A practical look at what changes a page's performance — structure, message and speed — and what only looks like it does.

Three things move the number more than anything else, and none of them is the thing most people ask us about first.

Structure

What each page is for, and in what order the argument is made. A visitor who cannot tell within one screen whether a page is for them will leave, and no amount of motion design will hold them. Structure is not the same as layout — a page can be visually polished and structurally confused if the order of information does not match the order a real visitor actually needs it in.

Why structure gets skipped

Structure is invisible in a mockup the way a typo is invisible in a font specimen — everything looks finished, because "finished" in a design review usually means every section is styled, not that the sections are in the right order for a real decision-making process. A stakeholder reviewing a static design rarely asks "would I actually keep reading here as a visitor with this specific problem," because they already know what the page is trying to say. The visitor arriving cold does not have that advantage, and structural problems only become visible once real traffic starts behaving differently than the review room expected.

Message

The words a company uses internally are almost never the words its customers search for. The gap between those two vocabularies is, in our experience, the single largest source of wasted traffic. A company that calls its offering "enterprise workflow orchestration" internally is invisible to the buyer searching "how to stop losing track of approvals," even though those two phrases are describing the same product. Closing that gap is not a copywriting exercise so much as a research one — it means finding the actual words real prospects use, in support tickets, sales call transcripts, and search queries, and writing toward those words instead of the internal vocabulary a team has settled into over years of talking to itself.

Speed

Not a Lighthouse score — the time between clicking and being able to read. Those are different numbers, and only one of them is felt by a customer. A page can score well on every synthetic benchmark and still make a visitor wait through a blank screen before the content that convinced them to click appears in the first place — and a visitor who is still waiting has not yet started being persuaded by anything on the page.

What does not move it

Carousels. Testimonial sliders nobody waits through. Illustration styles chosen because a competitor had one. Cookie banners that block the first screen. These are not neutral — several of them measurably cost you enquiries — but the reason they persist is that they are easy to decide and easy to approve. A carousel gets approved because it lets five stakeholders each get their slide, which resolves an internal disagreement about whose message goes first without anyone having to actually choose. The cost of that resolution is a feature that, by design, only ever shows a fraction of its content to a given visitor, and shows nobody the slide that was actually relevant to them.

The list of things that feel important and are not

Beyond carousels, the recurring offenders are predictable: a hero video that autoplays and adds load time before saying anything a static headline could not; a chatbot that pops up before a visitor has read a single sentence of context; an animated counter ticking up to an impressive-sounding statistic that nobody can verify by clicking through. Each of these gets proposed with genuine confidence that it will help, usually because it worked, once, somewhere else, for a business in a different situation with a different problem to solve. None of them address structure, message or speed — they are activity that resembles progress without being progress.

How we measure it

Enquiries per thousand sessions, segmented by landing page and by channel. It is a blunt number and it is honest, which is more than can be said for time on page. We take a baseline before anything ships, because a redesign without a baseline is just an opinion with a launch date. Segmenting by landing page matters specifically because an aggregate site-wide number hides which pages are actually underperforming — a strong homepage can mask three weak category pages if you are only ever looking at the total.

What a baseline actually requires

A baseline is not a vague sense that "the old site converted okay." It is a specific number, measured the same way you intend to measure the new site, over a long enough window to smooth out weekly noise — usually a full month, sometimes a full quarter for lower-traffic sites. Without it, a redesign's success is judged by whether people in the room like the new version more than the old one, which is a real signal about taste and no signal at all about enquiries. We have seen redesigns that everyone preferred visually underperform the site they replaced, and the only reason anyone caught it was that a baseline existed to compare against.

Why we ignore time on page

Time on page is one of the most commonly cited engagement metrics and one of the least useful for a commercial site, because it cannot distinguish between a visitor absorbed by a compelling argument and a visitor confused and re-reading a paragraph trying to work out what you are offering. Both produce a long session. Only one of them is good news. Enquiries per thousand sessions does not have this ambiguity — it counts the thing you actually wanted to happen, not a proxy that could mean the opposite of what it looks like it means.

The order to fix them in

Structure first, because a well-structured page with mediocre copy still gets a confused visitor to the right information eventually, while a beautifully written page with confused structure loses that same visitor before the copy gets a chance to work. Message second, once the structure is settled and there is a stable place to put the right words. Speed last is not because it matters least — it is because it can usually be fixed without touching structure or message at all, so there is no dependency forcing it to come first, and fixing it after the other two avoids re-testing performance every time a content decision changes.

A short diagnostic

If conversion is underperforming and you are not sure which of the three is the problem, this order of questions usually narrows it down fast. Can a visitor tell within one screen whether this page is for them — if not, that's structure. If they can tell it's relevant but do not act, read the headline and ask whether it uses your internal vocabulary or your customer's — that's message. If the page says the right thing to the right person and still underperforms, check how long it takes to become readable on a throttled connection — that's speed. Most conversion problems trace to exactly one of these three, and fixing the wrong one first is how a redesign budget gets spent without moving the number it was meant to move.

In short

Fix structure, close the vocabulary gap between how you talk and how customers search, and make the page fast to become readable. Everything else on the usual redesign wishlist is a distraction from those three, however confidently it gets proposed.