Build the tracking before you spend the budget
Most underperforming accounts are not a bidding problem or a creative problem. They are a measurement problem that was never fixed at the start.
When we audit an account that is not working, the diagnosis is boring more often than it is interesting. The targeting is usually fine. The creative is usually fine. What is broken is the thing nobody wanted to spend week one on.
The pattern
A campaign launches on a Monday. Conversion tracking is “mostly set up.” Two events fire on the same page. One of them counts a scroll. The other counts a thank-you page that also gets hit on refresh.
Six weeks later the platform has optimized enthusiastically toward a signal that was never real. The algorithm did exactly what it was told. It was told the wrong thing.
What good looks like before launch
One conversion action per meaningful outcome. Not seven. Not a scroll depth. If a stakeholder cannot explain in one sentence what a conversion action represents, it should not be a conversion action.
Server-side where it matters. Browser-only tagging loses a meaningful share of events to blockers and privacy settings, and the loss is not evenly distributed across audiences, which quietly biases your optimization.
Values attached, not just counts. A conversion worth $80 and one worth $8,000 should not be the same row. Passing value lets the platform bid toward revenue rather than volume.
A test conversion you triggered yourself. Walk the funnel personally. Fill in the form. Confirm the event lands in the platform, in GA4, and in the CRM. If the three disagree, find out why before spending.
The cost of skipping it
Two weeks of setup feels expensive when a client wants to be live. Six weeks of budget optimizing toward a phantom event is more expensive, and you only find out after the money is gone.
We do not launch without it. It has cost us a couple of deals. It has never cost us a client.