The only website metrics that mean anything

Ask a marketing manager to list the metrics on their website report and you will usually get somewhere between twelve and twenty, in a spreadsheet with a tab for each month, most accompanied by a small green or red arrow. The instinct behind the length is not laziness. It is the reasonable belief that more measurement is more rigor — that a twenty-line report looks like someone is watching closely, and a six-line one looks like someone gave up early. That belief is wrong often enough to argue against directly, because the length of a report and the amount it actually explains pull apart past a certain point, and most business reporting sails straight past that point without noticing.

A business site needs about six numbers. Two of them come from the analytics tool. The other four come from the CRM and the inbox, which is exactly the part most reporting habits skip, because it takes a spreadsheet formula rather than a dashboard export. That split — and the fact that the harder half of the report is the more important half — is the actual argument here, and it holds regardless of which measurement stack sits underneath it, a question covered on its own terms in GA4 versus Plausible versus server logs.

Before getting into the six numbers themselves, it's worth pausing on who this report is actually for.

Where this gets harder for a one-person site

Everything above assumes a CRM, a shared inbox and enough deal flow to separate signal from noise across a quarter — the normal shape of a company site with more than one person selling. A single-person site, a portfolio or a freelancer's page, usually has none of that, which is worth naming alongside the tool built for exactly that smaller case: reach turns a CV and a photo into a live one-page site in under two minutes rather than an empty template. For a solo operator that speed argument outweighs any metric here — a site with zero visitors measures nothing, however good the six numbers look on paper.

reach also ships with no visitor analytics of its own, stated plainly rather than softened, and that makes the CRM-and-inbox half of this report the only half that exists for it. Anyone running a reach site is already living the argument above by necessity: enquiries and response time tracked by hand in an inbox, because there is no dashboard to lean on instead, and rarely a second page or channel to split by. None of this is the right fit for a company with a services tree, a blog and a sales team — that is the case the rest of this piece was written for.

The six numbers, each with what it cannot tell you

Qualified enquiries. Not every form submission — every one that clears whatever the business already treats as real: a phone number that isn't obviously fake, a message that matches something the company actually sells. This is the base number everything else in the report is measured against. What it cannot tell you: why someone enquired, or whether they were ever going to buy. It counts interest, not intent.

Enquiry-to-call rate. Of the qualified enquiries, how many turn into an actual phone or video conversation rather than stalling in email. A form that produces plenty of enquiries but few live conversations usually has a friction problem somewhere between submission and contact — a slow reply, an awkward booking link, a request for information the prospect already gave. What it cannot tell you: whether the conversations that do happen go anywhere. That is a separate number.

Close rate by source. Of the people who had that conversation, how many became paying customers, broken out by where they came from. This is the number that keeps the report honest about quality rather than volume, in the same way argued at length in what good actually looks like for a conversion rate: a channel producing twice the enquiries at a quarter of the close rate is not twice as good, and a report that stops at enquiry count will say it is. What it cannot tell you: the reason a source underperforms. That takes a conversation with sales, not a query.

Non-brand entrances. Sessions landing on the site from a search that did not contain the company's own name — the closest measure of whether the site is finding new people rather than being found by people who already knew to look. What it cannot tell you: quality. A page can pull in non-brand traffic for a term unrelated to what the company sells, and the number will still climb.

Page-level entrance share. Of all entrances, what portion land on each of the handful of pages that actually matter — the pricing page, the main service page, the one case study that keeps getting linked. This is diagnostic, not a score: its job is showing where attention lands so a drop in the numbers above can be traced to a specific page rather than shrugged at. What it cannot tell you on its own, with nothing else moved: anything. A diagnostic metric only means something once a scoring metric has already said something is wrong.

Time-to-first-response. From the moment a qualified enquiry lands to the moment a human replies. Every other number on this list is downstream of this one more than most reports admit — a five-minute reply and a five-day reply are, for practical purposes, two different products. What it cannot tell you: whether the reply itself was any good. Speed and quality are not the same axis, and a report that only tracks the first will eventually reward a fast, useless answer over a slow, correct one.

Why bounce rate, average session duration and total sessions are off the list

All three are on almost every report this one is replacing, and all three get argued out for the same underlying reason: each can mean two opposite things and the number alone cannot tell you which.

A bounce is a single-page visit. It happens when someone hits a broken page and leaves in disgust, and it happens when someone lands on the contact page, finds the phone number they needed in four seconds, and calls without loading a second page — a genuine success dressed up as a failure. A report that treats a rising bounce rate as bad news is guessing.

Average session duration has the mirror problem: a long session means someone was absorbed by the content, and it means someone got lost looking for a shipping policy they couldn't find. Both produce the same number, and only one of them is good news.

Total sessions — the chart that anchors most monthly reports — fails for a different reason: it moves for reasons that have nothing to do with the site doing its job. A shared LinkedIn post, a crawler having a busy week, a competitor's outage — all of it shows up as growth, none of it is evidence the site is converting better. A number that moves for reasons unrelated to what you're measuring isn't a metric. It's weather.

The four numbers that live outside the analytics tool, and the joining problem

Of the six above, only non-brand entrances and page-level entrance share come out of an analytics platform in any direct sense. Qualified enquiries, enquiry-to-call rate, close rate by source and time-to-first-response all live in the CRM, the booking calendar, or a shared inbox — places analytics tools were never built to see into.

That split creates the actual work of reporting: joining the two halves by hand. Analytics knows a session arrived from a given source; it does not know whether that session became a customer, because the CRM record that would confirm it lives in a different system with no shared identifier. The honest fix is not a fancier attribution setup — it is a single required field on the enquiry form asking how someone heard about the company, which produces a source label the CRM can carry through to the close-rate number, uncorrupted by whatever an ad blocker or device switch did to the analytics session upstream. It is a smaller fix than most teams want, and it is the one that closes the loop.

Cadence: what to look at monthly, quarterly, and essentially never on its own

Qualified enquiries and time-to-first-response are monthly numbers — cheap to pull, sensitive to real operational problems like a slow reply queue, and stable enough in a normal month that a genuine shift is worth acting on.

Close rate by source and enquiry-to-call rate are quarterly: they move on smaller sample sizes and slower cycles — a deal from March might not close until June — and reporting them monthly mostly reports noise dressed as a trend.

Non-brand entrances and page-level entrance share are almost never worth reporting on their own. They are diagnostic, not evaluative: useful once a scoring number above has already moved and someone needs to find out where, useless as a standalone row in a monthly deck built for a story it was never built to carry alone.

The number to push back on

The number a board asks for most often, and the one worth resisting hardest, is total site traffic. It is intuitive, easy to chart, and nearly disconnected from anything the business can act on, for the same reason total sessions was cut from the list above: it rewards a viral post and a bot crawl exactly as much as it rewards a genuinely better site. Reporting it without qualified enquiries beside it lets a bad quarter look like a good one, and a report that lets that happen is failing at its one job.

What a report built on six numbers actually looks like

Once the list is this short, a monthly report stops being an exercise in filling a template and starts being an actual account of what happened: qualified enquiries this month, the response time on each, the source split from the form question rather than the platform's guess, and — quarterly — whether any of it turned into revenue. Nothing on that list needs an arrow or a color to be understood, and none of it can be explained away by a crawler having a busy week. That is the same discipline argued in full in measuring what a website does: a report earns its length by what it explains, not by how many charts it took to fill the page.