Our sales fell 73% when we switched to Meta’s Incremental Attribution. And that’s okay.
Incremental Attribution is Meta’s attempt to measure the conversions that truly wouldn’t have happened without an ad. Unlike the standard 7-day click / 1-day view model, which credits every conversion after an ad interaction, Incremental Attribution tries to approximate the impact of an ad in a more “real-world” sense. Meta uses machine learning, historical Lift studies, and probabilistic modelling to filter out conversions that likely would have happened anyway. In theory, it’s a smarter way to see the true value of our spend.
The principle is sound. Every marketer wants incremental results, sales that wouldn’t have existed without an ad. So we tested it.
We ran an A/B test: standard attribution vs incremental attribution. Then we scrutinised the results across Meta, GA4, and Fospha. You have to assess incremental attribution across multiple sources, as the in-platform data will always be the same - fewer sales.
In Meta:
The sales drop was expected, but engagement metrics suggested we were reaching a higher-quality audience.
In Google Analytics:
This mirrored Meta’s signals: fewer attributed sales, but more meaningful interactions.
In Fospha:
And here is where the picture became less encouraging. While Meta and GA4 hinted at potential, Fospha’s results cast serious doubt.
For now, Incremental Attribution is a no-go for us. Smaller accounts can’t absorb a 73% sales drop, and larger accounts need a retest before moving forward. The concept is compelling. The execution, however, is still rough around the edges.
We’ll revisit it when the data becomes more consistent. Until then, this experiment serves as a reminder: new attribution models are interesting, but their impact can be dramatic, and sometimes costly, in practice.
