How to choose a web analytics tool

Most teams choose a web analytics tool by default — it was already installed, or it was free — and only discover what they needed when a question arrives that the tool cannot answer. Choosing deliberately takes an afternoon. This guide walks through the four decisions that actually separate the options, in the order that eliminates the most candidates fastest.

1. Decide where the data is allowed to live

Start here because it is the only question with a hard yes or no. If your organisation, your clients or your regulator require that visitor data stays on infrastructure you control, or inside a particular jurisdiction, then a large part of the market is gone before you look at a single feature.

Three deployment models exist:

• Vendor cloud, vendor's terms. The vendor hosts the data and may use it, in aggregate, for its own products. Google Analytics 4 is the familiar example.
• Vendor cloud, your terms. The vendor hosts it, contractually as your processor only, often in a region you pick. Piwik PRO and the hosted versions of Matomo and Plausible sit here.
• Self-hosted. You run it. Open-source tools make this possible; it moves the cost from licence fees to the people who keep it running.

If nobody in your organisation has an opinion on this, ask legal before you ask marketing. The answer is cheaper to get now than after implementation.

2. Decide what you need to measure

Be honest about the questions you ask monthly, not the ones a demo makes look exciting.

"How much traffic, from where, to which pages?" Every tool on the market does this. A lightweight, cookieless product will answer it with a script a fraction of the size, no consent banner in many jurisdictions, and a dashboard that needs no training.

"Which campaigns and channels produce conversions?" You need reliable UTM parameters handling, goal or conversion tracking, and ideally a connection to your ad platforms so that results flow back into bidding. This is where integration with the advertising ecosystem you buy from matters most.

"What do people do inside the product?" Once the questions are about sequences of actions by identified users, you have left web analytics. Look at product analytics tools instead; they are built around event tracking, funnel analysis and retention analysis and will serve you better than a web tool stretched to fit.

"All of it, across brands, joined to offline data." That is the enterprise tier — Adobe Analytics or the paid tier of Google's product — and it comes with an implementation project, not a script tag.

3. Decide how you will handle consent

Privacy regulation changed what "measuring everything" means. Whatever you pick, you need an answer to two things.

First, what happens when a visitor declines tracking? Some tools drop the visit entirely. Some collect an anonymous, cookieless hit. Some use consent mode to model the missing behaviour statistically. None of these is wrong, but they produce different numbers, and you should know which one your reports are showing. If you compare year on year across a change in consent handling, you are comparing two different measurements.

Second, how resilient is collection? Browser tracking prevention and ad blockers remove a meaningful share of client-side hits. server-side tagging and a first-party data approach recover some of that, at the price of more engineering. Ask each vendor what they support and what it costs to run.

4. Decide who will use it

A tool nobody opens is the most expensive option at any price.

• If the audience is a founder and a marketer, choose the simplest product that answers question one. Depth you do not use is just navigation you have to learn.
• If you have analysts, they will want raw data. Check that you can export event-level data to your data warehouse without paying for the top tier, and how delayed it is.
• If you work with agencies, remember that they know the market leader and will bill you to learn anything else.

Definitions differ between tools more than people expect. One product's engagement rate is not another's inverse bounce rate, and a session does not end at the same moment everywhere. Whoever owns reporting should read the definitions page before the contract is signed.

What it really costs

Licence price is the smallest part. Budget for:

1. Implementation — tagging, a data layer, conversion definitions, testing. Days for a brochure site, months for a large commerce estate.
2. Running it — for self-hosted tools, servers and the person who patches them. For hosted tools, the tier you will be on in two years, not today; most are priced by traffic volume.
3. Migration — historical data almost never moves between tools. Plan to run old and new side by side for at least one full business cycle so you can explain the differences.
4. Attention — someone has to own the numbers. Without an owner, every tool decays into a dashboard nobody trusts.

A short way to decide

Shortlist two, install both on the same pages for a month, and compare them on your real questions. A trial costs less than the first meeting about a wrong choice.