My website looks good but doesn’t convert: ten causes you can check

The site looks great. Visits come in. And yet almost nobody signs up, requests a quote or buys. The reflex is to redesign. It is rarely the right starting point: a page can be beautiful and still leave visitors without an answer to the three questions they ask in the first seconds — what is this, is it for me, what should I do now. Here are the ten places I look at first, in the order a visitor meets them, with what you can check yourself.

First: is it really the site?

A page does not convert visitors who are not your audience. Before changing anything, look at where the traffic comes from: a poorly targeted campaign, an article that attracts the curious or an ambiguous keyword produce visits that could never convert. Then measure the main action at each step — view, click, submit, confirmation — with events that carry no personal data. Without that baseline, every fix remains an opinion.

A word on volume: below a few thousand visits a month, an A/B test almost never reaches a conclusion. Telling two versions apart takes several hundred conversions per variant. With little traffic, a usability test with five people from your audience reveals more, faster.

The ten causes, in one table

1. The first screen does not say what you do

A headline like “Welcome”, “Revolutionize your workflow” or “The all-in-one platform” could sit on a thousand sites. The simplest test: hide your logo and read the first screen. If the sentence could describe a competitor, it does not describe your offer yet. A good headline names the outcome, for whom, and if needed what people no longer have to do; the subheading explains how, in one or two sentences. To validate, show the page for five seconds to five people who do not know it, then ask: “What does this site offer, and to whom?”

2. Google’s promise is not the page’s

The visitor clicked on a promise: the title and description shown in search results, or the preview of a shared link. If they land on a different message, they wonder whether they are in the right place. Check that the <title> tag names the offer rather than “Home”, that the meta description restates the page’s promise, and that a share image exists — without it, a link posted on LinkedIn or WhatsApp shows up with no visual.

3. The form asks for too much, too soon

Every field costs effort and needs a reason to be filled. For each one, ask: “do I need this to reply?” A required phone number, the company name, a detailed budget and a password confirmation can often wait for the first exchange. Check the mechanics too: a visible label above every field (example text that disappears as you type is not a label), the right field type for email and phone, and the autocomplete attribute on personal fields — WCAG asks for it (criterion 1.3.5) and auto-fill shortens the form on a phone.

One non-negotiable point: no newsletter or consent box ticked in advance. In the EU a pre-ticked box is not valid consent (GDPR recital 32; CJEU Planet49 ruling, 2019). A good default is reversible and never decides a purchase or a consent on someone’s behalf.

4. After the click, the unknown

A “Get started” button with no hint of what follows lets people imagine the worst: a long form, a credit card, a sales call. Write under the button what really happens — number of steps, time, whether a card is needed, when the first result appears. If sign-up is long, ask for the minimum to create the account and the rest afterwards, in short steps with real progress. People speed up when they see the goal getting closer (the goal-gradient effect) — provided the progress is true, never inflated.

5. Everything asks for commitment before giving anything

If every button leads to “Sign up”, “Buy” or “Contact us”, visitors must commit before seeing the value. A real sample deliverable, a demo without an account or a small tool that produces a first result changes the exchange: people give back more readily after receiving something useful. One condition: it must really be useful, with no hidden catch.

6. Visitors cannot create anything of their own

A page that only explains leaves the visitor passive. A price estimate, a plan or a selection they build themselves is worth more to them, simply because they made it (the IKEA effect). Then offer to save it, without holding it hostage: what they created stays theirs. If the page speaks to several audiences, let visitors say who they are (“I am…”) instead of making them sort everything.

7. The risk of saying yes is never addressed

A possible loss weighs more than an equal gain. Near prices and forms, state honestly what is reversible: no commitment or minimum term, refund or not, data export, what happens to the email address. The opposite is a trap: a countdown that restarts, or an unverifiable “only 3 spots left”. If the urgency is real, give its reason; if it is not, remove it — in the EU, falsely stating that an offer is only available for a very limited time is a banned commercial practice (Directive 2005/29/EC, Annex I, point 7).

8. Hidden prices, or a doubtful anchor

Without a price or an order of magnitude, part of your visitors leave rather than ask. A “from” price, a range or a real priced example is often enough; if everything is quoted, say how the quote is calculated. With several plans, name the profile each one fits. And be careful with crossed-out prices: in the EU a reduction must refer to the lowest price charged in the previous 30 days (Directive 98/6/EC, art. 6a); in Switzerland the Price Indication Ordinance also regulates comparison prices. Comparing two real plans keeps the contrast effect without the risk.

9. Too many actions, empty labels

When the first screen offers three equally weighted actions, none stands out. Keep one main action, clearly more visible, and at most one secondary one. Then reread the labels: “Learn more”, “Discover” and “Submit” say neither what the click brings nor what happens next. Write them as the end of the sentence “I want to…”: “See a sample report”, “Estimate my price”, “Book a 20-minute call”.

10. Slow, unreadable or doubtful on a phone

Most of these checks are done on a phone. Google considers the main content’s render time (LCP) good under 2.5 seconds and poor above 4; a layout that shifts while loading (CLS above 0.1) makes people tap the wrong thing. Text must reach a 4.5:1 contrast (WCAG 1.4.3) and zoom must not be blocked. Then trust: who is behind the site? In Switzerland anyone selling online must clearly state their identity and a contact address, email included (UWG/LCD art. 3 para. 1 let. s). And real proof — a named client with consent, a result you can show — beats ten unverifiable testimonials. Never invent one.

Six behavioural levers, and their guardrails

The recommendations above rely on six well-documented effects. Each can help a decision — or force it. The difference is the guardrail.

Validate without fooling yourself

Every fix is a hypothesis: “if the headline names the outcome, more visitors will reach the main action”. Nobody knows the figure before measuring it — be wary of anyone who promises a percentage uplift in advance. To validate, in this order: the five-second test for clarity, a usability test with five people for the flow, funnel events compared over a few weeks before and after, and only with enough traffic, an A/B test.

Where to start

The fastest way is to run the free scan of your page: it reads the page’s real code (and, when possible, its rendering on a phone by Google Lighthouse) and gives you three priority findings, each with what was read, why it matters and how to fix it. Nothing is invented: what cannot be checked automatically is labelled as such.

• Fix it yourself: every finding in the report points to the correction and how to check it.
• Learn to do it: Build School teaches you to build and verify a product with AI, measurement included.
• Hand it to me: Conversion Rescue — an on-screen review of your pages, or the fixes, on a written quote.
• If the site looks dated or confusing before anyone reads it: Design Rescue.
• If something does not work (a form, a payment, a button): Code Rescue. Before a launch: Launch Rescue.