How Does User Behaviour Affect On Page SEO?
A subject full of confident claims that are not established. This page separates what is known from what is assumed. It then gives the reason for building usable pages that holds regardless of which turns out to be true.
Do Not Optimise For The Metric
A page can improve its numbers by getting worse. The clearest example is moving the answer further down: people stay longer, scroll more and read more of the page, because you made it harder to get what they came for.
How the reversal works. The measure rewards friction.
Time on a page rises when a page is slower to use. A measure intended to indicate value can be improved by removing value, which is why it should never be the target.
Why it happens deliberately. Somebody was asked to improve a number.
Given a target expressed as a metric, the fastest route is to affect the metric. That is a rational response to the instruction rather than bad faith.
Other versions of the same move. Several.
Splitting an article across several pages, hiding content behind interactions, autoplaying video. Each improves a measure and worsens the experience the measure was meant to describe.
What the metrics are actually for. Noticing.
They are useful for spotting that something may be wrong with a page. They are not useful as objectives. The difference between those two uses is the whole of this block.
Why it is hard to spot afterwards. The report improves.
A change that worsened a page shows up as better numbers, so the reporting confirms the decision. Nothing in the data distinguishes a page that got clearer from one that got harder to use.
What to target instead. The thing itself.
Whether somebody arriving can get what they came for quickly. Improve that and the measures follow. If they do not, the page is still better.
Are These Ranking Factors
Widely assumed and not established. We state what has been published rather than what is asserted. Nothing published describes bounce rate, time on page or engagement as ranking factors.
What is commonly claimed. Direct influence.
That a search engine watches how people behave on your page and adjusts your position accordingly. Stated everywhere, rarely with a source attached.
Why the claim is attractive. It sounds obvious.
It would be reasonable for a search engine to notice that nobody stays on a result. Plausibility is not evidence. A great deal of this subject rests on plausibility alone.
What can be said. Less than people would like.
Search engines have said they use interaction data in various ways. None of the published statements amounts to bounce rate or time on page being a ranking factor. The gap between those two things is where most of this subject lives.
Why the distinction matters commercially. Work gets sold on it.
A business told these are ranking factors will pay for metric improvement. Told they are indicators, it will spend the same money on the page instead, which is the better purchase.
What we will not do. Promise a mechanism.
Where something is unestablished we say so rather than presenting it as settled. That costs us the easier sales argument. It is also the only position that survives somebody checking it.
Where the measures themselves are covered. Elsewhere.
What these numbers mean, how to read them and where they mislead is a subject of its own, covered in our measuring performance material. This page is about what to do to the page.
The Argument That Survives Either Way
A page people can use, understand and act on is worth building whether or not any system measures it. That reasoning holds no matter how the previous block eventually resolves, which is why everything below rests on it.
Why this framing is stronger. It cannot be falsified.
Advice depending on an unproven ranking claim collapses if the claim turns out to be false. Advice depending on customers being able to use your site does not.
What the page is actually for. Enquiries.
Somebody who cannot find what they came for does not enquire. That cost is immediate and commercial. It arrives whether or not a search engine ever notices.
Why this is not a hedge. The actions are identical.
Everything you would do if these were ranking factors is what you would do for a reader anyway: answer early, write clearly, make the next step obvious. The uncertainty changes nothing about the work.
What it does change. The reporting.
It changes what should be promised. Nobody should be selling improvements to a measure on the basis of a mechanism nobody has published.
How to test your own pages. Ask somebody.
Give a page to somebody unfamiliar with the business, ask them what it offers and what to do next. Their hesitation is more informative than any metric on this subject.
Bounce Rate
Only one thing here is a page-side matter: a high rate on a page meant to lead somewhere is worth investigating. What the measure means, plus why a high rate is frequently fine, is covered in our measuring performance material.
The page-side question. Was it meant to lead somewhere.
A page whose purpose is to send somebody onward, where nobody goes onward, has a problem worth looking at. That is the only inference this page will draw.
Why the qualifier matters. Most pages are fine.
A page answering a question completely should expect people to leave having got their answer. That is success. It registers identically to failure in the measure.
What to look at when it is worth investigating. The next step.
Whether a route onward exists, whether it is visible, whether it is relevant to what the page just said. Frequently one of those is missing.
What not to do. Chase the number.
Any technique aimed at the measure rather than the route is block one happening. The measure is the symptom.
Where the definition lives. The measuring material.
Including how the figure is calculated, why it varies by page type and why comparing against an industry average is close to meaningless.
Readability
Shorter sentences, plain words, meaningful headings and space. That is most of what makes a page usable, it costs nothing, then it matters most for businesses whose customers are not reading for pleasure.
Shorter sentences. The single biggest change.
Long sentences require holding several things in mind at once. Splitting them is mechanical work that improves comprehension immediately. It is also the easiest edit available.
Plain words. Not simple ideas.
