Diagnose the problem before changing the page

Why Is My Website Not Converting?

Written by Neil Webley, founder of FreeCROTool and a CRO and A/B testing specialist with more than ten years of experimentation experience · Last updated 1 September 2026

A website that appears not to convert may have too little evidence, the wrong visitors, unclear messaging, weak trust, excessive friction, an unconvincing offer, a technical fault or broken measurement. These are different problems and need different responses.

Start by locating the problem. Changing headlines, buttons or colours before doing that can waste time and may make the website worse.

Step 1

Do you actually have enough relevant traffic?

A lack of conversions is not automatically a conversion-rate problem. With a small number of relevant visits, zero sales or enquiries over a short period can be normal variation.

Suppose a service page receives 25 relevant visits and its underlying conversion rate is around 2%. That would produce only half an enquiry on average. Zero enquiries from those 25 visits would not, by itself, show that the page is broken. At 500 relevant visits with no enquiries, the case for investigating becomes much stronger.

This is not a rule that every small business must wait for thousands of visitors. Broken forms, misleading information and obvious usability failures should be fixed whenever they are found. It means separating “I do not have enough evidence yet” from “I have enough relevant traffic and the outcome looks unusual.”

Step 2

Are the right people arriving?

A website can have plenty of traffic but no sales because it is attracting people who were never likely to buy. That is primarily an acquisition problem, not proof that the page fails to convert.

Use Search Console to understand intent

Open the Performance report, filter by the landing page and review the queries producing clicks. Classify the leading queries as informational, commercial or irrelevant. A visitor searching for a definition, template or free image behaves differently from someone comparing a paid service.

Also check countries and devices. Traffic from a market you do not serve, or to a page that is difficult to use on mobile, can depress the overall rate. The guide to using Google Search Console for CRO explains how query and page data can reveal this mismatch.

Use GA4 to compare acquisition sources

In GA4, compare landing page performance by session source/medium or campaign. Look beyond the blended conversion rate: do paid-search visitors reach product pages while social visitors leave after reading one article? Do branded and non-branded visitors behave differently?

For paid campaigns, compare the advert’s promise, keyword and audience with the landing page they receive. An advert offering a free guide that lands on a sales page creates a message mismatch even if both advert and page look good separately. See the practical guide to Google Analytics for conversion optimisation.

A high exit rate is not automatically bad. Someone who lands on a contact-details page, finds a phone number and calls may have completed the task. Interpret behaviour in the context of the visitor’s intent and the page’s purpose.

Step 3

Don't guess. Find where people leave.

When conversion looks poor, it is tempting to start changing headlines, colours, layouts or calls to action. First examine how visitors progress through the journey: landing page → service or product page → primary CTA → form or cart → confirmation.

Every funnel loses people, and the largest numerical drop is not automatically the problem. Many product-page visitors may reasonably decide not to add an item to their cart. Someone who starts checkout and then leaves has shown much stronger buying intent, so unexpected friction there may be more useful to investigate.

Immediate exits

Investigate search or advert intent, message match, loading problems and whether the page makes its purpose clear.

Views but no CTA interaction

Investigate clarity, perceived value, differentiation, trust and whether the CTA offers a sensible next step.

CTA clicks but no completion

Investigate form friction, validation errors, forced accounts, unexpected questions, mobile usability and reassurance.

Checkout starts but no purchases

Investigate unexpected costs, delivery options, payment failures, discount-code distraction, trust and checkout complexity.

Funnel data tells you where to look, not why the behaviour occurred. A drop may reflect normal decision-making, a problem earlier in the journey or friction at the stage itself. Use analytics alongside form testing, recordings, support questions and customer feedback to identify a plausible explanation.

CR shows the percentage of all visitors who reached each stage. Comparing progression can highlight a stage worth investigating, but a drop is not proof of a user-experience problem. The numbers need to be interpreted in light of visitor intent and what the step asks them to do.

What this checkout-friction example demonstrates

The hypothesis is deliberately narrow: reducing friction around the delivery step may allow more already-interested visitors to continue through checkout. The variant does not improve the whole funnel. In a representative run it can have slightly fewer product-page visits and add-to-cart actions, yet more visitors progress through Delivery and Checkout and ultimately purchase. That later pattern is relevant because it appears where the change was made.

For example, a simulated overall conversion-rate movement from 2.62% to 3.01% is an absolute difference of 0.39 percentage points and a relative increase of approximately 14.9%. That can matter at scale, but this simulation does not establish statistical significance, predict a real-world uplift or show that simplifying every delivery form will help.

The practical sequence is to observe behaviour, identify a plausible source of friction, form a hypothesis, test a focused change and measure the result. That produces more useful evidence than redesigning several elements because the overall conversion rate looks low. See how to tell whether a website change actually worked.

About this example: The figures are simulated and each run varies. Variants do not always finish ahead. A losing test can still prevent a harmful change from being made permanent and can help refine the next hypothesis.

Do not stop at the first apparent lead: early results fluctuate and the version in front can change. Collect sufficient evidence over an appropriate period before drawing a conclusion; there is no universal sample-size threshold. The guide to interpreting A/B test results explains the checks to make.

Step 4

Investigate the likely cause

Once you know which visitors and journey stage are affected, examine the most plausible category. Avoid making several unrelated changes at once: even if results improve, you will not know what caused the change.

Clarity and message match

What you might notice: relevant visitors leave early, scroll without interacting, or ask basic questions the page should answer. The headline may name a product without explaining the outcome, target customer or reason to choose it.

