The models, and what each one favors
Last click gives all the credit to the last touchpoint before the conversion. Simple and stable, it favors the channels that close (branded search, retargeting, email) at the expense of the ones that sparked the interest.
Rule-based models (first click, linear, time decay, position-based) shift the credit according to a rule chosen in advance. The data-driven model splits the credit by comparing the journeys that convert with the ones that do not.
According to Google's documentation, Google Ads now offers only two models: last click and data-driven, which has become the default for most conversions. The rule-based models were retired in 2023, in Google Ads as well as in Google Analytics 4.
Why ad platform figures never add up
Each platform attributes with its own rules, on its own data: it only sees the touchpoints it served itself. A sale preceded by a Meta ad and then a click on a Google ad will be claimed by both, each in good faith.
Attribution windows widen the gap: a platform can count a conversion that happened several days after a click, and sometimes after a mere impression, with no click. The longer the window, the more sales it credits to itself, and the further the platforms' total exceeds actual sales.
Then there is signal loss: declined consent, browser blocking, journeys across several devices. Some conversions are seen by no tool at all, and the platforms estimate part of them through modeling.
Attributing is not proving
Attribution tells you which ads came before the sale. It does not tell you which sales would not have happened without advertising, the only question that matters when deciding on a budget. Retargeting and brand campaigns shine in attribution because they reach people who had already made up their minds.
Only incrementality answers that second question: you compare an exposed group with a group that sees no ads, split by region or at random, and the difference in sales measures the real effect.
A sound setup combines three levels: attribution for day-to-day steering, incrementality tests to recalibrate channels at regular intervals, and an overall indicator such as MER to check that the whole holds up. None of them is enough on its own.
Frequently asked questions
Which attribution model should you choose?
In Google Ads, the data-driven model, which splits the credit based on actual journeys. Last click remains useful as a stable reference, as long as you know that it undervalues discovery channels.
Why do the platforms' conversions add up to more than my sales?
Because each platform credits itself with the sales it preceded, over its own window, without seeing the others. The same sale is claimed several times. Only your back office gives the real total.
What is an attribution window?
The period during which a conversion can be linked to an ad after a click or an impression. It varies across platforms and settings; comparing two platforms without aligning their windows amounts to comparing two counting rules.
What is the difference between attribution and incrementality?
Attribution splits sales among the ads that preceded them. Incrementality measures the sales that would not have happened without advertising, by comparing an exposed group with a control group. The first describes; the second proves.