Website Redesign Checklist — What to Measure Before, During, and After

Short answer: before a redesign, record four weeks of baselines by device for each funnel step, save heatmaps and a set of recordings of the current site, and list what is working so you do not remove it. Launch in stages where you can. Afterwards, check that tracking still works, compare the same steps over matching periods, and wait for the novelty period to pass before declaring a result.

Table of Contents

1. Why Redesigns Go Wrong
2. Before: Baselines and Evidence
3. During: Design Decisions
4. Launch
5. After: Judging the Result
6. Key Takeaways

Why Redesigns Go Wrong

• Everything changes at once, so no change can be linked to any effect.
• Things that worked are removed because nobody knew they worked.
• The brief is aesthetic, not a list of problems to solve.
• Tracking breaks in the move, and the "drop" is partly or wholly a measurement fault.
• URLs change without redirects, and search traffic falls.
• Returning visitors are disoriented.
• No baseline exists, so success is a matter of opinion.

Before: Baselines and Evidence

Capture numbers. At least four weeks, by device and by source.

Save qualitative evidence. Retention is limited: 30 days on DeepSync's Free plan, 90 on Pro, 180 on Business. Old-site recordings and heatmaps will age out, so export or screenshot what you need.

• Click and scroll heatmaps for each key page, by device.
• Twenty recordings of successful conversions.
• Twenty of abandonment at each major step.

List what works. From heatmaps and recordings:

• Which elements get the clicks that lead to conversion?
• Which content do converting visitors read?
• Which navigation items are used (navigation)?
• What do visitors search for (site search)?

List what is broken. Rage clicks, dead clicks, steps with heavy drop-off, pages nobody reaches (audit checklist). This list is the real brief.

During: Design Decisions

• Tie each change to a problem. If a change does not address something on the broken list or protect something on the working list, ask why it is being made.
• Keep what is used. If a plain text link in the footer drives a third of pricing page visits, it stays, whatever the new footer looks like.
• Do not remove content that ranks. Check organic landing pages before cutting or merging.
• Design mobile first. Most visitors are there.
• Set a performance budget. The new site should be no slower than the old at the 75th percentile.
• Test prototypes with people before building (unmoderated testing).
• Plan the tracking. List every event, goal, and funnel step, and how each will be reproduced. Elements identified by CSS selectors or page URLs will change.

Launch

URL and SEO.

• Map every old URL to a new one, with a 301 redirect, in a single hop.
• Keep canonical URLs, the sitemap, and internal links consistent with the final URL form (a trailing-slash example).
• Carry over titles, descriptions, headings, and structured data.
• Check robots.txt and that a staging "noindex" did not ship to production.

Tracking.

• Analytics script present on every template.
• Consent banner working, and recording starting only after consent.
• Each event firing once. Place a test conversion and reconcile with your backend.
• Masking still applied: new templates have new class names, so re-check ds-mask and ds-block on sensitive elements (masking).
• Funnels and goals updated for new URLs and events.

Staging the release.

A full switch on one day is sometimes unavoidable. In that case, the checks below matter more.

The first 48 hours (post-release checks):

1. New JavaScript error groups, sorted by sessions affected.
2. The funnel by device, step by step.
3. Twenty recordings on mobile, of the main flow.
4. Rage clicks on the new templates.
5. A test purchase or signup on a real phone.
6. Server logs for 404s from old URLs.

After: Judging the Result

Do not judge in the first week. Returning visitors behave differently when a site changes. Some explore the new design, some are lost. Both effects fade (novelty effect). Give it three to four weeks.

Compare like with like.

• Same weekdays, equal periods (comparing periods).
• By device and source, not blended.
• New visitors separately from returning. New visitors have no memory of the old site, so they give the cleaner reading.
• Funnel step by step.

Check tracking before believing any change. A large move in either direction on launch day is a measurement fault until shown otherwise (why numbers differ).

Remember what a before-and-after cannot prove. Without a concurrent control, season, campaigns, and traffic mix are mixed in. A staged percentage rollout is the only clean comparison (test pitfalls).

Watch organic traffic for several weeks. Search engines take time to process redirects, and a temporary dip is common. A continued fall means lost pages or broken redirects.

Then iterate. A redesign is a starting point. Use the new heatmaps and recordings to find what the new design got wrong, and fix those one at a time.

Key Takeaways

• Capture baselines and save evidence before changing anything.
• Know what works so you do not remove it.
• Verify tracking, redirects, and masking at launch.
• Judge after the novelty period, by segment and by step.

Conclusion

A redesign is the riskiest change a website makes. The teams that come through it well are the ones who knew, in numbers, what they had before they started.

Start free with DeepSync.