Six numbers beat forty dashboards

The dashboard nobody opens

Almost every small business site we take over has analytics installed. Almost none of them have analytics that anyone reads.

The pattern is consistent. Someone set it up at launch. It collects everything. It has twenty-four reports, four of which are about audience demographics that are largely inferred, and none of which answer the question the owner actually has, which is: is this website getting me work?

So it gets opened when a marketing company asks about it, and otherwise never.

That is not a discipline failure. It is a design failure. A tool that requires forty minutes and a mental model of event parameters to answer a one-sentence question will not get used by someone whose actual job is running a business.

Two tools, two questions

Before the numbers, the split that clears up most confusion:

Search Console tells you what happens before the click. Which queries showed your pages, where you ranked, how often people clicked through. Google's own record of Google's own behaviour. There is no substitute for it and it is free.

Analytics tells you what happens after the click. Which pages they read, how they moved, whether they did the thing you wanted.

People install the second and skip the first, which is backwards for most small businesses — the pre-click data is scarcer and harder to reconstruct. Install both. Read them for different reasons.

The six numbers

Here is the set we put on a one-page monthly review for clients. It is not exhaustive, and it is deliberately not exhaustive.

1. Contact events

Form submissions, tel: link taps, and direction requests, added together.

This is the number. Everything else on this list exists to explain movements in this one. If it goes up and nothing else does, the site is working. If it goes down while traffic goes up, something in the path to contact broke.

Count intent, not just completion — a tel: tap is a person choosing to call you, even if you cannot see whether the call connected.

2. Contact rate

Contact events divided by sessions, as a percentage.

Rate matters more than raw count because it separates two different problems. Falling traffic with a stable rate is a visibility problem — an SEO or listings problem. Stable traffic with a falling rate is a site problem — something about the page stopped persuading people, or stopped working. Those get fixed in completely different places.

Do not compare your rate to an industry benchmark. Compare it to your own last quarter.

3. Sessions, split by channel

One number with four rough buckets: organic search, direct, referral, social.

You do not need per-source detail. You need to know whether the growth came from search (durable, compounding), from a one-off referral (nice, temporary), or from a post that did well (nice, extremely temporary). That distinction should change what you do next, and it usually does not get made because the report is too granular to read.

4. Top five landing pages

Not top pages — top landing pages. The page people arrived on.

This is consistently the most surprising number for owners. A great many small business sites get most of their search entries on a page nobody thought of as important: one service page, one location page, one blog post from three years ago. If a page is bringing people in, it deserves a proper headline, a clear next step, and a place in your refresh rotation.

5. Mobile share

The percentage of sessions on a phone.

You need this figure because it decides where your attention goes. It is not the same for every business — a B2B supplier serving office workers and a taco shop have very different splits — and using someone else's number as a proxy for your own is how sites end up designed on a monitor for an audience on a phone.

6. Core Web Vitals from real users

LCP, INP and CLS as experienced by actual visitors, not a lab simulation.

Search Console reports this if you have enough traffic. If you do not, you can collect it yourself with a small script using the browser's own Performance API, or you can fall back to lab tools. Field data is the version that counts, because it includes the visitor on a five-year-old Android on a weak connection — the visitor a lab test on a simulated device never quite represents.

The events worth defining

Most analytics tools track pageviews automatically and everything else on request. For a small business site, four custom events cover it:

• contact_submit — a successful form submission. Fire it on the confirmation, not on the button click, or you will count failures as successes.
• call_click — a tap on any tel: link. Include which page it happened on.
• directions_click — a tap on a map or address link.
• quote_start — if you have a multi-step form or a booking flow, an event when someone begins it. Paired with contact_submit, this gives you an abandonment rate, which is the only reliable way to find out whether your form is asking for too much.

That is four events. Not forty. Every event you define is something you have to maintain, and unmaintained events go quietly wrong — a class name changes in a redesign, the listener stops firing, and six months of data are missing before anyone notices.

Measure the measurement

An analytics script is a third-party resource on every page. It has a size, an execution cost, and, usually, a DNS lookup and a TLS handshake to a domain that is not yours.

Two habits keep it honest:

Load it deferred. defer or async, always, so it cannot block rendering. There is no version of analytics that is important enough to delay your content.

Test the page with and without it. Run your homepage through a lab tool twice — once as shipped, once with the analytics script removed. The difference is what measurement costs you, on every visit, forever. Some tools cost almost nothing. Some cost a surprising amount, particularly the ones bundled with tag managers that then load four more things.

Knowing the number does not mean you must remove it. It means you are making the trade deliberately, which is the whole point of treating page weight as a budget rather than an accident.

A review that takes fifteen minutes

Once a month:

1. Open the six numbers. Write them down next to last month's.
2. Ask one question of each direction of change: what would explain that?
3. Open Search Console's query table, sorted by impressions, and read the first twenty rows.
4. Pick one thing to change. One.

That last step is the part that makes the rest worth doing. Analytics that never leads to a change in the site is a hobby. The value is not in the measuring, it is in the fact that next month you will find out whether the change worked — which requires that you only changed one thing.