Performance marketing has a reporting bias built into it. Ad platforms report revenue and ROAS because those are the numbers they can see. They cannot see your cost of goods, your fulfilment cost, your payment processing fees, your refund rate or your support load. So they optimise toward a number that is only loosely related to whether you are making money.
The gap widens exactly where it hurts most: at scale. Early campaigns often run on your best audiences and highest-margin products, so revenue and profit move together. As budget increases, the marginal customer is more expensive to acquire and frequently buys the discounted, lower-margin item. Revenue keeps climbing. Contribution margin quietly inverts.
The fix is unglamorous. Feed real margin data back into the optimisation loop — either through value-based bidding with true contribution values, or by managing to a blended contribution target rather than a channel-level ROAS. Both require finance and marketing to agree on a single number, which is usually the actual obstacle.
Second-order costs deserve the same treatment. A campaign that drives high refund rates or heavy support volume carries a cost that never appears in the ad dashboard. Segmenting refund and support rates by acquisition source frequently reveals that the cheapest traffic is the most expensive customer.
None of this argues for spending less. It argues for knowing, at the point of decision, which incremental rupee of spend is productive — because the alternative is discovering it a quarter later in a P&L review.