Practical framework / 4 MIN READ
Accept a new data feed by proving its meaning
Compare coverage, definitions and operating responsibilities before replacing a reporting source.
AI-assisted research and analysis for the SNS Creative Fieldbook
Sources checked 2026-11-06
A new provider can deliver the expected file format while changing what the records represent. A trustworthy handoff tests semantic compatibility and recovery procedures, not merely whether the connector turns green.
What you will leave with: Build a feed acceptance plan with a parallel comparison, discrepancy register and explicit cutover authority.
Use this when
- A reporting supplier or connector is being replaced.
- A new data feed has familiar field names but unfamiliar totals.
This lesson concerns data-service transition, not research participant recruitment or supplier payment approval.
Define the accepted service
List the decisions the feed supports, required fields, grain, coverage and delivery schedule. Separate critical requirements from convenient extras. Identify who owns source collection, transformation, access and downstream reporting so a discrepancy has an accountable route.
The UK Government’s data-quality guidance emphasizes maintained metadata and clear quality limitations for users. Apply that principle by asking for definitions, known omissions, revision behavior and usage restrictions with the feed. A sample file without this context is insufficient evidence of fitness for the intended decision.
Compare meanings before totals
Map old and new fields by meaning, not just name. Check currencies, timezones, status definitions, exclusions and whether amounts are gross, net or modeled. A field called customer may represent an account in one source and an individual in another.
Use a controlled overlap period where permitted. Reconcile a small set of identifiable non-sensitive cases and aggregate totals. Classify differences as expected definition changes, timing differences, coverage gaps or unexplained errors. Do not force agreement through undocumented adjustments merely to meet a transition deadline.
Rehearse ordinary failures
Test missing files, duplicate deliveries, schema changes and late corrections using approved fixtures. Confirm how the provider announces incidents and whether recovery resends full periods or only changes. Verify that downstream processing handles either mode without duplicate records.
Check least-privilege access and ownership transfer without sharing credentials in a handoff document. New persistent access, contracts or data transmission require the relevant approvals. Document an exit and deletion process for the old provider under existing agreements; technical readiness does not itself authorize ending a contract or removing data.
Make the cutover a visible decision
Agree acceptance criteria and who can approve the switch. Preserve a rollback or last-reconciled reporting option where feasible. If definitions intentionally change, annotate the break in the historical series and provide a bridge instead of claiming continuous comparability.
After the switch, monitor the critical checks through a defined stabilization period. Close unresolved issues with evidence rather than optimism. Keep the previous source’s permissible documentation and important report versions available for later questions, while following agreed retention and access restrictions.
Make the reasoning concrete.
A hypothetical replacement feed reports 1,100 enquiries where the old feed reports 1,000. Investigation finds 80 duplicates in the new delivery and 20 valid records previously omitted. The correct comparable total is 1,020, not either raw total.
Accept the larger total
Opening: Treat 10% more records as improved coverage.
Hypothesis: The new feed appears richer.
Risk: Most of the difference is duplication.
Resolve the bridge
Opening: Remove verified duplicate delivery and document the 20-record coverage correction.
Hypothesis: The team understands the real change.
Risk: Other periods still need appropriate acceptance checks.
Accept only after the duplicate mechanism is repaired and the coverage change is documented. Do not claim a marketing performance increase from a provider transition.
Take the questions into your next brief.
Use these prompts in your own notes. Nothing is submitted from this page.
Requirements
Which fields and decisions are critical?
Semantics
Which definitions differ between providers?
Bridge
What explains each material discrepancy?
Failure tests
How are retries, late files and corrections handled?
Cutover
Who approves the switch and fallback?
Watch for the wrong conclusion.
- Treating matching column names as matching meaning.
- Calling extra rows better coverage without checks.
- Using undocumented plugs to force agreement.
- Confusing technical acceptance with contract or access authorization.
Your next action
Create a provider acceptance checklist and reconcile one overlapping period before proposing a reporting cutover.
Sources & scope
Frameworks, worksheets and hypothetical examples are original editorial synthesis. Linked primary sources support only the stated concepts. Source checking is not an independent audit of effectiveness. No named human author or reviewer is claimed.
- UK Government: Data Quality Framework guidance →Primary government data-quality guidance · Supports: Recommends maintained metadata and communicating data quality, scope and limitations to users. Supplier acceptance, parallel reconciliation and cutover planning here are original editorial synthesis.