In short
Measure a redesign against the problems it was commissioned to solve. Record the pre-launch baseline; track priority user tasks, qualified business actions, search performance, accessibility, Core Web Vitals, editorial work, and operational errors; segment by page and audience; and interpret changes alongside campaigns, seasonality, pricing, and product changes.
Begin before launch
Measurement starts while the old site still exists. Record the baseline for the same pages, journeys, devices, countries, and time periods you will compare later.
Write the redesign logic:
> We changed X because evidence showed Y, expecting it to improve Z, while watching guardrail G.
Example:
> We reorganised services around buyer tasks because prospects entered through articles but could not identify the relevant capability. We expect more qualified movement from articles to service evidence without reducing informational usefulness.
Use five evidence groups
1. User-task evidence
- completion of priority tasks in usability testing;
- path success and abandonment;
- form error and completion;
- search-result selection;
- comprehension of offer and next step;
- support questions caused by unclear information.
2. Business evidence
- qualified enquiries, applications, bookings, or transactions;
- lead quality and sales progression;
- self-service completion;
- revenue or pipeline influenced, with attribution limits;
- cost or time removed from internal work.
3. Search and discovery
- clicks and impressions by query cluster and page type;
- indexed pages and errors;
- movement from informational to commercial pages;
- branded and non-branded visibility;
- referral traffic from search and answer engines.
4. Experience quality
- Core Web Vitals field data;
- accessibility defects and manual test results;
- mobile and browser errors;
- broken links and failed integrations;
- layout and JavaScript regressions.
5. Editorial and operational evidence
- time to publish common content;
- developer requests for routine changes;
- duplicated or inconsistent records;
- approval and translation delays;
- incidents, recovery, and maintenance load.
Choose one primary measure per outcome
Avoid a dashboard of twenty equal numbers. For each objective, choose:
- one primary outcome;
- supporting diagnostics;
- one or two guardrails;
- responsible owner;
- review window.
If qualified form completion is primary, diagnostics may include landing-page relevance, field errors, and mobile completion. A guardrail may ensure lead quality does not fall.
Segment before concluding
Site-wide averages conceal problems. Compare:
- new versus returning visitors;
- mobile versus desktop;
- country and language;
- page type;
- acquisition channel;
- branded versus non-branded demand;
- priority audience where identifiable without intrusive tracking.
A redesign may improve service journeys while a migration defect harms articles. One average cannot represent both.
Account for confounding changes
Record campaigns, media spend, seasonality, product releases, price changes, sales staffing, outages, consent changes, analytics changes, and search updates.
Do not claim the redesign caused every movement after launch. Combine quantitative trends with tests, interviews, and operational evidence that connect the changed experience to the observed behaviour.
Review on three horizons
First days
Errors, forms, analytics, redirects, indexability, performance, and operational failures.
First month
Journey behaviour, content gaps, conversion diagnostics, editor workflow, and early search processing.
Following quarters
Qualified outcomes, search trends, repeat use, content productivity, maintenance load, and strategic fit.
Search migrations and field performance need time; critical functional defects do not.
Common questions
What is the best metric for a website redesign?
The best primary metric is the outcome closest to the problem the redesign addressed, such as qualified enquiry completion, task success, or publishing time. No single metric fits every site.
When should results be evaluated?
Check technical and functional health immediately, early journey evidence within weeks, and longer-term search or business outcomes over comparable periods.
Is conversion rate enough?
No. It can change with traffic mix, demand, pricing, and measurement. Pair it with lead quality, task evidence, page-level diagnostics, and qualitative research.
How do we measure a redesign with low traffic?
Use moderated usability tests, interviews, form and error evidence, sales and support feedback, performance, accessibility, and editorial workflow alongside the limited analytics.
What if traffic falls but qualified leads improve?
That may be a positive result if irrelevant demand was removed and business value increased. Check lost queries and pages before accepting the trade-off as intentional.
