Roughly a third of the accounts we inherit are measuring the wrong thing entirely. Not badly, wrong. The numbers look plausible enough that nobody questions them, and every optimisation decision downstream inherits the error.
Conversions
Duplicate conversion events are the most common fault we find. A thank-you page fires a conversion, the form submit handler fires another, and a tag manager trigger fires a third. Your reported conversions can be double or triple the real number, which makes your cost per acquisition look far better than it is.
Check next that your key events are actually marked as conversions, that they are not counting on every page view, and that offline outcomes are being imported. A form fill is not revenue. If your CRM knows which enquiries closed, GA4 should know too.
Traffic quality
Internal traffic filters are frequently defined but never activated. Developer and staging traffic then sits inside every report. Referral exclusions for payment gateways are equally often missed, which breaks session attribution at exactly the moment revenue is recorded.
Cross-domain and consent
If your checkout, booking engine or CRM forms live on another domain, cross-domain measurement has to be configured explicitly. Without it, a returning buyer is counted as a brand new direct session at the point of purchase, and every channel that actually earned the sale loses its credit.
Consent mode is the newest source of quiet loss. Configured incorrectly it either blocks measurement that users did consent to, or models data you never had permission to collect. Both are worth checking before you trust a single report.
What to do with the list
Work through it in order of revenue impact rather than effort. Fix conversions first, because everything else is judged against them. Then traffic quality, then attribution. Most brands recover twenty to thirty five percent of previously lost signal without spending another rupee on media.