How to choose a web analytics tool

Start with the question, not the tool

Write down the three things you would actually look at in a dashboard, then stop. Almost everybody writes something like: how many people came, where did they come from, and which pages did they read. That is a web analytics problem and it is solved by a kilobyte of JavaScript.

If instead you wrote down which cohort of signups still returns in week four, you have a product analytics problem and you want a different category of tool, a designed event taxonomy and somebody whose job includes answering it. If you wrote down why nobody clicks the pricing button, that is a behaviour question and needs heatmaps or replay. Buying across categories is the single most expensive mistake in this market, and it is made constantly.

Work out what the consent banner is costing you

This is the number almost nobody has and it is usually the largest single factor in the accuracy of their reporting. If your analytics sets cookies, you need consent in the UK and EU, and every visitor who declines is simply absent from your data. Add ad-blocker loss on top, which hits Google Analytics hardest because it is the most blocked script on the web.

You can measure the gap in an afternoon. Run a cookieless tool alongside your existing one for a month and compare the totals. The difference is what you have been making decisions without. It is common for it to be large enough to change which pages you think are working.

Price the script, not just the subscription

Every analytics tool hands your site a performance bill on every single page view, and it is paid by your visitors rather than by you, which is exactly why it gets ignored. The spread on this list is enormous: the lightest tools are one to three kilobytes, and the product analytics platforms and session replay tools are tens of kilobytes, before a tag manager adds its own.

On a content site funded by attention, that is a real cost for reports nobody opens. If your Core Web Vitals matter to you, weigh this the way you would weigh an image.

Read the metering unit and the overage policy before the price

Two vendors quoting the same price can bill you very differently, and the detail is always published but rarely compared.

Check what counts. Some tools meter pageviews only; several count custom events against the same allowance, so an event-heavy site climbs the ladder faster than its traffic implies; Matomo meters hits, which includes events, downloads and outlinks. Then check whether bots count. Roughly a third of arrivals at a typical site are automated, and a tool that filters them before the meter runs is materially cheaper than its headline suggests.

Then read the overage policy, because they differ more than the products do. On their own published terms: Fathom emails at 30 per cent over and locks the account on day seven. Plausible locks dashboards after two consecutive months. Pirsch stops collecting after five days, which leaves a permanent hole rather than a temporary lockout. Absolutely Analytics keeps collecting and keeps the dashboard open, moving you up a rung only after three consecutive months with notice. A traffic spike is the moment you most want your analytics, so this is worth more attention than a three-dollar price difference.

Decide whether you need behaviour tooling, and price it honestly

Heatmaps and session replay answer a question traffic analytics cannot: not how many people arrived, but what they did once they were there. For anyone iterating on a landing page, that is the difference between guessing and knowing.

The thing to price honestly is that most of the cookieless category does not offer it. Plausible, Fathom, Pirsch and Simple Analytics have no heatmaps at any tier, which means a second subscription. And if that second subscription is Hotjar or Microsoft Clarity, it sets cookies and brings the consent banner straight back, which may undo the reason you left Google Analytics in the first place. Tools that bundle behaviour tooling into a cookieless script avoid that, and it is the cheapest way out of the trap.

Ask where the data lives and whether you can leave with it

Two separate questions hide inside this one. Where is it stored, which is a compliance question your legal team may have a view on. And can you take it with you, which is a commercial question that matters the day you want to switch.

Self-hosted tools answer both absolutely: Matomo, Umami, GoatCounter and the rest put every raw event on your hardware. Hosted tools vary a great deal. Check for CSV export, an API, the retention period on your tier, and whether history survives cancellation. Platform-native tools are the weakest here by design, because leaving the platform means leaving the history behind.

A shortlist by situation

If you run a website and want honest numbers with no banner, shop in the cookieless category and the choice comes down to whether you want heatmaps and how many sites you have.

If you are building software, you want product analytics, and the free tiers there are generous enough that you can try two before paying for either.

If your budget is zero, self-host something open source, or use a free platform-native tool as a baseline, and accept that free behaviour tooling is paid for in data.

And if you buy Google Ads, keep Google Analytics for the conversion signal regardless of what else you run. It is the one genuinely exclusive capability in this market.