Meta · Published 2026-08-25 · 4 min read
Conversion tweaks will not lift performance
The classic app funnel that experts describe on streams:
Creative in META → App Store page → onboarding inside the app → purchase
They say: “Raise conversion at every step and you will get a higher conversion at the last one, the purchase. Raise conversion on the first onboarding screen and more users will buy.”
A pile of UA managers and product managers keep their jobs and avoid being fired by telling the client: “look, CTR is up! that means purchases are coming.”
🔥 Easy to grasp. Sells beautifully. Bullshit.
Curse me in the comments, but I have never met a project that lifted performance materially through small improvements.
A working creative, or a working bundle, always appeared out of nowhere, and performance rose sharply and unpredictably.
🙋 Why that is. I unpacked the mechanics in the post about micro-audiences. By raising conversion at any early step you are pushing a less loyal audience further down the funnel, and their numbers are worse than those of the audience that used to get through. The funnel has to be understood and tracked as widely as possible — from the campaign creative all the way to the purchase.
At any step of the funnel, a rise in conversion drags down purchase CPA and per-user LTV.
🎮 How I walked into this myself. I spent a year of my life understanding this simple idea. As they say: do not try this at home. Five years ago I was buying traffic for mobile games. Spend in the millions, A/B tests on 10–50 thousand users, so there was plenty of data for statistically significant conclusions. We built the funnel from the campaign click to level completion: first level, second, third and so on. From those days there is still a channel with detailed write-ups of those tests (no need to subscribe — I do not run it any more, though I still read the statistics literature).
- We raised first-level completion from 50% to 90%; it took six months of development. I was waiting for purchases to grow — the number did not move. We had pushed further down the funnel users who were not interested in the game and were about to leave anyway.
- On the same projects, at the META level, we drove creative CTR up to 5–8% and got non-paying users out of the campaign. My reasoning at the time was: “if Playrix runs misleads, they must perform, so they will work for me too.” For those outside gamedev, a mislead is a high-CTR creative that shows a different game from the one the user ends up downloading.
It turned out to be nothing of the sort. Our misleads collapsed purchase conversion. Playrix runs misleads because they have excellent games with a stratospheric LTV that lets them earn even on misleads. Something like Gardenscapes has an organic LTV of around $800; a mislead cuts it to $200, and Gardenscapes is still in profit. Those figures are hearsay, I do not have exact ones. Your app's LTV is usually in the $20–$50 range, and a luxury like high-CTR misleads is not something you can afford.
Making a high-CTR creative is comparatively easy. It buys you nothing.
💳 Web funnels. Here the cost of the mistake is higher still. Pushing up conversion on the steps of a web funnel gets you less loyal users, and with them a rise in chargebacks that can end in the payment provider banning you. Anton Shakhtin (Funnel Fox) and Lucas Lovell covered this paradox brilliantly on the Paddle stream.
⚖️ Does that mean conversion is not worth working on? No. If the task is to tune performance by a few per cent, it works.
🔥 So what do you do? When the economics do not close, tuning will not fix them. You have to flip the game and change the positioning.