Website re-audit: compare before and after.
A client-review checklist for agencies. Published 19 November 2026. Implementation reviewed 19 November 2026.
A re-audit is a new audit after changes, not proof that revenue improved. Compare the same conditions, preserve evidence and separate website findings from commercial results.
Run a free audit to establish your baseline. The editable setup opens without payment or sign-in.
1. Save a baseline before changing the site
Start with the pre-pitch website checklist. Choose the findings the client agreed to fix, then preserve the original report and supporting evidence.
- Page and time. Record the exact audited URL, final address after redirects, audit date and report link. Note the site version or release reference.
- Conditions. Record report language, browser, viewport dimensions and consent state where known. Separate desktop from mobile evidence.
- Coverage. Copy the overall score, pillar scores, check counts and any partial or not-checked reasons. Keep the original limitations visible.
- Assumptions. Write down visits, booking rate and revenue per booking, with their sources and periods. Label estimates and unknowns.
- Agreed work. Give each selected finding a short identifier, its observed evidence, proposed fix, responsible person and verification method.
A missing original capture cannot be recreated as historical proof. Label the gap. Do not take a new screenshot and call it the before view.
2. Repeat the conditions, not just the URL
After the release is live, run a new audit of the same page. Check that redirects still lead to the intended version.
For manual browser checks, match the original viewport, zoom, browser and consent state. Note login status, locale and any experiment variant. Use a fresh session when testing first-visit behaviour, without treating it as equivalent to a returning visitor.
Compare natural consent captures with natural captures, and provider-cleared captures with cleared captures. A provider hiding a banner does not prove refusal, withdrawal or tracking behaviour.
PitchGap’s rendered analysis can vary between runs. Capture availability, consent state and model interpretation can change. Inspect the attached evidence and tested view before attributing a changed finding to the release.
If conditions cannot be matched, record the difference beside the finding. Retest that issue manually under matching conditions rather than presenting a clean before-and-after claim.
3. Compare each finding against evidence
Use a short evidence record per agreed fix, not a list of score differences. Match findings by problem and page, not their report position.
- Before. Quote the original observation and link its dated report or capture. State whether it was observed, inferred or a signal to verify.
- Change. Record what shipped and when. A changed title tag and a redesigned enquiry path need different verification.
- After. Link the new evidence. Repeat the relevant manual check and name the tested viewport or page element.
- Status. Use resolved in the tested scope, still observed, not comparable, or not checked. Give the reason and next action.
A finding disappearing from the new report is not enough to mark it resolved. The check may not have run, or the evidence may differ.
For a contact-route fix, verify navigation to the intended form or booking page. Test delivery only with the client’s permission and an agreed test process.
For EU Compliance, record the observable change separately from legal review. A newly detected privacy link does not establish that the notice is adequate.
4. Read coverage before reading the score
PitchGap scores the tested checks, not the whole website. Each pillar starts at 100 and subtracts finding weights: critical 12, high 8, medium 5, with a floor of zero.
The overall score is the rounded average weighted by each checked pillar’s check count. Older reports without check counts use a plain average. Do not treat those methods as interchangeable.
A failed visual analysis can leave Website Experience partially checked from HTML. Truncated page content also reduces coverage. A not-checked pillar has no score; it is not a pass.
Compare partial reasons and individual captures as well as the reported count. The count alone does not guarantee both viewports were assessed or every obscured question was conclusive.
Local Visibility and Social Presence currently use narrow homepage checks. Competitor Gap is not checked and is excluded from the overall score. A 100 means the tested checks found no scored problem, not complete visibility or legal assurance.
If coverage differs, compare only the findings with equivalent evidence and checks. Report the overall scores separately with their limitations, rather than claiming the score difference measures the release’s effect.
5. Keep the euro assumptions unchanged
For its enquiry-to-booking scenario, PitchGap uses visits × an assumed 0.5–1.0 percentage-point uplift × booking rate × revenue per booking.
The scenario is independent of the score and finding count. Fixing a finding does not automatically lower it. Identical inputs produce the same estimate before and after.
Keep visits, booking rate and booking revenue unchanged for an assumption-controlled comparison. Record them outside the report too: edits in the report panel recalculate locally and do not save back or carry into the next audit.
If new client records replace an assumption, show a separate revised scenario and explain the changed input. Do not label its difference recovered revenue.
Without a booking rate, the report stops at estimated extra enquiries. Detected shops, SaaS and directories need their own model-specific numbers; uncertain models receive no numeric scenario.
The lost-enquiry estimate guide explains the arithmetic, rounding and limits. A scenario is neither a measured loss nor a forecast.
6. Measure commercial outcomes separately
Agree the commercial measurement before shipping. Define relevant visits, qualified enquiries, paid bookings and realised revenue consistently, using records the client is authorised to share.
Compare suitable periods with the same definitions and tracking setup. Show counts alongside rates, and note traffic mix, advertising, seasonality, pricing, capacity and other concurrent changes.
Small samples and unfinished bookings limit conclusions. If tracking changed, explain the break rather than joining incompatible periods into one trend.
A higher score supports a claim about tested website checks. It does not establish more enquiries, more revenue or causation. Even an observed conversion increase can have other explanations.
Where commercial data is unavailable, say so. “The tested contact route now works” is useful evidence without an invented return on investment.
7. A hypothetical review note, not a customer result
This example is fictional. It illustrates a manual review record, not a PitchGap progress feature or a real client outcome.
- Before. The selected mobile capture showed the consent panel covering the enquiry action. The agency recorded its viewport and consent state.
- After. A matching manual test shows the action visible with the panel present. The agency marks this issue resolved in that tested view.
- Limitation. The new audit’s visual analysis failed. Its overall score is not used as comparable progress evidence. Desktop behaviour still needs verification.
- Money. Scenario inputs stay unchanged. No enquiry or booking records are available, so the note claims no measured commercial improvement.
- Next action. Verify desktop behaviour and consent functionality, then review the agreed commercial records when the measurement period ends.
End each client review with what is evidenced, what remains unknown and one ranked next step. Set the next review in your own calendar.
8. Use PitchGap for the next audit, compare manually
PitchGap creates a separate report for each new audit. Its signed-in workspace lists past completed reports with their URLs, dates, languages and scores.
It does not currently group client history, calculate before-and-after progress, track resolved findings or send monthly re-audit reminders. Keep your comparison record and review schedule yourself.
See the labelled sample report for the format and audit plans for the paid offer. A new audit uses the applicable free or paid allowance; the guide does not add extra audits.
PitchGap’s site and business are built and run by AI agents on NanoCorp. Begin with a free audit, then preserve the evidence before proposing changes.