TLDR: A “10 common GA4 mistakes” checklist can only find the problems every site shares. The errors that actually break your reporting are specific to your build, your platform and what your site is for, and no checklist author can see them. A real audit is a layered, repeatable process, not an article you read: universal hygiene, then platform, then intention, with every finding tied back to a business question. Some of it a tool can settle. The judgement stays human.
Click to discuss this article with your favourite AI
Search “GA4 audit” and you will find a hundred checklists. Ten common mistakes, twenty things to fix, a tidy list you can run down in an afternoon. They are not worthless. But they can only ever contain the problems that are the same on every site, because that is the only kind of problem a stranger writing an article could know you have. Duplicate base tags, missing enhanced measurement, a conversion marked that should not be. Real, worth fixing and shallow. They get sorted in a morning.
The errors that quietly break your reporting are not on any list, because they depend on things no article could know: how your site is built, which platform runs it, what it is actually for. Reading about someone else’s mistakes will never surface yours. Running a proper audit will, and a proper audit is not a checklist. It is a layered process.
Three layers, and only the top one is about your questions
Think of the audit in three layers, from the most universal to the most specific to you.
The bottom layer is hygiene, and it is the same on every ecommerce site on earth. Is the user count deduplicated. Is internal and staff traffic filtered out. Is attribution set sensibly. Nobody ever asks for these. No stakeholder has said “please confirm our users are not double counted.” But this layer holds up everything above it, because if the user count is wrong then every rate built on it, every conversion rate, every per-user figure, is wrong too. On one site I audited the reporting counted 2.28 million users where the deduplicated figure was 1.77 million, close to a third too high. Not one person had asked me to check that. It was quietly bending every ratio on the dashboard.
The middle layer is your platform. Every ecommerce platform, and every analytics extension bolted onto it, has its own known behaviours. On one build the platform’s own GA4 extension and the tag manager were both firing the purchase event, so a share of orders were counted twice. You only think to check that if you know what platform you are dealing with and how its plugins behave. A checklist cannot, because it does not know your stack.
The top layer is intention, what your site is for. An ecommerce site lives and dies on a specific set of events: add to cart from every surface, a clean purchase, working promotion tracking on the merchandising. On that same site the events behind the homepage carousels were dead in the tag manager, so the business could not see whether any of its banners drove a sale. A content site would not care about that at all. It would want reading depth, scroll and subscriptions. The intention layer is where your business questions finally shape what gets checked.
Notice the order. The questions only touch the top layer. This matters, because it is tempting to say the audit should just check whatever the stakeholders asked about. It should not. The hygiene layer runs regardless of what anyone asked, because trustworthy numbers are not optional and nobody thinks to request them. What the questions decide is not whether a check runs, but what matters most once it does, and what you fix first. Questions are the lens on the audit, not the gate to it.
Some of it a machine can settle, the rest is judgement
Here is the honest part most audit pitches skip.
A good chunk of an audit can be automated. Whether the purchase event fires, whether users are deduplicated, whether attribution is set to data-driven, whether the custom dimension your reporting relies on actually exists, these are facts sitting in the API, and I run a tool that checks 69 of them in a single pass. If that were the whole job, you would not need me. You would need a script.
But the checks that decide whether you can actually trust a number are not like that. Does the purchase event fire only when someone really buys, or also every time they refresh the thank-you page or hit back and land on it again. Does the add-to-cart number include every surface from the map, or just the product page. Does this metric, set up the way it is, answer the question the business asked, or a subtly different one. None of that is visible in the API. It needs someone in the site with the console open, following the event the whole way through, thinking about what the number will be used for. The tool clears the mechanical checks so the human time goes where the judgement is. That split is the whole point, and pretending the tool does the hard part would be a lie.
What you get is not a list of red marks
Run all of this and the output is not 69 ticks and crosses. It is a short, ranked picture of what you can trust and what you cannot, with every problem attached to the business question it threatens and a severity that says how much to care. The doubled purchases are critical, because they inflate revenue, the number the whole business steers by. The dead carousel tracking is high, because it blocks a real merchandising decision. A cosmetic naming inconsistency is low. You finish knowing exactly which of your numbers are safe to build on, which are not and why.
That is the difference between running a checklist and running an audit. One hands you a tidy list of generic fixes. The other tells you, specifically, which of your own numbers you are allowed to believe.
And the moment you know that, you can do something you could not before: write down each metric with an honest note of what it can and cannot tell you yet. That is the next step.
Ready to start trusting your data?
If you have a dashboard you are not quite sure you can trust, an audit is how you find out which parts you can. I run layered GA4 and GTM audits for ecommerce businesses and hand back a ranked picture of what is safe to build on.