Paid media judged by what it returns

The most expensive failure in paid media is not a poor campaign; it is a good campaign optimised against a broken signal. A duplicated conversion event teaches the platform to buy the wrong traffic, and the account can look successful the entire time it wastes money.

We test the measurement path before launch: submit a real form, follow the event to the platform and the analytics property, confirm it appears once, and check that consent handling has not silently removed a segment of it. Only then does budget go live.

Every advertising platform reports on its own performance using its own attribution window, and each will claim the same conversion. Adding up the dashboards routinely produces more revenue than the business actually received, which makes cross-channel budget decisions impossible.

We treat platform figures as one input and reconcile them against analytics and back-end records. Where volume allows, incrementality is tested by holding out a region or audience rather than inferred from a model, because a channel that reports well can still be buying customers who would have arrived anyway.

Paid accounts decay. Search terms drift, competitors change bids, creative fatigues, and automated bidding responds to seasonal noise as if it were a trend. An account left alone for a quarter reliably spends more per outcome than it did at launch.

Management means a regular cycle of query review, negative keyword maintenance, placement exclusions, creative rotation and pacing checks, with changes recorded so their effect can be read later. We would rather return unspent budget than let it run into segments that have stopped performing.