Does a 200-record gap make the cutover safe?
Explain a technical trade-off without hiding offsetting errors.
The question
Ten producer teams are migrating a shared event archive. The source contains 1,000,000 unique event IDs. The target has 999,800 rows. A duplicate is an extra row beyond the first row for an event ID. An AI summary calls this a 0.02% gap and recommends cutover.
| Records | |
|---|---|
| Source unique event IDs | 1000000 |
| Target rows | 999800 |
| Duplicate extra rows in target | 300 |
Each valid source ID should appear once. There are no unexpected target IDs outside the source set.
Zero missing source IDs, at most 50 duplicate extra rows per million source IDs, accepted reconciliation from every producer and a tested rollback. A row-count match alone does not pass.
“Reconcile the offsetting errors and state the evidence required to cut over.”
Not saved by this site. Your browser may restore text when you navigate back.
A strong answer, step by step
Open each step in turn. This is one strong answer, not the only one.
Step 1: Count unique target IDs
What would you do here? Open the move you’d make.
Separate missing IDs from duplicate extras.
The strong move. Missing rows and duplicate extras offset in the total row count.
Use the 200-row net gap as total defects.
Tempting, but it misses something. Missing IDs and duplicate extras offset in the row total: the net gap hides 500 missing IDs and 300 duplicate extras.
See how a strong candidate says it
“The target has 999,800 − 300 = 999,500 unique IDs. Against 1,000,000 source IDs, 500 are missing. The net 200-row gap hides 800 defects: 500 missing plus 300 duplicate extras.”
Why it works: Missing rows and duplicate extras offset in the total row count.
Step 2: Apply both technical gates
What would you do here? Open the move you’d make.
Hold and assign reconciliation work to producer owners.
The strong move. A low percentage does not waive an explicit correctness requirement.
Accept the gap because it is below one percent.
Tempting, but it misses something. Any missing ID fails the zero-missing rule, and 300 duplicate extras exceed the cap of 50.
See how a strong candidate says it
“Five hundred missing IDs fail the zero-missing rule. Three hundred duplicate extras also exceed the 50-per-million cap. I would hold cutover and send each producer its reconciliation evidence.”
Why it works: A low percentage does not waive an explicit correctness requirement.
Step 3: Check what still remains
After repair, no source IDs are missing and 25 duplicate extra rows remain. Nine of ten producer owners accept reconciliation. The rollback test has not run.
What would you do here? Open the move you’d make.
Wait for the tenth owner and tested rollback.
The strong move. Passing the data counts does not establish operational acceptance.
Cut over as soon as the row metrics pass.
Tempting, but it misses something. One producer has not accepted reconciliation, and rollback has not been tested.
See how a strong candidate says it
“The new snapshot has zero missing IDs and 25 duplicate extras per million, within the numeric cap. Only nine of ten producers have accepted reconciliation, and rollback is still untested. I would keep cutover on hold.”
Why it works: Passing the data counts does not establish operational acceptance.
Step 4: Give a conditional recommendation
What would you do here? Open the move you’d make.
Make approval conditional on the remaining gates.
The strong move. A technical programme decision needs evidence, ownership and a recovery path.
Replace the missing acceptance with the AI summary.
Tempting, but it misses something. An AI summary cannot supply the missing producer acceptance or a passing rollback test.
See how a strong candidate says it
“I would approve a cutover window only after the last producer accepts, rollback passes and the reconciled snapshot is frozen. I would name the go/no-go owner and the condition that triggers reversal.”
Why it works: A technical programme decision needs evidence, ownership and a recovery path.
The answer, in short
- Decision
- Keep cutover on hold after the data repair.
- Because
- One producer’s acceptance and the rollback test remain open.
- What would change it
- All producer sign-offs and a passing rollback test could permit a scheduled cutover.
Common traps
- Confusing net difference with gross defects.
- Treating count accuracy as full readiness.
Check your answer
- I removed duplicate extras before counting unique IDs.
- I found 500 missing IDs.
- I applied both numeric gates.
- I preserved acceptance and rollback gates.
Last reviewed Nov 10, 2026