Most analytics installations grow by addition. A tag manager gets added, then a heatmap tool, then a handful of conversion pixels, then whatever the last agency recommended, until a business is sitting on more dashboards than anyone opens. The result is not better measurement — it is a reporting interface nobody actually reads, next to a decision-making process that runs on gut feeling anyway.
Start from the questions, not the tools
Before installing anything, the useful exercise is to write down the three or four questions the business actually needs answered: which pages bring in enquiries, where visitors give up before contacting, which channel is worth the next dirham of spend, and whether a recent change made things better or worse. Everything measured should trace back to one of these. A metric that does not answer one of the real questions is not neutral — it is a line of noise competing for attention with the ones that matter.
Fewer events, tracked properly
A handful of well-defined events — a form submission, a click on a phone number, a completed booking — tell a business more than fifty vaguely-named events nobody can explain a year later. The instinct to "track everything in case it's useful someday" usually produces a dataset too messy to trust for anything, because nobody can say with confidence which events fire correctly and which quietly stopped working after the last site update.
A report that fits on one screen
If a monthly report needs a meeting to interpret, it has already failed at its job. The version that gets read is short: enquiries this month against last month, the one or two pages doing most of the work, and anything that moved sharply enough to deserve a question. Everything else can live in the underlying tool for the rare occasion someone needs to dig deeper, rather than being pushed to the front of a report every time.
Privacy settings are part of accuracy, not just compliance
How consent is configured changes what the numbers actually mean, and getting that wrong quietly skews every report built on top of it. This is a genuinely technical and legal area with rules that vary by market and change over time, so the specific configuration is worth confirming against current guidance rather than assumed from a template — but the broader point holds regardless of the specifics: analytics set up without attention to consent produces numbers that look precise and are not.
Reviewing on a schedule, not only when something feels wrong
Numbers left unchecked for months tend to be checked for the first time exactly when something has already gone wrong — a tracking code broken by a redesign, a goal that stopped firing after a form was rebuilt. A short monthly look, even five minutes against the same three questions, catches this early enough that it is a two-minute fix rather than months of unmeasured traffic.
Configuring measurement around what a business actually needs to know, rather than what a tool happens to offer, is part of the software and systems work we do alongside a build, because the point of the site described in our piece on treating a website as a business system is to generate answers someone will actually use — not a dashboard nobody opens.
