Every few months a founder sends us a spreadsheet showing a return on ad spend of 4× and a bank balance that disagrees. The campaigns are not lying, exactly. They are answering a different question from the one being asked.
Before you move another shilling of budget, it is worth knowing which question your numbers are actually answering.
The three numbers that disagree
Pull up any growth meeting and you will usually find three versions of the same metric in the room:
- Platform-reported revenue — what Meta and Google each claim they drove. These overlap, and both platforms count the same sale.
- Analytics revenue — last-click, usually under-crediting anything that happens early in the journey.
- Actual revenue — what finance recognises, net of refunds, discounts and the cost of goods.
The gap between the first and the third is where budget decisions go wrong. We have seen accounts where platform-reported revenue was 2.6× the real figure, purely from deduplication failures across channels.
Fix things in this order
1. Get server-side events working
Browser-side tracking loses somewhere between 15% and 40% of events to ad blockers, Safari's tracking prevention and plain network flakiness. Server-side conversion APIs recover most of that, and give you an event stream you control.
The important part is the deduplication key. Send the same event ID from browser and server so platforms can collapse duplicates rather than counting the sale twice.
2. Agree one definition per metric
Write down what a conversion is. Not "a purchase" — is a purchase counted at checkout or at payment capture? Does a refund reverse it? Does a subscription count once or monthly?
This sounds bureaucratic until the first meeting where marketing and finance stop arguing about whose number is right.
3. Report on contribution margin, not revenue
Revenue-based return on ad spend hides the fact that your product mix has wildly different margins. A campaign driving 5× on a 20%-margin product is losing money. A campaign driving 2× on a 70%-margin product is printing it.
Once you switch the dashboard to contribution margin, the ranking of your campaigns changes. Usually dramatically, and usually uncomfortably.
4. Then, and only then, scale spend
With a trustworthy signal, scaling is mostly mechanical: increase budget on what clears your payback threshold, cut what does not, and keep creative volume high enough that fatigue never becomes the bottleneck.
What this costs you
Two to four weeks, typically, and a quarter of uncomfortable conversations while everyone adjusts to smaller-looking numbers. Set against the alternative — scaling spend against a broken instrument — it is the cheapest work in the entire growth programme.