Lesson 6 of 6
Parallel runs and consolidation sign-off
10 min
Listen to this lesson
A consolidation tool is proven when it produces the same group result as the system it replaces, or when every difference is explained. The way to show that is a parallel run: close the same quarter in BPC and in Group Reporting, compare, explain, and sign. Contoso runs two parallel quarter-ends, because a single quarter rarely exercises every rule: Q4 brings year-end entries, and Q1 tests the balance carryforward.
What to compare
| Level | Compare | Expected result |
|---|---|---|
| Reported data | Per unit and FS item: BPC input versus ACDOCU level 00 | Identical; differences mean mapping or release errors |
| Translation | Per FS item for 1910: USD amounts and the CTA | Identical if rates and methods match; otherwise explained by method |
| Eliminations | Per pair of units and FS item: eliminated amounts and differences | Identical, or explained by matching improvements |
| Group statements | Consolidated balance sheet and income statement by group FS item | Within the agreed tolerance, every difference explained |
| Management view | Profit center and segment figures at group and business level | Explained where BPC used different allocation logic |
The five comparison levels of a consolidation parallel run.
Compare from the bottom up. A difference in the consolidated balance sheet is hard to explain; the same difference at reported data level for one unit is easy. If reported data matches, look at translation; if translation matches, look at eliminations. Most parallel-run differences at Contoso-sized groups come from three sources: FS item mapping that differs from the BPC account mapping, rate types that differ from what BPC used, and elimination logic that BPC handled in script and Group Reporting handles by task.
Tolerances and classification
Agree tolerances before the run, not after. Contoso's (illustrative) rule: reported data and eliminations must match exactly; translated amounts may differ by up to USD 1.00 per FS item from rounding; group statement lines may differ only where a classified difference explains them. Every difference gets one of four classes, and the class decides the next step.
| Class | Meaning | Action |
|---|---|---|
| Defect in Group Reporting | Wrong mapping, method or task setting | Fix, rerun, retest |
| Defect in BPC | The legacy result was wrong | Document; Dan decides whether a prior period needs correcting |
| Intended design change | New logic agreed in design, such as different elimination timing | Reference the design decision; no fix |
| Rounding | Within the agreed tolerance | Record and accept |
Classifying parallel-run differences.
Running the parallel quarter
- Freeze inputsSame trial balances, rates and ownership data into both systems
- Close in bothBPC as usual; Group Reporting by the documented task sequence
- Compare by levelReported, translated, eliminated, consolidated, management view
- Classify and fixEvery difference classified; Group Reporting defects fixed and rerun
- Sign the quarterDan signs the comparison pack; open items have owners and dates
Freezing inputs matters more than it sounds. If BPC receives an extract on day 2 and Group Reporting releases ACDOCA on day 3 after a late posting, the systems compare different data. Both systems must work from the same trial balances, the same exchange rates and the same ownership percentages, and the evidence pack should say so.
What would you do?
In the Q4 parallel, everything matches except the CTA for 1910, which differs by USD 18,200. Reported data and translated balance sheet items match. What should the team check first?
When the data path changes mid-parallel
Contoso's two parallel quarters straddle the Wave 2 data migration. In the Q4 parallel, Alternadores (1910) is still a non-integrated unit: its trial balance comes from the legacy ERP by flexible upload into Group Reporting and by the usual extract into BPC. By the Q1 parallel, 1910's books are in SAP S/4HANA, and its reported data arrives through the release task. That means the Q1 parallel is the first test of 1910 as an integrated unit, with new failure modes: G/L accounts created during migration that are not mapped to FS items, and intercompany balances with Contoso that need trading partners on every migrated open item.
Dan's team treats the switch as a separate test inside the parallel run. They change the consolidation unit's data collection settings in the right period only, check that Q4 data uploaded by file is not overwritten when Q1 data is released, and reconcile the 1910 opening balances in Group Reporting with the balances migrated into the general ledger. The Wave 2 migration's financial reconciliation (sap-dmg-302) and the consolidation parallel run should use the same key date and the same figures, so that one reconciliation does not contradict the other.
What Dan signs
- The comparison pack for the quarter, by level, with every difference classified and closed or owned.
- The task logs and status of both monitors, with completed tasks blocked.
- Evidence of the frozen inputs: trial balances, exchange rates and ownership data used in both systems.
- The rule inventory showing each BPC rule's Group Reporting equivalent was exercised.
- After the second clean quarter: the decision to make BPC read-only, co-signed by Grace's steering group.
What goes wrong in testing
| Pitfall | Effect | Prevention |
|---|---|---|
| Comparing only consolidated totals | Offsetting differences hide each other | Compare by unit, FS item and posting level |
| Rates loaded separately in each system | Translation differences that mean nothing | One rate file feeds both systems |
| Fixing Group Reporting to match a BPC error | The old defect is rebuilt in the new system | Classify first; never fix to match without a decision |
| No owner for differences | Items stay open into the next quarter | Every difference has an owner and a due date |
| Parallel run in a test client | Results do not reflect production data | Run the parallel on production data, in production or a fresh copy |
Parallel-run pitfalls seen in consolidation moves.
- Run at least one parallel quarter-end, ideally two that include a year-end.
- Compare bottom up: reported, translated, eliminated, consolidated, management view.
- Agree tolerances first and classify every difference.
- Freeze inputs so both systems consolidate the same data.
Check your understanding · not graded
Lesson check
1. Why does Contoso run two parallel quarter-ends instead of one?
2. A group balance sheet difference appears. Where should analysis start?
3. A difference turns out to be a BPC script error that existed for years. How is it classified?
4. Why must inputs be frozen for a parallel run?