Last click lies

The default model in ad platforms is last click: the final touch before a conversion takes all the credit. It spread because it is manageable, not because it is accurate. Reporting a user who searched for the brand and clicked as 'won by search advertising' makes every touchpoint that taught them the brand invisible.

The problem is structural rather than a flaw in the tool. A purchase decision is not formed in a single touch; impressions, memory, recommendation and search form a chain. A model that counts only the last link pushes the chain itself out of the budget. In practice that means awareness campaigns look permanently inefficient and get cut.

The attribution window is part of the decision too. A common default on Meta is seven-day click and one-day view; the narrower the window the fewer conversions a campaign claims, the wider the more. The same campaign can show two different performances purely because the window changed. When reading a report, knowing the window matters as much as the number.

A sounder method is incrementality: comparing a group that saw the ads with one that did not. A full experiment can be hard on small budgets; geographic splits, or simply pausing a campaign for a period to see the baseline, will do. The measurement will be imperfect — just less imperfect than last click.

One habit in reporting prevents most of the errors: always place total sales and a period-on-period comparison next to the conversion count. The ad platform overstates its own share; revenue does not.