Measurement · 6 min read
SKAN, incrementality and trusting your numbers again
Why your attribution drifted, and how blending SKAN, MMM and incrementality gives you numbers you can actually defend.
Privacy changes broke the neat, deterministic attribution everyone leaned on - and a lot of teams have spent the time since arguing about whose numbers are right. The way out is not a single perfect source of truth. It is triangulation.
No single number is the truth anymore
SKAN tells you something, your MMP tells you something else, and your finance team sees the bank balance. None of them is wrong, exactly - they measure different things, with different lag and different blind spots. The mistake is treating any one of them as gospel.
Blend three lenses
- SKAN for privacy-safe, network-level signal - only as good as the conversion-value schema behind it.
- Media mix modelling for the top-down view that does not depend on user-level tracking at all.
- Incrementality tests to answer the only question that really matters: what would have happened anyway?
Design the SKAN schema on purpose
Most noisy SKAN setups are noisy because the conversion-value schema was an afterthought. Map your values to the events that actually predict a paying, retained user, and the signal sharpens immediately.
The point is calmer decisions
When three independent lenses point the same way, you act. When they disagree, you have found something worth investigating rather than an argument to win. That is what trustworthy measurement buys you - not certainty, but speed.
Let's find your next 10 million installs.
Tell us about your app and your targets, and we'll show you where the growth is - and what it would take to get there.