Moving data does not by itself justify switching operations. Define reconciliation, business acceptance and rollback criteria as explicit project deliverables.
Start by separating data reconciliation from business acceptance. Agree the workflow, comparison window, discrepancy policy and approver. Matching totals can conceal offsetting duplicate and missing records or changed date interpretation.
What the official documentation establishes
Google Cloud documents using the Data Validation Tool to compare source data with BigQuery after transfer, including selected count, sum and average aggregations.
Google Cloud: Schema and data transfer overview
The following are For f design and review recommendations. They are scoping considerations, not customer results or a guarantee of outcomes.
Compare the same time window and definitions
Snapshots taken at different times can produce differences unrelated to migration defects. Align cutoffs, ingestion delays, deleted records and NULL versus empty-string handling. For sales, confirm order versus posting dates and the treatment of returns and cancellations.
Classify discrepancies and record acceptance decisions
Separate count, amount, duplicate-key, missing-record and transformation differences. Do not close acceptance with unexplained discrepancies. Record the scope, rationale and approver for accepted exceptions. Tool output informs a decision; it does not define company-specific business rules.
Include downstream reports and rollback
Test reports, BI, scheduled distribution and downstream feeds against migrated data. Include restricted users and missed refreshes. Define postponement criteria, restoration steps and treatment of records created after cutover, as well as the cutover time.
Example: accepting a monthly sales migration
For a hypothetical monthly sales migration, test order-level lines, returns, post-close corrections and payments across month boundaries separately. Keep tests that reproduce the old report separate from tests of an agreed new calculation. Assign an owner, cause, correction plan and retest result to each discrepancy. Be prepared to defer switching a report with unexplained differences. This is a test-design example, not a customer case study.
When comparing proposals, check report counts, comparison periods, exception types, parallel operation and the scope of discrepancy remediation, rather than data volume alone. Specify an explanation of acceptance results and an operational handover as deliverables, making the boundary between transferring data and making it usable explicit.
Deliverables to agree before engagement
| Area | What to review | Evidence to retain |
|---|---|---|
| Data integrity | Business keys, aggregation grain and time window | Reconciliation results, open differences and owners |
| Operational acceptance | Representative reports, permissions and refresh timing | Business-user acceptance results |
| Cutover decision | Postponement, rollback and parallel running | Execution plan and approval record |
Prepare for an initial conversation
- List the workflows, reports and BI assets to migrate
- Identify source data for comparison and business reviewers
- Include reconciliation and rollback in the estimate
How many sample records are enough?
There is no universal sample count. Combine aggregate checks with important transactions, exceptions, time boundaries, NULLs, updates and deletions. Decide which checks require all records and which can use samples based on impact and verification effort.
Discuss the scope with For f
Explore the related service / Review the approach and scope
References and verification date
Official page last updated: 2026-09-18. Information checked: September 21, 2026. Product features and conditions can change; check the official documentation and your environment before implementation.
The thumbnail is an AI-generated concept image, not an actual system screen or measured result.