What Are Engagement Metrics in SEO and Do They Matter?
Dwell time, time on page and pages per visit share one problem. All three are approximations of something nobody can actually observe, which is whether a person found what they came for.
All Three Are Guesses At The Same Thing
Every engagement measure is trying to answer one question that cannot be measured directly. Was this person satisfied. Nothing recorded by any analytics platform observes satisfaction, so each of these figures stands in for it. Every one of them stands in badly in a different way.
What they are all doing. Inferring from behaviour.
Somebody stayed a while, so perhaps they were reading. Somebody looked at four pages, so perhaps they were interested. Both readings are plausible and neither is evidence.
Why the inference frequently fails. Opposite causes.
A long visit can mean absorbed or confused. Several pages can mean interested or lost. The measure records the behaviour and not the reason behind it.
The fourth member of the family. Bounce rate.
Covered separately because it carries its own argument. Everything here applies to it identically. Our bounce rate guide has it.
Why they get so much attention. They feel closer to the truth.
Traffic counts arrivals while these appear to describe experience. Appearing to describe something is not the same as describing it.
What this page does. Three at once.
What each one measures, what it cannot measure and where it is useful anyway. Taken together they are more informative than any of them alone.
Dwell Time
Dwell time describes how long somebody spends on your page before returning to the search results. The important thing about it, which almost nobody is told, is that it is not a figure any standard report gives you. It is an inferred concept rather than a measured number.
Where you will find it. Nowhere.
Open your analytics and look for it. It is not there. Neither is it in your search reporting, because neither platform is positioned to see both ends of the journey.
Why it cannot be measured by you. Two systems.
Knowing when somebody left the results and when they came back requires seeing the search engine and your site at once. You only ever see one of those.
What people usually mean by it. Something else.
Anybody telling you to improve your dwell time is generally talking about time on page, which is measurable and is a different thing. Block three covers it.
Why the term persists. A useful idea.
The concept is a reasonable way to think about whether a page satisfied a search. As a concept it earns its place. As a metric it does not exist for you.
What to say when it comes up. Ask to see it.
If somebody reports your dwell time, ask where the figure came from. That question usually resolves the conversation quickly.
The related claim. Ranking effects.
This is the measure most often described as a ranking factor, which block five addresses directly.
Time On Page
Time on page is calculated from the gap between one recorded event and the next. That sounds unremarkable until you consider what happens when there is no next event, which is exactly what occurs on the last page of every visit.
How the calculation works. Between two points.
Something is recorded when somebody arrives, something else when they move on. The time between the two is the figure. Straightforward while a visit continues.
What happens at the end. Frequently nothing.
When somebody leaves the site entirely, there may be no closing event to measure against. The visit ends and the time is recorded as nothing at all.
Which pages that affects. The ones that worked.
Pages answering a question completely are the last page of the visit by design. They are penalised in this figure precisely because they succeeded.
The consequence in a report. Backwards.
Your most effective single answer pages will show the lowest times. Read without knowing this, that looks like a list of your worst content.
What modern platforms have improved. Some of it.
Newer measurement can record additional signals during a visit, which recovers part of this. It varies by how a site is set up, which is block seven.
How to use it despite that. Compare like with like.
Set one guide against another guide rather than against a site average. Within a category the blind spot affects everything equally.
Pages Per Session
How many pages somebody viewed in one visit. Universally treated as better when higher, which is the reframe worth making here. Somebody who found the answer on the first page has been served better than somebody who had to look through four.
The two readings. Opposite meanings.
A rise means people are exploring. Or it means people cannot find what they came for. Both produce the same figure and the second is more common than anybody admits.
Where high genuinely is good. Shops and catalogues.
Somebody browsing products, comparing options or working through a range. There, more pages is engagement doing what it should.
Where high is a warning. Service sites.
Somebody looking for your prices, your area or your telephone number should reach it immediately. Four pages means your navigation failed them.
The pattern worth checking. Which pages.
Not how many but which. Several visits to a contact page suggests somebody who could not find something rather than somebody keenly interested.
What it cannot distinguish. Purposeful from lost.
The count is identical either way. Only knowing what the pages were, plus what the site was supposed to do, tells you which happened.
The trap in improving it. Splitting content.
Breaking one complete answer across several pages raises the figure and makes the site worse. That is the same failure our bounce rate guide describes.
Are They Ranking Factors
All three are widely assumed to affect where a site appears in search results. That position is not established. Our standing approach across this site is to say what is published rather than what is asserted. We neither claim they are used nor claim they are ignored.
