Website statistics: understanding visitors and taking action
Visits, bounce rate, sources and conversions: learn how to read website stats with concrete examples and choose an improvement to test.
Sur cette page6 sections
Your site received more visits this month. Good news? Perhaps. If these visitors land on an article that answers their question, but are not the people you actually sell to, the numbers can go up without your phone ringing any more often.
Conversely, a page can receive few visits yet regularly bring in very useful enquiries. I prefer to look at what people are trying to do on the site rather than deciding that a big upward curve is necessarily a good curve.
Statistics are for understanding this. Not for giving the site a grade, but for spotting where it helps your visitors and where it leaves them to fend for themselves.
Page views indicate recorded displays. If someone returns to your pricing page three times, it could mean they are comparing before deciding. It could also mean they aren't finding the answer. The figure alone doesn't distinguish between the two.
The entry page tells you where the visit begins. This is important: not everyone goes through the homepage. Someone might land directly on your article, your service, or your contact page. These pages must be self-explanatory without forcing the visitor to start their journey from the beginning.
And an exit isn't necessarily a drop-off. After a booking confirmation, leaving is normal. In the middle of a form, it's more interesting to examine. We look at what was supposed to happen next, not just the number of doors closing.
This is why the bounce rate shouldn't be read as a dissatisfaction rate. Its definition depends on the tool and the actions recorded. In Google Analytics 4, it corresponds to the share of sessions without engagement; this engagement depends notably on time, views, and key events. Google explains its calculation method here. Matomo utilise une autre logique, décrite dans its bounce documentation. We don't compare the two figures as if they were telling exactly the same story.
Engagement time gives another clue, but it's not a reading stopwatch. In GA4, it is based on the time the page is in the foreground, as explained in Google's documentation. This doesn't tell you if the person understood your offer or if they liked it.
A long visit can be encouraging on an article. On a page where someone is just looking for a number, it can reveal a difficulty. I would therefore look at the duration alongside the action taken: a call click, a form submission, a booking. Without this follow-up, you have a clue, not yet an explanation.
Traffic sources allow you to separate these situations: organic search, social media, advertising, a site citing you, or an email campaign. The "direct" category isn't proof that everyone has memorised your address. It can also contain visits whose origin wasn't correctly identified.
If you run a campaign, tagging its links helps track what it brings in. Then we look at the landing page and the enquiries received. An advert can get many clicks but send people to an offer that doesn't match what was promised.
Take a tradesperson working in a few towns. A national article might bring them many readers. A local page might bring fewer, but more enquiries from their area. The right comparison isn't just "which channel gets the most visits?". It's also "which one brings in people I can actually help?".
Start by going through the journey on your own phone. Does the keyboard hide a field? Does the button work? Is it impossible to attach a document? You don't need a major audit to discover an error that happens every time you try.
Speed also matters in this journey. A page might display its title quickly but take much longer to make its button usable. The Core Web Vitals distinguish between the arrival of main content, responsiveness, and layout shifts. These are useful benchmarks for investigating, not an obligation to win a scoring contest.
Look at important pages and their real visitors when these metrics are available. Optimising only the homepage doesn't fix a slow booking page where campaigns land. And an average calculated on your office connection doesn't describe all mobile phones.
A click on "send" isn't enough either if the form displays an error immediately after. And a click on the phone number shows an intent to call, not proof that a conversation took place. The measurement must match the action it claims to track.
I advise testing this chain before interpreting its figures. You make a test request, check its receipt, and see what the tool recorded. If it counts three conversions for the same test, your priority is to fix the counter, not celebrate the result.
Then, you relate the measurement to your business. Ten forms filled with out-of-area requests or spam are not worth ten prospects. Statistics show what the site records; your sales tracking completes what they cannot know.
You check the journey, fix the field that's blocking them, then observe the results over a comparable period. If you change the form, the prices, the advertising, and the page title all at once, you will have a much harder time understanding where the difference comes from.
With few visits, you must also accept not jumping to conclusions too quickly. Two requests versus one the previous week is one extra request, not yet a solid trend. Time periods, seasons, and campaigns must be factored into the comparison.
At Cadarsir, monthly SEO monitoring is used to turn these observations into understandable decisions: reworking a page, testing a journey, improving information, or working on a visibility source. If the site regularly requires changes, this work is coordinated with its maintenance and evolution.
Ultimately, I want you to be able to answer "what have we understood and what are we going to change?". That is more useful than "the rate went up by three points", especially if no one knows what it's measuring.
Conversely, a page can receive few visits yet regularly bring in very useful enquiries. I prefer to look at what people are trying to do on the site rather than deciding that a big upward curve is necessarily a good curve.
Statistics are for understanding this. Not for giving the site a grade, but for spotting where it helps your visitors and where it leaves them to fend for themselves.
