Analytics gives you hundreds of numbers, and a monthly review usually turns into scrolling through them looking for something reassuring. Almost none of it changes a decision, which is why most owners stop opening it within a year.
Five numbers are enough. Each one, when it moves, tells you to do something.
1. The outcome, counted directly
Orders, qualified enquiries, sign-ups — whatever the site exists to produce, counted as a number rather than a rate. This is the only figure that survives contact with your bank account.
Count it where it lands, not where analytics thinks it happened: your order system, your inbox, your CRM. Analytics loses conversions to consent, ad blockers and payment redirects, and the gap is larger than most people expect. If the two disagree, believe the one with money in it.
2. Conversion rate, split by device
Not overall conversion — that number moves with your traffic mix and mostly describes your marketing rather than your site. Split by phone and desktop and the picture changes: most businesses have a desktop-shaped conversion rate and a mobile-shaped majority of visitors, and the average describes nobody.
A gap of more than about a third between the two is a defect, not a fact of life. It usually points at the same short list from where stores lose buyers: forms, costs revealed late, and one slow step.
3. Value per source, not sessions per source
The channel report ranks by visits, which flatters whichever channel sends the most curious and the least committed. Rank it by what each source produced instead — orders, enquiries, revenue.
The result is regularly uncomfortable: the channel with a third of the traffic delivers most of the money, and the one everybody talks about in meetings delivers almost none. Budget follows this number, or it follows a feeling.
4. Real-world speed on the two pages that matter
Not a lab score for the homepage. Field measurements — what actual visitors experienced — on the page they land on and the page where they pay or enquire. Those two carry the money; the rest of the site is a rounding error by comparison.
Watch the slow end rather than the average. Averages are pulled down by fast devices on office connections and hide the person on a mid-range phone at the end of the day, who is the visitor you paid for. When this number drifts, something was added: a script, an image, a tag somebody installed and forgot — the mechanism is in the third-party tax.
5. One operational number
The fifth is yours to choose, and it should be about work rather than traffic: hours your team spends on tasks the system should handle, support tickets grouped by cause, orders needing manual intervention, the percentage of enquiries answered within a day.
This is the number that tells you whether the business is getting easier to run. Revenue can grow while the operation quietly becomes unbearable, and nothing in a standard analytics dashboard will warn you.
What to do when one of them moves
A number on its own is trivia. What makes the review worth twenty minutes is having decided in advance what each movement means.
Outcomes fall while traffic holds: something on the site broke or got worse — start at the funnel. Mobile conversion falls behind desktop: the phone experience, not the offer. One source stops producing value while still sending visits: the traffic changed, not the site. Field speed drifts upward: something was added since last month, and the change log will name it. The operational number rises: the business grew without the machinery growing, which is the moment to automate rather than to hire.
Write those responses down once. It stops the monthly review from becoming a discussion about whether the numbers are good.
The ones to ignore
- Sessions and pageviews. Interesting when they collapse, meaningless otherwise. Traffic is an input; you can buy it, and buying it proves nothing.
- Bounce rate. A visitor who read your page, got the answer and phoned you counts as a bounce. It measures a definition, not satisfaction.
- Average time on page. Rises when people are confused as reliably as when they are engaged.
- A single lab performance score. Useful while building, misleading as a monthly metric: it is one run, on one simulated device, from one location.
- Follower counts and impressions. They correlate with activity, not with revenue. If they mattered, they would show up in the value-per-source number.
Twenty minutes, once a month
Write the five numbers in the same place every month — a spreadsheet is more than adequate — with a one-line note about anything that changed on the site or in the marketing. That note is what turns a table into a cause-and-effect record, and it is the part everyone skips.
Three months of that beats any dashboard, because you can finally see whether the thing you changed in May did anything by July. Most sites are managed on the memory of the last conversation instead.
The one to put on the wall
If only one number is allowed, use cost per outcome: everything you spent to get customers, divided by the customers you got. It contains the traffic, the conversion and the price in a single figure, and it is impossible to feel good about while the site is broken.