two & a half gamers -🎯 How to Do a UA Review Properly by Matej Lancaric

Summary

“The CPI is going up, so change the creatives” is not a UA review — it’s just narrating the dashboard. A real review has next steps, digs under the naming conventions, and usually finds money you’re already wasting without spending a cent more. This is how to do one properly.

FULL ARTICLE: https://lancaric.substack.com/p/ua-re…

Matej Lančarič goes solo for a UA-audit masterclass, built from his popular Substack piece and 12+ years of spending on and reviewing user acquisition. He walks the five layers he audits every time — the strategy behind the spend, the campaigns, the creatives, the MMP/dashboards, and the people making the decisions — and the order that matters (data and raw conclusions first, context and conversations second, so nobody colors your read before you’ve formed it). Along the way: why you must strip out fraud before you trust anything (one game looked like five healthy channels until fraud removal revealed one channel eating 70% of spend), why you should never trust a campaign’s naming convention (build the spend-by-campaign-by-geo pivot and look underneath), how splitting a “worldwide” campaign by geo and OS surfaces small high-ROAS pockets (his own US spend was dragging the whole campaign down until he split it four ways), the cohort discipline most teams skip (fully-baked D7/D28/D60, apples to apples), the danger of deciding on blended organic-plus-paid numbers, and the LTV-curve and payback questions to ask before you touch anything. The punchline: most reviews don’t call for more budget — just reallocation — and you should run one every month.

📌 KEY TAKEAWAYS

— There’s a review, and there’s a review. Pulling 90 days of data, glancing at the MMP, and saying “CPI is up, you’ve got creative fatigue, change the creatives” isn’t a review — it’s narrating the dashboard. A real review produces next steps and digs several layers under the surface.

— Audit five layers, in order: the strategy behind the spend, the campaigns, the creatives, the MMP/dashboards, and the people making the decisions. Knowing who owns UA and who makes which calls is as important as the data.

— Look at the data before you talk to people — deliberately. If you gather context first, people’s framing colors your read. Matej wants raw data and raw conclusions first, then asks for context to catch anything he overlooked. Order protects objectivity.

— Never trust the campaign naming convention. Teams silently add geos into campaigns without updating the name. Build a spend-by-campaign-by-geo pivot table and look at what’s actually inside each campaign, not what it’s called. Go one layer deeper, always.

— Splitting “worldwide” by geo and OS surfaces hidden pockets. On his own game, the US was dragging the whole campaign down — so he split it into four campaigns. “Best practice” says run one worldwide campaign, but best practice has exceptions. Bucket by geo using your own LTVs (not someone else’s), and remember geo performance on iOS doesn’t predict Android — audit by OS too. Tier one isn’t only UK/Canada/Germany/Australia; small high-ROAS pockets (Norway, etc.) hide in the data.

— Cohort discipline is non-negotiable, and blended numbers lie. Compare apples to apples with fully-baked cohorts (real D7/D28/D60, not half-baked windows). Some teams don’t even know what a cohort is — a serious red flag. And beware decisions made on blended organic-plus-paid numbers: one audit had a team cut budget based on blended figures, which painted a totally false picture. A day-0-to-day-7 multiplier collapsing (e.g. 12.3x to far lower) should be flagged by the team — and if it isn’t, the reviewer flags it.