, accept with conditions, or reject; signatures and date | Dan, Priya, Noor |
Contoso's data sign-off evidence pack. Store it with the release in SAP Cloud ALM.
Writing it so it survives an audit
- Every result has a timestamp, a run or job ID and the exact query or check used.
- Every difference has a classification and a resolution; "under investigation" is an open item, not a pass.
- The data product versions and source release (SAP S/4HANA 2025, support package level) are stated, so a later reader knows what was tested.
- Screenshots are optional; query outputs exported with their selection criteria are not.
- The pack names what was not tested, such as company code 1910, which joins the model in Wave 2.
Conditions and their limits
Some issues may be accepted as conditions: a cosmetic description error, a slow refresh outside business hours, a non-critical attribute with a known fix date. Some may never be: an unexplained difference in finance amounts, a failed access test, a delta that loses changes. Contoso's rule, reused from migration sign-off (sap-dmg-302): unexplained financial differences always block; every condition has an owner, a deadline and a tracking item.
- Is any financial amount difference unexplained?Yes → reject · No → next
- Did any access or delta-loss test fail?Yes → reject · No → next
- Are all open items owned, dated and non-financial?No → reject or resolve · Yes → accept with conditions
- Are there no open items?Yes → accept
Contoso stores the pack with the release in SAP Cloud ALM: the test cases hold the stored query pairs, each test run attaches its outputs with timestamps, defects carry key lists, and the decision is recorded as an approval on the release. After go-live, the same structure is reused: each production regression run adds a short record to the pack, so a year later an auditor or a new team member can see what was proven, when, and against which versions. When company code 1910 joins the margin model in Wave 2, Lucía, the Monterrey O2C key user, adds the MXN cases and the pack is extended rather than rewritten.
What would you do?
Two days before the margin dashboard goes live, one open item remains: customer group descriptions are blank for 12 aftermarket customers; amounts reconcile and all access and delta tests passed. What should Dan decide?
Finally, remember who reads the pack. Dan reads the decision, the finance reconciliations and the conditions. An auditor reads the method, the timestamps and the versions. The next tester reads the stored queries and the baseline. A finance agent team reads the scope, lineage and known limitations before it lets an agent use the model. Write a one-page summary at the front that each of them can use, and keep the detail behind it.
- The evidence pack is the single artifact behind a data sign-off decision.
- It covers scope and lineage, strategy, replication, reconciliation, model results, regression baseline, defects, operations and decision.
- Every result is timestamped, rerunnable and classified; untested scope is stated.
- Unexplained financial differences, failed access tests and delta loss always block.
Check your understanding · not graded
Lesson check
1. Which item can be accepted as a condition under Contoso's rules?
2. What makes evidence rerunnable?
3. Why should the pack state what was not tested?
4. When should the evidence pack be built?