Steadier, not just cheaper: scaling Cleanup on Snapchat
Written by Codeway
Listen to this article
3:30
Introduction
A cheap install is easy to buy once. We moved Cleanup onto Snapchat’s unified attribution to find out whether a cheap install is something you can buy repeatedly.
Every growth team has bought a cheap install. The harder question is whether you can buy the next hundred thousand at the same price — because a channel that gets more expensive the moment you lean on it is not a channel, it is a coupon.
Cleanup is our phone storage app: it finds the duplicates, the screenshots and the forgotten video files that quietly fill a device. It is a broad product with a broad audience, which makes it a good test bed and a demanding one. The people who need it are not a niche you can target your way to, so performance depends on the platform reading intent better than we can describe it.
The measurement problem underneath the media problem
On iOS especially, the gap between what happens and what a channel can prove has been widening for years. Every platform sees a slice of the truth. Snapchat sees its own first-party signal; our mobile measurement partner sees a modelled, cross-network view. Optimise against either one alone and the bidding is working from a partial picture.
Unified Attribution combines the two — Snap’s first-party attribution alongside modelled signals from the measurement partner — so campaigns optimise against a fuller account of what an install was worth. We ran it across several countries and both operating systems rather than in one clean market, because a result that only survives in one geography is not a finding.
What changed
- lower cost per install
- 18%
- more scale at a maintained cost per purchase
- 1.8×
- months of continuous measurement
- 4.5
The cost per install fell 18%. The number that actually changed our behaviour was the second one: we pushed 1.8× the volume through the channel without the cost per purchase drifting. Cost per install tells you what you paid at the door. Cost per purchase tells you whether the people who came through it were the people you wanted, and holding it flat while the spend climbs is the rare part.
“Unified campaigns changed how we scale on Snapchat. Across multiple geos and operating systems, we’ve built performance that’s steadier and more sustainable.”
Why steadiness is the metric
A volatile channel gets managed defensively. You underspend it, you cap it, you check it daily, and you never find its ceiling. A predictable one gets a bigger budget, because the forecast holds. That is the real return on better attribution: not a lower number in a dashboard, but a channel the finance conversation can plan around.
It also changes what a test is for. When the measurement is noisy, most experiments are unreadable and you learn nothing but the noise floor. When it settles, the creative and audience work underneath it starts producing signal you can act on — which is where the compounding actually comes from.