How to investigate: compare the search query or advert with the first screen of the landing page. Ask representative users what they think is offered, who it is for and what they should do next. Check whether benefits are buried beneath features and whether competing CTAs obscure the primary action.

What to change or test: make the primary benefit and audience explicit, carry the acquisition promise onto the page, explain unfamiliar terms and give the page one clear priority. Do not assume that a shorter headline is always clearer.

Trust appropriate to the decision

What you might notice: visitors view pricing, guarantees, contact or company pages repeatedly but do not proceed; sales conversations include questions about legitimacy, security, returns or what happens next.

How to investigate: review the page where it asks for money, personal information or commitment. Check that contact and company details are verifiable, testimonials are specific and authentic, claims have evidence, total costs are visible and policies are understandable.

What to change or test: answer the real concern close to the hesitation. That could mean delivery information beside “Add to cart” or clear data handling beside a form. Decorative trust badges are not a universal remedy and unfamiliar badges can create new doubt.

Necessary and unnecessary friction

What you might notice: visitors start forms or checkout but abandon particular fields or steps, repeatedly trigger validation, or perform much worse on mobile.

How to investigate: complete the journey on real phones and browsers. Review field-level errors, recordings and support reports. Check for hidden costs, forced accounts, awkward keyboards, distracting CTAs and information requested before it is needed.

What to change or test: remove or postpone unnecessary work, use clear labels and error messages, preserve entered data, explain why sensitive information is required and make the next step predictable. A lead form may genuinely need qualifying questions, but it does not need to feel difficult.

Motivation and the offer

What you might notice: the journey is usable and credible, but qualified visitors still do not see enough reason to act. They may compare prices, return several times or choose a competitor.

How to investigate: compare the customer’s actual problem with the promised outcome, price and alternatives. Review sales objections, interviews, on-site search and competitor positioning. Ask whether the CTA communicates value: “Get a tailored quote” is more informative than “Submit”.

What to change or test: strengthen relevant differentiation, make value easier to judge, package the offer around the real job and clarify genuine deadlines or availability. Do not manufacture scarcity; false urgency trades short-term pressure for long-term distrust.

Technical problems

What you might notice: sudden conversion loss, sharp device or browser differences, CTA clicks without subsequent page views, or reports that “nothing happened”.

How to investigate: submit every important form and place a test order where possible. Test common phones, browsers and screen sizes. Check buttons, error messages, payment methods, confirmation emails and browser errors. Try the journey with the site’s cookie choices, ad blockers and slower connections. Consent banners and third-party scripts can cover buttons or prevent functionality, not just tracking.

What to change or test: fix reproducible defects rather than A/B testing them. Prioritise failures that stop the primary journey, then mobile layout, validation and performance issues that materially affect use. A performance score alone does not prove a conversion problem; the user experience does.

Measurement problems

Sometimes the website is converting and analytics is wrong. An apparent zero across every funnel stage should trigger a measurement check before a redesign.

Verify a real test conversion from beginning to end. Confirm the GA4 event fires once, contains the expected parameters and is configured as a key event where required. Check thank-you page rules, form callbacks, consent states, duplicate tags and cross-domain journeys. Compare analytics with orders, CRM leads, inbox enquiries or booking records. Missing events hide conversions; duplicate events exaggerate them.

Optimising against broken measurement can reward the wrong change. Document the conversion definition and test tracking after releases, tag changes and consent updates. FreeCROTool experiments use a supported Google Analytics workflow, so accurate event collection is part of experiment setup, not an afterthought.

Website conversion diagnostic decision tree

Use these questions in order. Each answer narrows the investigation.

Do you have enough traffic to the relevant page and enough time to judge?
No: collect more evidence, widen the period or use qualitative research. Yes: continue.
Does the traffic match the offer, market and buying intent?
No: investigate keywords, targeting, source/medium and campaign message match. Yes: continue.
Do visitors progress beyond the landing page?
No: investigate intent, message match, clarity, technical performance and first-page experience. Yes: continue.
Do qualified visitors interact with the primary CTA?
No: investigate the value proposition, offer, trust, distraction and whether the CTA is a sensible next step. Yes: continue.
Do people start but fail to complete the form or checkout?
Yes: investigate friction, errors, unexpected costs, mobile usability and reassurance. No: validate the funnel definition and measurement.
Do business records show enquiries or sales that analytics does not?
Yes: repair event, consent or cross-domain measurement. No: prioritise the strongest evidence-backed explanation and form a hypothesis.

Step 5

Turn the diagnosis into a testable hypothesis

A diagnosis should lead to a specific explanation, not a random list of page edits. For example:

Visitors may not understand the main benefit of the service, so making that benefit explicit near the top of the page may increase qualified CTA engagement.

Create an alternative that tests that idea, decide which outcome would support it, then run the experiment for an appropriate period. Keep the change only if the evidence supports it. Fix confirmed bugs and compliance issues directly; experimentation is for uncertainty, not for deciding whether a broken form should remain broken.

FreeCROTool can help you create selected page variants using its visual A/B testing editor and measure experiments through Google Analytics. The documentation explains how to create a test and interpret A/B test results.

Diagnose first, then test what matters

A low conversion rate is a symptom. Establish whether the cause is evidence, acquisition, clarity, trust, friction, motivation, technology or measurement before changing the site.

Start Free See how FreeCROTool works Review the conversion checklist