Why the belief is so widespread. It sounds right.
A search engine wanting to show useful results might reasonably notice whether people stay. Plausible is not demonstrated. This industry has repeatedly treated the two as the same.
The practical difficulty with the theory. The data is yours.
Time on page and pages per visit are calculated by your own analytics, configured by you, using conditions you can change. Figures you control are a strange foundation for anybody else's assessment.
The exception in the family. Dwell time.
Because it concerns behaviour on a results page, it is at least theoretically observable from the other side. That establishes nothing about how it is used.
What we will not do. Assert either way.
Claiming these are ranking factors would be repeating an assumption. Claiming they are definitely not would be the same error in the opposite direction.
What follows for you. Do not chase them.
Work aimed at moving these figures in the hope of improving rankings is aimed at an unestablished connection. It frequently damages the page as well.
The consistent standard. Across the site.
The same evidence rule applies wherever we discuss what search engines do, including everything in our generative engine optimisation material.
Where They Are Genuinely Useful
These measures earn their place in three situations. All three share one feature. They compare things within your own site rather than treating any figure as meaningful on its own.
Comparing similar pages. The main use.
One service page against another. Or one guide against another. Same kind of page, same audience, same measurement conditions, so a difference between them means something.
Spotting a page that is obviously failing. Genuinely useful.
Where one page sits far outside the pattern of everything like it. That is worth investigating and it is the kind of outlier these figures find well.
Watching a direction over time. On a single page.
Whether a page you changed behaves differently afterwards. Holding everything else still makes the before and after comparison legitimate.
What they are never useful for. Absolute judgement.
No value is good or bad in itself, which is why we publish no target for any of them and why you should distrust anybody who does.
The comparison to avoid. Against another business.
Two sites are frequently not measuring the same thing at all, which is block seven and it makes external comparison worthless.
The rule that covers all three. Within, not against.
Comparison inside your own site is informative. Comparison against anything outside it is not.
Two Sites Are Not Measuring The Same Thing
These figures are affected by how a site is built and how its measurement is configured. Two businesses can produce completely different numbers for identical visitor behaviour, which quietly undermines every comparison anybody makes with them.
How the site affects it. Page structure.
A site loading content as somebody scrolls behaves differently from one loading a fresh page for each step. The same reading produces different records.
How the configuration affects it. What counts.
Measurement can be set to record additional signals during a visit or not to. That choice changes every engagement figure on the site.
What consent does. Reduces the sample.
Where visitors decline tracking, their behaviour is not recorded. The figures then describe the people who agreed rather than everybody.
Why this is rarely mentioned. It undermines the numbers.
Acknowledging it means acknowledging that these figures are not comparable between sites, which removes most of what they are used for in proposals.
The one thing it does not affect. Your own trend.
If your configuration stays constant, your figures remain comparable against themselves. That is why internal comparison survives all of this.
What to check first. Did the setup change.
A figure that moved sharply with no change in enquiries is usually a measurement change. Our Google Analytics guide covers the platform.
What To Do With Them
Treat all three as diagnostic prompts rather than as results. They suggest where to look rather than telling you what is wrong. Used that way they are genuinely helpful. Used as objectives they do real damage.
What a prompt looks like. A question raised.
One page behaving unlike everything similar to it is worth a look. That is the whole contribution and it is a useful one.
What you do next. Look at the page.
Read it as somebody arriving from that search. The figure told you where to look and your own judgement does the rest.
Why they must not be targets. All three can be gamed.
Splitting content, hiding answers and adding interruptions improve every one of these while making the site worse for everybody using it.
What they never belong in. The top of a report.
They describe behaviour on your site rather than anything the business gained. Which measures matter covers where they sit.
What to fix a page against instead. Its purpose.
What the page was built to achieve, written down. Judging engagement without that is judging against nothing.
Who owns the fixing. Not this series.
What to actually change on a page belongs with our on-page material. This series covers what the measures mean.
Where everything sits. The hub.
All nineteen guides are on the measuring performance guide.
Dwell time is
not a figure
any report
gives you.
Open your analytics and look for it. It is not there. Neither is it in your search reporting, because knowing when somebody left the results and when they returned requires seeing both the search engine and your site at once. You only ever see one of those. Anybody telling you to improve it is generally talking about something else entirely.
How we treat engagement measures:
Your best single answer pages will show the lowest times on page, because a visit ending there frequently records nothing at all.
Every guide.
One question.
What each number actually means, how to read the data without being misled, what belongs in a report and what does not, then how to connect any of it to the enquiries the business is actually paying for.
