Return on ad spend is a useful number that is wrong more often than people realise. Before cutting a campaign that reports badly, it is worth knowing which of these you are looking at, because three of the four are measurement problems rather than performance problems.
1. The attribution window is shorter than your sales cycle
If it takes a customer three weeks to decide and your platform attributes over seven days, the sales are real and the credit goes missing. This hits considered purchases hardest — anything expensive, anything B2B. Compare the window against how long your customers actually take, which your own records will tell you.
2. You are counting revenue, not margin
A 4x return on a product with a 20% margin loses money. A 2x return on something with 70% margin is excellent. Platforms report revenue because that is what they can see. Until you feed margin into the decision, you are optimising toward whichever products are cheapest to sell rather than most profitable.
3. Every channel is claiming the same sale
Add up the conversions each platform reports and compare it against orders in your actual system. If the platform total is higher — and it usually is — several channels are claiming the same customer. The fix is not picking a winner, it is trusting your own order data as the source of truth and treating platform numbers as directional.
4. It genuinely is bad
Sometimes it is. The tell is consistency: poor return across every product, audience and placement, holding steady over time. A real performance problem is broad and stable. A measurement problem is patchy, and looks different depending on where you stand.
The check worth doing first
Pick a month. Take total revenue from your own system, and total ad spend from the platforms. Divide. That blended number is cruder than any dashboard and much harder to fool. If it looks healthy while individual campaigns look terrible, you have a measurement problem, and you were about to cut something that works.