The short answer
- A wrong measurement is more dangerous than no measurement: you decide with conviction on noise.
- The five checks: Consent Mode v2, server-side measurement, full conversion return, a fitting attribution model and own key events.
- Last-click attribution punishes channels that work early in the customer journey; always compare two views.
- Measure your request route step by step; standard events do not tell that story.
Are you measuring this yet?
| What | Why it counts | How to check |
|---|---|---|
| Consent Mode v2 | Without correct consent signals you measure too little and not cleanly | Check in GA4 whether consent signals come in after a consent choice |
| Server-side measurement | Browser measurement leaks through blockers and limits | Check whether your tags run through an own server container |
| Conversions returned in full | Ad algorithms learn on what you send back | Compare platform conversions with your own records |
| Attribution setting | Last-click punishes channels that work early in the journey | View your attribution model and compare channels in both views |
| Own key events | Standard events say little about your funnel | Walk your request route: is every step measured as an event? |
Why correct measurement comes before everything
Every marketing decision rests on the measurement beneath it: which channel gets budget, which page gets rebuilt, which campaign stops. If the measurement is off, all those decisions are guesswork with a dashboard as an alibi.
The creeping mistakes are the most dangerous. A double tag counts everything twice. A consent setting that lets half the data through skews per channel differently. And a forgotten event makes your best-performing page invisible.
So walk the table above first. Five checks, not dozens; the rest is refinement.
A dashboard gives confidence. Only correct measurement gives you a case.
The two big distortions
Distortion one: the leak. Browsers and blockers let measurement leak away, and without a server-side setup and full return your ad platforms learn on a shrunken reality. How that leak closes is in our piece on cookieless measurement.
Distortion two: attribution. By default the last click often gets all the credit, so channels that work early in the journey structurally get too little credit and are unfairly cut.
The remedy is simple discipline: view channels in two attribution views side by side before you shift budget. If the conclusion changes per view, you know the model is steering your judgement.
Own events: measure your funnel
GA4 measures mainly generic behaviour by default. Your money is earned on a specific route: from landing page via form to request. Every step in that route should be an own event, otherwise you never see where it goes wrong.
So set up the key events around your request route and mark the most important as a conversion. Our own site measures exactly like that, and only after consent; neat and useful do not exclude each other.
The new AI insights layer in GA4 can flag anomalies and answer questions in plain language. Handy, but it mainly amplifies what lies beneath: feed your AI noise and you get confident noise back.
Frequently asked questions
How do I know if my GA4 double-counts?
Compare your measured requests with your own inbox or records over the same period. Structurally too-high figures point to double tags or events; a measurement audit finds the source.
Do I have to be able to do everything in GA4 myself?
No. You have to understand the five checks and be able to read the monthly figures; the setup is one-time specialist work. After that the report should speak in plain language.
Is GA4 allowed under the privacy rules?
With Consent Mode v2, measurement after consent and a correct setup it can be done neatly. That is how it is set up on our own site too.
Do you dare trust your dashboard?
Let us run the five checks on your measurement. You hear per point what is in place, what is missing and what that skews.