Measure whether a website helps its visitors

Give each indicator a question

For each key service, describe the expected journey: discover the offer, read its conditions, review an example and make contact. Choose a small set of measures that can reveal a break, such as visits to the service page, contact page arrivals and enquiries actually received.

Write a possible decision beside every indicator. If no action would follow a change, collecting it may not be a priority. Keep visitor volume separate from enquiry quality and feedback from customers.

• Question to answer.
• Minimum measure and comparison period.
• Possible decision if the result changes.

Validate collection before interpreting results

Test events on desktop and mobile. A click on a button does not prove the form arrived. Look for duplicate events, loading errors and internal visits. Record tagging changes so that two periods with different definitions are not treated as equivalent.

Analytics can involve trackers and personal data. Review the configuration, retention and recipients. The French data protection authority explains that exemption from consent applies only to certain measurements under conditions; do not assume a particular tool qualifies.

Compare like with like

Note campaigns, publications, holidays, outages and offer changes that may explain movement. Compare the same page family and, for low volumes, look at several weeks instead of one day. Laboratory performance scores, audience figures and sales enquiries answer different questions.

Read searches with no results, questions sent through contact and feedback from people who use the site. These often reveal missing information that a chart cannot show.

Make one change and check its effect

Choose a specific improvement: clarify a price, make a condition visible, repair a link or simplify a form. Record the hypothesis, affected page, change date and expected observation. Check the rendered page and actual delivery of enquiries before drawing a conclusion.

If the result does not change, revisit the observed journey before adding more dashboards. Incomplete measurement should be described honestly; it cannot support a promise about conversion or rankings. The maintenance guide suggests a review rhythm after each change.

Fictional example: 200 visits produce 10 enquiries, or 5%; another period has 500 visits and 15 enquiries, or 3%. The rate falls while enquiry volume rises by 50%. Before choosing a change, compare relevance, traffic sources and the definition of a visit. A rate alone cannot establish whether the website has become more useful.

To connect SEO and SXO, keep the search question, landing page, intended action and enquiry actually received together. For GEO, separately record the assistant’s answer, cited source and test date. These observations have different denominators and purposes; adding them into one visibility score would hide those differences.

Observe performance without assuming every browser exposes identical measurements.

Synthetic case: a dashboard expects an entry type missing in a browser. Check supportedEntryTypes and distinguish unavailable data from zero. Define collection start and end. Use technical labels without email addresses or visitor input; this initial local check needs no remote data collection.

• Available measurements are identified, missing data stays unknown and labels contain no personal information.

• MDN PerformanceObserver

  Interface receiving new browser performance timeline entries for supported entry types.

Decide in a concrete situation

Fictional example: visits increase while usable enquiries remain stable. Examine journeys, periods and event definitions before concluding; exposure does not automatically measure visitor usefulness.

Document event, unit, scope, missing data and collection method. Compare consistent periods and note consent or tool changes; a counting difference need not indicate a different audience.

• Manage website enquiries after submission

  From form confirmation to closure: route, qualify and follow up enquiries without losing context or exposing personal data.

From 4% to 5%: what changed?

Fictional example: 20 enquiries from 500 visits is 4%; 25 from 500 is 5%. The difference is 1 percentage point, or a 25% relative increase. It does not establish that a website change caused it. Check comparable periods, definitions and enquiry quality before interpreting the change.

Find a missing page before concluding there were no visits

1. 1Identify the report
2. 2Keep period and segment
3. 3Examine Others
4. 4Compare archiving

In a Matomo report, an “Others” row may group entries during archiving. Separate three stages: collected data, aggregation and the displayed list. Showing more rows does not separate an entry that has already been aggregated.

Before changing limits, record the version, hosting mode and relevant setting. Compare in a test copy; reprocessing older periods can require time, memory and storage. Do not turn an invisible row into an absence of visitors.