Before opening website analytics, write three questions: Are people reaching the page that matters? Are they taking its intended action? Are they completing the action you actually want? Give each question a possible decision. This turns a dashboard visit into a focused review rather than a search for numbers that look interesting.
You do not need to measure every part of your website at once. Start with one important page and one useful outcome. The trade-off is a narrower view, but that makes it easier to check what the numbers mean and choose a change you can realistically make.
Choose a page with a specific job
Example: A freelance illustrator has a page describing editorial illustration services. Its job is to help suitable prospects understand the offer and send a project enquiry. The immediate decision is whether to improve how people find that page, explain its content or investigate the enquiry process. This is a hypothetical planning example, not a report of business results.
Write the page’s job in one sentence before selecting measures. “Help visitors understand my services” is a useful starting point, but “help suitable prospects submit a project brief” gives the review a more focused destination. Analytics may show recorded actions; it cannot, by itself, establish whether every visitor was suitable or understood the offer.
Question 1: Are people reaching the page that matters?
Look for visits to the chosen page and, where available, how those visitors arrived. Keep the question specific: “Are visits from my newsletter reaching the services page?” is more actionable than “Where does my traffic come from?” It points towards a decision about the newsletter link, its destination or the promise made before the click.
Check that the link leads to the intended page and that the reporting period includes the promotion you are reviewing. Source labels may be incomplete or inconsistent, so treat an unexplained category as a measurement question, not evidence that a particular channel is failing. More visits are not automatically better if the page attracts people seeking something else.
Question 2: Are visitors taking the intended action?
Choose one observable action that moves someone towards the page’s goal. In the illustrator example, that might be opening the project enquiry form. Avoid collecting several convenient measures merely because they are available. Scrolling, time on page and button clicks describe different behaviours; none automatically demonstrates interest or satisfaction.
Before interpreting the action count, verify what triggers it. Does the event record a click on the enquiry button, the form becoming visible or something else? Check whether repeated clicks can produce repeated records. If the definition is ambiguous, your next task is to explain measurement—not rewrite the page based on a number you cannot explain.
Question 3: Are people completing the useful action?
Separate starting from finishing. Opening an enquiry form is not the same as successfully submitting it. Choose a completion signal that matches the website’s job, then verify that it represents success rather than an attempted submission. Where useful, compare recorded completions with the enquiries actually received, without copying personal details into your review notes.
If starts appear more common than completions, several explanations remain possible: visitors may be exploring, the form may be demanding, a technical problem may exist or tracking may be incomplete. Analytics identifies a place to investigate; it does not prove which explanation is correct. Manually checking the process is a sensible action before changing the offer.
Run one small review, then choose one action
Use a short note with three rows, one per question. Record the measure, the period, any known gaps and the decision it could inform. Use comparable periods, noting promotions or site changes that make comparison less fair. With sparse activity, extend the observation period rather than treating one visitor’s behaviour as a dependable pattern.
- Confirm the chosen page and its intended outcome.
- Check what each recorded action actually counts.
- Test the enquiry path, including successful completion.
- Record missing data or unusual activity alongside observations.
Finish with one decision: improve the route to the page, explain the action, investigate completion or collect more evidence. Change one meaningful element where possible and record when you changed it. A later difference is a reason to investigate further, not proof that your edit caused it.