Project logic
Problem, system, output.
The framework treats redirect QA as evidence reconciliation. It does not assume that a redirect is correct because it resolves, or that two similar-looking destinations answer the same user intent.
Migration risk hides inside familiar-looking spreadsheets.
Redirect maps, crawl exports, chains, status codes, and final destinations need to be reconciled before anyone can understand which rows are safe and which need attention.
Planned mappings are compared with crawl evidence.
The local workflow validates the files, normalises URL data, joins planned redirects with Screaming Frog evidence, and classifies each row with conservative reason codes.
Reports separate technical evidence from stakeholder readouts.
Internal workbooks keep the detail needed for QA, while simplified stakeholder-facing reports focus attention on status, risk, and rows that need review.
Pipeline
From source inputs to reviewed output.
- Input
Mapping workbook
Planned redirects provide the baseline for comparison.
- Input
Crawl evidence
Observed status and final destination evidence are joined to the map.
- Logic
Normalisation
Formatting noise is reduced before the decision layer runs.
- Rules
Classification rules
Exact, accepted, ambiguous, and high-risk rows are separated carefully.
- Review
Review queue
Broad hubs, mismatches, broken targets, and missing evidence stay visible.
- Output
Stakeholder workbook
Technical evidence is converted into review-ready reporting.
Artifact proof
Two public-safe specimens.
Synthetic examples only. No private source material is shown.
Joined evidence
Synthetic redirect row joined to crawl evidence
The framework does not treat a plan as proof. It compares the map with crawl-backed evidence before classification.
| Layer | Synthetic value | Decision signal |
|---|---|---|
| Mapping source | Source URL A to Destination URL B | Expected destination |
| Crawl evidence | Observed final destination B | Evidence available |
| Final status | Resolved target | No broken final target |
| Classification | Review needed | Broad destination requires judgement |
Path specimen
Redirect path diagram
A tidy redirect chain is still checked against intent. Similar-looking destinations are not automatically equivalent.
- Old URLSource URL A
Synthetic source entry
- Final destinationDestination URL B
Observed target
- Final statusResolved
Evidence confirms a reachable final target
- ClassificationReview needed
Similar is not automatically equivalent
Mini representation
Classify this redirect
Choose a synthetic scenario to see how conservative classification turns evidence into a review status.
- Classification
- Review needed
- Reason code
- RQ-BROAD
- Human action
- Review intent fit before release
Simplified portfolio representation. Synthetic values only; no private crawl or migration data.
Governance / QA
Similar paths are not always equivalent destinations
A redirect can look tidy while still sending users and search engines to a weaker destination. The framework is intentionally conservative because migration QA needs evidence, not wishful matching.
- A broad hub is not automatically a valid replacement for a specific old URL.
- Heuristic semantic similarity is treated as advisory rather than final proof.
- Human review remains part of the release process for ambiguous or high-risk rows.
What this demonstrates
- Technical SEO migration risk management.
- Screaming Frog export interpretation.
- Local reporting automation.
- Conservative classification logic.
- Stakeholder-friendly QA outputs.
Next iterations
- Additional rule explainers for edge cases.
- Synthetic demo data for public documentation.
- More flexible input presets.
- Optional summary views for non-technical reviewers.