A visit is not a customer, a page view is not an opinion
The same visitor can open several pages, return the next day, or change devices. Tools don't group these visits in the same way, and some visits may escape measurement. Therefore, the counter doesn't represent an exact list of people sitting in front of your site.Page views indicate recorded displays. If someone returns to your pricing page three times, it could mean they are comparing before deciding. It could also mean they aren't finding the answer. The figure alone doesn't distinguish between the two.
The entry page tells you where the visit begins. This is important: not everyone goes through the homepage. Someone might land directly on your article, your service, or your contact page. These pages must be self-explanatory without forcing the visitor to start their journey from the beginning.
And an exit isn't necessarily a drop-off. After a booking confirmation, leaving is normal. In the middle of a form, it's more interesting to examine. We look at what was supposed to happen next, not just the number of doors closing.
"They leave quickly" can also mean you helped them
Imagine someone looking for your Saturday opening hours. They open the page, find the information, and close the tab. They might have spent five seconds on your site. Yet, you provided exactly the service expected.This is why the bounce rate shouldn't be read as a dissatisfaction rate. Its definition depends on the tool and the actions recorded. In Google Analytics 4, it corresponds to the share of sessions without engagement; this engagement depends notably on time, views, and key events. Google explains its calculation method here. Matomo utilise une autre logique, décrite dans its bounce documentation. We don't compare the two figures as if they were telling exactly the same story.
Engagement time gives another clue, but it's not a reading stopwatch. In GA4, it is based on the time the page is in the foreground, as explained in Google's documentation. This doesn't tell you if the person understood your offer or if they liked it.
A long visit can be encouraging on an article. On a page where someone is just looking for a number, it can reveal a difficulty. I would therefore look at the duration alongside the action taken: a call click, a form submission, a booking. Without this follow-up, you have a clue, not yet an explanation.
Where do these visitors come from, and what were they looking for?
Someone who types your name might already know you. Someone arriving from an advert is discovering a specific promise. Someone who finds a tutorial might simply want to solve a problem, without needing your service.Traffic sources allow you to separate these situations: organic search, social media, advertising, a site citing you, or an email campaign. The "direct" category isn't proof that everyone has memorised your address. It can also contain visits whose origin wasn't correctly identified.
If you run a campaign, tagging its links helps track what it brings in. Then we look at the landing page and the enquiries received. An advert can get many clicks but send people to an offer that doesn't match what was promised.
Take a tradesperson working in a few towns. A national article might bring them many readers. A local page might bring fewer, but more enquiries from their area. The right comparison isn't just "which channel gets the most visits?". It's also "which one brings in people I can actually help?".
The user journey often tells the story better than the average
On a computer, visitors fill in your form. On a phone, they start and then stop. If you only look at the overall rate, the good results from the first group can hide the difficulties of the second.Start by going through the journey on your own phone. Does the keyboard hide a field? Does the button work? Is it impossible to attach a document? You don't need a major audit to discover an error that happens every time you try.
Speed also matters in this journey. A page might display its title quickly but take much longer to make its button usable. The Core Web Vitals distinguish between the arrival of main content, responsiveness, and layout shifts. These are useful benchmarks for investigating, not an obligation to win a scoring contest.
Look at important pages and their real visitors when these metrics are available. Optimising only the homepage doesn't fix a slow booking page where campaigns land. And an average calculated on your office connection doesn't describe all mobile phones.
What do we actually want to count?
A conversion is an action you have defined as useful: a quote request, a booking, a purchase, or a sign-up. You can also track intermediate steps, like starting a form. But they must be named correctly: starting a request is not the same as sending it.A click on "send" isn't enough either if the form displays an error immediately after. And a click on the phone number shows an intent to call, not proof that a conversation took place. The measurement must match the action it claims to track.
I advise testing this chain before interpreting its figures. You make a test request, check its receipt, and see what the tool recorded. If it counts three conversions for the same test, your priority is to fix the counter, not celebrate the result.
Then, you relate the measurement to your business. Ten forms filled with out-of-area requests or spam are not worth ten prospects. Statistics show what the site records; your sales tracking completes what they cannot know.
Choosing a modification that can be understood later
To avoid spending the month looking at spreadsheets, choose one page and one question. For example: why do mobile visitors start the form without finishing it?You check the journey, fix the field that's blocking them, then observe the results over a comparable period. If you change the form, the prices, the advertising, and the page title all at once, you will have a much harder time understanding where the difference comes from.
With few visits, you must also accept not jumping to conclusions too quickly. Two requests versus one the previous week is one extra request, not yet a solid trend. Time periods, seasons, and campaigns must be factored into the comparison.
At Cadarsir, monthly SEO monitoring is used to turn these observations into understandable decisions: reworking a page, testing a journey, improving information, or working on a visibility source. If the site regularly requires changes, this work is coordinated with its maintenance and evolution.
Ultimately, I want you to be able to answer "what have we understood and what are we going to change?". That is more useful than "the rate went up by three points", especially if no one knows what it's measuring.
