Did the Website Redesign Help? Read Before-and-After Evidence Carefully

More enquiries after a website redesign are worth investigating. Before treating the increase as a result of the new website, check what was counted, how much traffic arrived and what else changed. A larger total can describe a useful business outcome while still leaving its cause unresolved.

Start with the question the project was meant to answer. If customers struggled to request a quote, look for evidence about that journey. A fresh visual identity, a working form and more received requests are different findings. Keep each one attached to the evidence that supports it.

Agree on the comparison before reading the result

Write down the outcome in terms the person handling enquiries can recognise. For example: a form submission received by the team, or a request that the team has checked and accepted as relevant. Those definitions produce different totals. A button click, a received message and a qualified opportunity should have separate labels.

Save the reporting source, date range, time zone, page or audience filters and counting rules. Use periods that cover comparable operating conditions, and note closures, seasonal changes or a staggered launch. Equal-length windows make the arithmetic easier to read, but do not automatically make the business conditions equivalent.

Check whether measurement changed at launch. Google explains that Analytics data depends on its implementation and configuration; missing tags and consent choices can affect what is recorded. Its guidance also distinguishes Search Console clicks from Analytics sessions. Replacing one measure with another does not create a continuous baseline.

Read the record behind the headline

Consider an invented service business comparing two non-overlapping 28-day periods. The redesign launched between them. The team also increased advertising spend and changed its report so that phone-link clicks were included under “enquiries”. These fictional figures are neither a ZXP client result nor a benchmark.

The after-period report contains more kinds of activity. A phone-link click does not establish that a call connected, and the same person might click and submit a form. The form counts are received submissions, not verified counts of distinct people or qualified opportunities. Those limits matter before any percentage is calculated.

Keep the arithmetic visible

Sessions increased by 400. Dividing that change by the before total gives 400 ÷ 1,000 = 40%. Advertising spend increased by $600, or 600 ÷ 600 = 100%: it doubled. These observations describe the example’s inputs and recorded traffic; they do not tell us which visits or requests came from the ads.

The report’s enquiry label rose from 20 to 42. Its arithmetic is (42 − 20) ÷ 20 × 100 = 110%. However, the label now includes phone-link clicks, so that figure is not a like-for-like increase in received enquiries. Reporting it without the definition change would hide an important part of the comparison.

The underlying records let us compare forms separately. Received submissions rose from 20 to 28: (28 − 20) ÷ 20 × 100 = 40%. Forms per recorded session were 20 ÷ 1,000 = 2% before and 28 ÷ 1,400 = 2% after. More forms arrived, alongside proportionately more sessions.

This is a submissions-per-session ratio, not a unique-customer conversion rate. Dividing all 42 after-period actions by 1,400 sessions gives 3%, but comparing that mixed-action ratio with the earlier form-only 2% would repeat the definition error. Nor does the unchanged form ratio prove that the redesign had no effect.

Separate the observation from the cause

The table gives the owner a narrower, defensible account: “We recorded 40% more sessions and 40% more received forms across these two periods. Forms per session were 2% in both. Advertising spend doubled, and the enquiry report began including phone-link clicks. This comparison does not isolate the redesign’s effect.”

Several explanations remain possible. A different mix of visitors may have arrived. The offer or sales response may have changed. For organic-search changes specifically, Google’s investigation guidance covers factors including seasonality, technical issues and migrations, and recommends inspecting comparable periods and affected segments. A launch date alone does not settle which explanation fits.

There is no statistical-significance finding in this example. It also lacks the revenue, margin, full project costs and attribution needed to support a return-on-investment claim. Dividing all received forms into advertising spend would not establish cost per paid lead, because the table does not identify which forms came from paid traffic.

Choose the next piece of evidence

Ask the person responsible for measurement to preserve both report definitions and recover the same outcome in each period where the records allow it. If the earlier events were never collected, leave that gap visible. Start a reliable prospective record instead of estimating an attractive baseline after the fact.

Then narrow the investigation. Where the data supports it, compare the same service pages and traffic sources, keeping their counts and denominators together. Ask the receiving team how many submissions became relevant conversations using a consistent qualification rule. Record other changes, including offers, staffing and advertising, rather than treating the website as the only moving part.

End the review with a decision and a reason: correct a tracking problem, investigate a particular journey or collect more comparable evidence. Choose the next review point around the question and the business’s operating cycle, rather than declaring every redesign successful or unsuccessful after a fixed number of days.

When discussing website work with ZXP, bring the problem you want to improve, the evidence you already have and any gaps in the baseline. Agreeing what can be measured gives the eventual review a clearer purpose without promising a particular result.