GA4 for a small business: the three reports that matter

Opening GA4 for the first time with a small website is a disorienting experience, and the disorientation is not a failure of understanding. The tool is built to serve organisations with many properties, large teams, app and web data combined, custom event schemas, advertising integrations and analysts whose job it is to use it. A business with one site, a few hundred visitors a week and one person looking after everything needs a small fraction of that surface, and almost none of the interface signals which fraction.

The practical response is not to learn the tool comprehensively. It is to decide which questions you actually need answered and find the smallest set of reports that answers them. For most small businesses there are three questions: where are visitors coming from, what do they do once they arrive, and does anything worthwhile happen as a result. Three reports cover those, and the rest of the interface can be left alone without cost until a specific question makes one of them necessary.

The acquisition reports group visitors by how they arrived — organic search, direct, referral from another site, social, email, paid campaigns. This is the report that tells you which of your efforts are bringing people at all, and it is the one most worth checking regularly, because it is the closest thing to a scoreboard for everything you do outside the site.

Two details prevent misreading it. First, direct traffic is partly a residual category: it collects visitors whose source could not be determined, including some clicked links in messaging apps and documents, so a large direct figure is not straightforwardly a measure of people who typed your address. Second, the channel groupings depend on links being tagged properly. Traffic from your own email or messaging campaigns lands in whichever bucket the link's parameters imply, and untagged links are frequently misfiled — which means the report is partly measuring your link hygiene rather than reality. Tagging campaign links consistently is the single highest-value habit for making this report trustworthy, and it costs nothing but consistency.

The engagement reports break traffic down by page and by event, answering which pages hold attention and which are passed through. For a small site the pages-and-screens view is the useful one: it ranks pages by views and, more informatively, by average engagement time, which tells you where people actually stay. A page with many views and very short engagement time is usually either mislabelled in search results or failing to deliver what the visitor expected, and both are fixable.

The concept worth understanding here is that GA4 records everything as events — a page view, a scroll, a click, a form submission are all events with parameters. This is why the tool feels abstract compared with older page-centric analytics, and it is also what makes it flexible. For a small business the payoff is narrow but real: you can mark the specific actions that matter to you and then count them. Resist the urge to instrument everything, because every additional event is something to maintain and to interpret, and a site with forty custom events and no clarity about which three matter is harder to read than one with three.

The third report covers what GA4 calls key events. The naming here has changed and the change causes real confusion, so it is worth stating precisely: Google renamed conversions to key events, describing them as the events that measure actions important to your business's success. The word conversion now refers specifically to a key event imported into Google Ads for campaign measurement. Two products, two meanings, one word — which is exactly why figures quoted from Ads and from Analytics so often fail to agree.

For a small business the setup is short. Decide the two or three actions that constitute success on your site — a form submitted, a call button pressed, a purchase completed, a quote requested — make sure each fires as an event, and mark those events as key events in the admin settings. Then the report counts them and lets you see them broken down by acquisition channel, which is the combination that answers the only question that really matters: not which channel sends the most visitors, but which sends visitors who do something. A high-traffic channel producing no key events is a channel bringing the wrong people, and no amount of traffic reporting on its own would tell you that.

Once those three reports are familiar, the efficient move is to build a single overview containing only what you look at, so the weekly check is one page rather than a tour. GA4's exploration and library features let you assemble a custom view; a handful of cards — visitors by channel, top pages by engagement time, key events by channel, all over a consistent period — replaces most navigation. Keep it deliberately short, for the same reason any report should be short: a comprehensive view requires the reader to decide what matters before learning anything.

What to leave alone is most of the rest, and leaving it alone is a legitimate decision rather than a gap in your knowledge. Audience building for advertising, attribution model comparison, cohort explorations, predictive audiences, cross-platform app reporting, custom dimension architecture: each exists for a real need, and none of those needs is likely to be yours yet. Adopting a feature before you have the question it answers reliably produces configuration you cannot interpret and will not maintain. The better sequence is to wait for a specific question you cannot answer, then find the feature that answers it.

It cannot tell you why anyone did anything. It records that a visitor arrived from search, read two pages and left, and it holds no information about what they were looking for or whether they found it. Intent, satisfaction and the reason for abandonment are outside what the tool measures, and the gap is usually filled by inventing a story around the numbers — which is where analytics does the most harm, because the story feels evidence-based. Asking customers remains the only route to that information.

It is also an incomplete record of who visited. Consent banners, ad blockers, privacy-focused browsers and cross-device journeys all remove or fragment data, and the size of what is missing cannot be measured from inside the tool. Attribution is a further layer of estimate rather than measurement: assigning a sale to a channel requires a model, and every model is an assumption about which touchpoint deserves credit. Treat these figures as directional evidence, reconcile the ones that matter against orders and payments you can count independently, and where the two disagree, trust the money.