How to measure the impact of a new website

Compare traffic, CTR, engagement and real enquiries before and after launch. With definitions, time periods and worked examples.

Editorial contact: Kjetil Fagereng
About 6 min readPublished Updated
Read the article
A person examines charts on a laptop at a desk.
The reports shown are illustrative. Photo: PRIME STOCK LAB / Adobe Stock.

Judge a new website on more than visitor counts. Compare traffic, usage and genuine enquiries before and after launch using matching periods and consistent definitions.

The most interesting question is often whether more of the right people find you and make contact. More page views can be positive, but may also result from more advertising, new articles or users having to visit more pages to find an answer.

Start with a clear period and a known launch date

For a seasonal website, two complete twelve-month periods are a useful starting point. Also compare each month with the same month a year earlier. This makes it easier to see whether a change is consistent, seasonal or linked to a particular moment.

Example: PEC launched its new website on 31 January 2025. February 2024–January 2025 and February 2025–January 2026 are therefore practical periods to examine. The final day of the before-period includes the transition itself, which should be noted. This is a comparison framework, not a results report for PEC. Read the project story.

If there have been major changes to domains, subsites or analytics configuration, first investigate which data actually covers the same business. Missing measurement must not be interpreted as zero activity.

Choose a small set of complementary metrics

  • Users and sessions: How much relevant traffic reaches the website? Consider mobile and desktop separately.
  • Search impressions, clicks and CTR: How often does the website appear in Google Search, and what share of impressions produces a click? Assess ad CTR separately.
  • Engagement time: Do people spend time with the content actively on screen? State whether the report shows an average per active user or per session.
  • Genuine enquiries: How many new, relevant requests arrived? Distinguish forms, confirmed phone calls and other contact methods.
  • Quality: How many enquiries match the business’s services and area, and how many are followed up?

GA4 engagement time measures time with the website in focus or the app in the foreground. It is not identical to the old ‘average time on page’ measure. See Google’s definition of engagement. A short contact page can do its job well even with a low time figure.

Calculate enquiry rates with the right denominator

Two useful metrics are:

  • Enquiries per 1,000 page views: Genuine enquiries divided by page views, multiplied by 1,000.
  • Share of sessions with an enquiry: Sessions with at least one verified enquiry divided by all measured sessions, multiplied by 100.

The first is a ratio. The second is a session-based conversion rate and requires an enquiry to be linked to a session. Dividing CRM enquiries by page views does not calculate the share of visitors who convert.

Calculation using fictional figures: 120 genuine enquiries and 40,000 page views give 3 enquiries per 1,000 views. If the next period has 180 enquiries and 50,000 views, the rate is 3.6. The number of enquiries has risen by 50%, while the rate has risen by 20%. Both statements can be correct, but they describe different things.

For percentage rates, distinguish percentage points from percentages. A change from 2 to 3 per cent is an increase of one percentage point, or 50 per cent in relative terms.

Worked example: 120 enquiries from 40,000 page views is 3 per thousand. 180 from 50,000 is 3.6 per thousand.
The article’s worked example. Figures are fictional and show why counts and rates must be assessed separately.

Assess cost per conversion alongside quality

Cost per conversion is cost divided by the number of recorded conversions. Use the same channel, period, conversion actions and counting method in the comparison. If phone clicks were included before, do not remove them from the later period alone.

When the conversion actions represent enquiries, this can provide a consistent measure of cost per enquiry. State what you count and also assess the quality of those enquiries. See Google’s definition of average cost per conversion.

Distinguish bounce rate from form abandonment

GA4 bounce rate is the proportion of sessions that are not considered engaged. It is not the same as the proportion of people who start a form but do not submit it. A form needs its own measurement points for starting and successfully submitting it.

Assess bounce rate and engagement time alongside the task the page is meant to support. A lower bounce rate alone does not demonstrate more valuable enquiries. Check that thresholds and key events are comparable. See Google’s explanation of engagement rate and bounce rate.

Clean enquiry data before counting

Exclude spam, test submissions, job applications and support from sales reports. Check whether the same request was recorded more than once. You can retain raw and cleaned counts internally with a simple explanation of the filtering.

A form event in analytics is not necessarily a qualified enquiry. Google recommends a separate event for the relevant lead form rather than marking every form submission as a key event. See the guide to measuring form enquiries. Also check that the submission actually arrived.

Check what might explain the difference

Note changes to ad budgets, campaigns, prices, service areas and available capacity. Look for data collection breaks when changing consent systems, analytics or forms. Do not combine overlapping analytics properties without checking for double counting.

Organic CTR is also affected by the queries and positions where the website appears. Use Search Console’s performance report to investigate relevant pages and search groups, not just the total. Historical exports may be needed for older periods.

Make the report useful for the next decision

Fictional example: A service business receives more visits and more recorded conversions after a launch. However, advertising has also increased in an area the business does not serve. The sales manager finds that more enquiries have to be declined. The overall report looks better, but it does not establish that the business is winning more suitable work.

Start by separating audiences and channels as far as the evidence allows. Investigate whether the service page explains the coverage area, whether the ads reach the right people and whether enquiries receive follow-up. Improving the copy or ad targeting may be more relevant than another redesign.

  1. Observe. Describe the change you can actually see, with its period, data source and the same definition used previously.
  2. Investigate. Look for technical faults, traffic changes and differences in quality before explaining the cause.
  3. Choose one action. Agree who will make the change, what you expect to learn and when to review the outcome.

When enquiry numbers are small, one additional case can produce a large percentage change. Examine several periods and the enquiries themselves before drawing a conclusion. An average can also hide poorer performance on mobile than on larger screens. The report should help you decide what to investigate, rather than merely display the most impressive percentage.

Make the conclusion as precise as the measurement

A good result claim states its period, scope, metric and source. ‘More enquiries after launch’ describes a change. ‘The website caused the entire increase’ is a stronger causal claim requiring more evidence.

Start with what you can actually verify. If enquiries are well documented but channel attribution is uncertain, the total can still be useful when the limitation is explained. Leave uncertain channel figures out of the conclusion.

See also the checklist for following up after launch and the new website guide.

Something in mind?

Let’s see if
we are a good fit.

Tell us what you want to achieve and what is getting in the way. We can then discuss the project and the way forward.

The Færder guide

Two quick choices. A practical starting point.

What can we help you with?

Choose what best fits your situation right now.

Go straight to contact