What Makes a Website Actually Convert

A website has one job, and it is specific: turn a visit into the next step. An enquiry, a booking, a call, a signup. Most redesign briefs never say that out loud. They talk about looking modern, feeling premium and being clean — all of which are real, and none of which can be tested. Here is what actually moves the number.

Start with one number

Before design begins, agree three things in writing: what counts as a conversion, what the rate is today, and what one conversion is worth. Without those, every design decision becomes a matter of taste and the loudest person in the room wins. With them, arguments become experiments.

If you have no baseline because tracking was never set up, that is the first job and it takes a day. A redesign launched with no before-figure cannot be judged, and you will spend the next year guessing about whether it worked.

The first screen answers three questions or it fails

What is this, who is it for, and what do I do next. A visitor decides whether to keep reading in a few seconds, and they decide on the headline more than anything else. A concrete noun phrase describing what you do outperforms an aspirational slogan almost every time, because the slogan makes the reader do the work of translation.

The test is simple and slightly brutal: could a competitor put their logo above your headline without changing a word? If yes, the headline says nothing. Replace it with something only you could truthfully claim — a specific service, a specific market, a specific constraint you work within.

Speed is a conversion feature, not a technical detail

Aim for a largest contentful paint under two and a half seconds, interaction latency under two hundred milliseconds, and effectively no layout shift. Those are the published thresholds and they are achievable on ordinary hosting. In practice, four things cause almost all failures: oversized hero images served at desktop resolution to phones, custom fonts loading late and reflowing the page, third-party scripts, and autoplaying video.

Third-party scripts deserve particular suspicion. A chat widget, two analytics tools, a heatmap recorder, a consent banner and three advertising pixels can easily outweigh the entire site. Audit them yearly and delete the ones nobody has opened a report from.

Forms are where most of the loss happens

• Ask only for what you need to make the next contact useful — every extra field costs completions
• Put labels above inputs, not inside them as placeholder text that vanishes on focus
• Validate inline, and write error messages a human can act on
• Use correct input types so phones show the right keyboard
• Make the submit button describe the outcome, not the mechanism
• Show a real confirmation state that tells the person what happens next and when

There is an honest trade here worth naming. Removing the phone number field usually increases completions. Adding one qualifying question usually decreases completions and increases lead quality. Neither is universally right — decide which problem you have, then choose deliberately rather than by default.

Proof, and what to do when you do not have any

Testimonials work, which is exactly why so many sites invent them. Do not. Fabricated proof is both a legal risk and, increasingly, an obvious one to readers who have seen a thousand generic quotes with stock photographs.

When you have no reviews yet, substitute specificity. Publish your process step by step. Publish what is included and what is not. State your response time and then meet it. Show real work, even one piece, described honestly. Name the tools and standards you work to. A visitor cannot verify a five-star quote from an initial and a first name, but they can verify that your process page reads like it was written by someone who has actually done the work.

Structure that carries intent

One primary action per page, with secondary actions visually quieter. Informational pages should link to the commercial page that answers the question they raise. Service index pages should explain rather than list — a page of eleven links with three words each is a menu, not a page, and it converts like one.

Navigation should reflect how buyers describe their problem, not how your company is organised internally. If a service is called something only your team uses, the page will never be found and never be clicked.

Mobile is the default, not a variant

• Primary actions within thumb reach, not stranded at the top of a long page
• Tap targets large enough to hit while walking — around forty-four pixels square
• No interaction that only exists on hover
• Sticky call or messaging button on service sites where enquiries arrive by phone
• Test on a real mid-range device on mobile data, not just a desktop browser resized

What to measure after launch

• Conversion rate per page template, not just site-wide
• Form abandonment field by field, which tells you exactly which question people refuse
• Scroll depth on long pages, to find where reading stops
• Internal site search terms, which are free feedback about missing content
• Enquiry-to-client rate, because more enquiries at worse quality is not a win

The hardest part is deciding what to remove

Most underperforming pages are not missing anything. They are carrying six competing messages, four calls to action and a slider nobody has ever reached the third slide of. Conversion work is usually subtraction, and subtraction is politically harder than addition because every element on the page belongs to somebody.