Using the ordinary word rather than the impressive one. This is about vocabulary rather than substance. Complicated ideas can be expressed in plain language.
Meaningful headings. Navigation for scanners.
Headings that state something rather than naming a topic, so somebody scanning finds the part they need. The headings pillar covers the structural side.
Space. Underrated.
Shorter paragraphs and room between them make a page approachable. A wall of text is refused before it is read, particularly on a phone.
Why this matters more in the trades. The reader is not enjoying this.
Somebody researching a problem with their house is not reading for interest. They want an answer and will leave for a page that reads more easily, which is a real commercial cost.
What we do not use. A readability score.
Scores measure sentence and word length mechanically and cannot tell whether a page is clear. Reading it aloud finds more in a minute than any score.
Answer Early
A reader arrived with a question. Making them scroll for the answer loses them. No amount of quality further down recovers somebody who left in the first few seconds.
What answering early means. In the first screen.
The substance of the answer where somebody sees it on arrival, before any preamble. Not a summary of what the page will cover. The answer itself.
What gets in the way. Introductions.
Pages routinely open by establishing that the subject is important and that many people wonder about it. Nobody arrived needing to be persuaded the question matters.
Why writers resist it. It feels abrupt.
Leading with the answer reads as blunt to the person writing. It reads as helpful to the person who came for it. Those two experiences are frequently confused.
What it does not cost you. The rest of the page.
Somebody who gets the answer early and wants detail keeps reading. Answering first does not shorten the page, it reorders it.
Where this is consistent with our other material. Blogging.
The same instruction appears in our blogging guidance, because it is the single most reliable improvement available to a page that already ranks.
UX Design
Four things affect whether a page succeeds, none of them a design lecture. Navigation, obvious next steps, forms that work and a page that behaves on a phone.
Navigation. Knowing where you are.
Whether somebody arriving mid-site can tell what else exists and how to reach it. Most people arrive on an interior page rather than the home page, which is frequently forgotten.
Obvious next steps. One, visible.
What to do having read the page, stated clearly. A page with three competing requests leaves somebody choosing between them rather than acting.
Forms that work. The commonest silent failure.
Forms that submit, confirm, then reach somebody. Broken forms are the most expensive fault on this list and the least likely to be noticed, because failure produces silence.
Behaving on a phone. Where most people are.
Readable without pinching, tappable without precision, forms usable with a thumb. This is where most visits happen and where most sites are checked least.
How to find the failures. Use it yourself.
Submit your own form, on a phone, then see whether anything arrives. Most businesses have never done this on their own site, which is why broken forms survive for months at a time.
Why these four. They affect outcomes.
Each one determines whether somebody who wanted to act could. Everything else in design affects how a page feels, which matters and is a different subject.
Dwell Time
An inferred concept rather than a number in any report. Nothing tells you how long somebody spent before returning to the results, which makes most advice about it advice about something unobservable.
What the term describes. A gap.
The interval between somebody clicking your result and returning to search again. Reasonable as a concept and not a figure available to you.
Why you cannot see it. It spans two places.
The measurement would need to start on a search engine and end on one. Your own analytics see only the middle, which is why no report contains this.
What people are usually looking at instead. Time on page.
A different measure with different weaknesses, most obviously that it cannot record time on the last page of a visit. The two get discussed interchangeably.
Why it persists as advice. It sounds measurable.
The name implies a figure, so it gets treated as one. Recommendations to improve it are recommendations about a number nobody has.
Where the measures are explained. The measuring material.
Which numbers exist, what each actually records and what none of them can tell you.
Accessibility Overlaps All Of This
Almost everything that makes a page usable for everybody makes it usable for people with access needs. That is a better reason than any ranking argument, because it does not depend on one.
Where the overlap sits. Nearly everywhere above.
Clear headings, plain language, working forms, tappable targets and answering early all serve somebody using assistive software as much as anybody else.
Why the reason is better. It is settled.
Whether these measures affect ranking is arguable. Whether a broken form excludes somebody is not. Building on the second is building on firmer ground.
Why it is also an obligation. Beyond preference.
Accessibility is a legal and ethical matter for businesses serving the public rather than an optimisation. That framing changes how the work gets prioritised.
Where else this appears here. Twice already.
Heading structure and image descriptions both rest on it. In each case the accessibility reason is what makes the advice defensible where a search claim is not.
What that pattern tells you. Look for the surviving reason.
Where advice depends on an unproven mechanism, ask what else would justify it. Credibility and image work both apply the same test. The full series is on the on page SEO guide.
Give the page to
somebody who has
never seen it.
Ask them what it offers and what to do next. Their hesitation tells you more about whether the page works than any engagement measure will, taking two minutes rather than a reporting period.
What we change on a page:
We will not sell you an improvement to a measure on a mechanism nobody has published.
Every guide.
One practice.
What a page says about itself, headings, keywords, search intent, images, page types, user behaviour, credibility and how to review an existing site.
