Most underperforming ad accounts we audit do not have a targeting problem or a creative problem. They have a measurement problem that has been quietly corrupting every decision made in the account for months. Here is the checklist we run before we let anyone increase spend.
1. Is the conversion event firing at the right moment?
The single most common fault is a conversion that fires on page load rather than on the action itself. A “Lead” event attached to the contact page instead of the form submission will report a conversion for every visitor who arrives and leaves. The platform then optimises hard toward cheap visits, and your reported cost per lead looks excellent while your inbox stays empty.
Test it yourself. Load the page without converting and confirm nothing fires. Then convert and confirm exactly one event fires.
2. Is it firing exactly once?
Duplicate events are common when a tag has been installed twice – once hard-coded in the theme and once through a tag manager or a plugin, which is easy to do when a site has changed hands or agencies. Symptoms include suspiciously round conversion counts, or platform-reported conversions running at roughly double what your CRM shows.
If you are sending events both from the browser and from the server, deduplication needs to be configured properly with a shared event ID, or you will count every conversion twice.
3. Do platform numbers reconcile with your source of truth?
Pick a single week. Count the leads or orders in your CRM, your inbox or your backend. Compare that with what the ad platforms report for the same window. A gap of a few percent is normal. A gap of thirty percent or more means something is wrong, and until you know which direction the error runs, you cannot tell whether you are scaling a winner or a rounding error.
Remember also that Meta and Google will each claim credit for the same conversion. Adding up platform-reported conversions across channels will always overstate reality.
4. Are you measuring the right event?
Optimising toward form fills produces form fills. If half of them are junk, the algorithm will happily find you more of the same people, because it is doing precisely what you asked. Where your sales cycle allows it, pass a downstream event back – qualified lead, booked call, or actual sale – and optimise toward that instead. It is slower to accumulate data but it points the system at revenue rather than at activity.
5. Does tracking survive consent and ad blockers?
Browser-side tracking loses a meaningful share of events to blockers, privacy settings and consent banners. That loss is not evenly distributed – it skews by device, browser and audience – so it distorts comparisons between segments, not just absolute totals. Server-side event forwarding recovers part of this, and it should be set up with consent handled correctly rather than bypassed.
6. Is the attribution window documented?
Two people comparing the “same” campaign on different attribution settings will reach opposite conclusions and both will be convinced they are right. Write down the window and model you report on, keep it constant, and note it on every report. Changing attribution mid-quarter and then celebrating the improvement is a good way to fool yourself.
7. Can you see it break?
Tracking breaks silently. A theme update, a plugin change, a new checkout flow or a consent tool update can remove a tag without anyone noticing until a month of data is gone. Set an alert on a sudden drop in conversion volume, and re-test the full path after every significant site change.
The point of all this
None of this is glamorous, and none of it will show up in a case study. But every optimisation decision, every budget increase and every “this campaign is working” conclusion rests on these numbers being right. Getting them right is usually the cheapest performance improvement available to a business already spending on ads.
If you would like us to run this check on your account, see our services or reach us through the contact page.