If you run user acquisition for a casual mobile game, you have probably had this conversation.
The CMO asks why Meta CPI is up 18% quarter on quarter. You explain that the casual gaming auction is more crowded than ever, that interest categories are saturated, that competitors are all bidding into the same "casual gaming enthusiast" bucket. You suggest tighter targeting. You build new lookalikes. You shuffle creatives. The needle moves a few percent. Then it slides back.
The problem is not your media plan. It is the audience definition you are buying.
What "casual gaming" means inside Meta's ad system
When you select casual gaming as an interest in Meta Ads Manager, you are buying a category that was largely assembled from public Page likes, app store browsing signals, and self-reported interests stitched together over time. It includes people who installed Candy Crush in 2019 and never played it again. It includes people who clicked one casual game ad in the last six months. It includes people who follow Solitaire pages on Facebook for nostalgia and have never spent a cent in any mobile game.
You cannot see who is in that bucket. You cannot suppress the lapsed users. You cannot weight the segment toward people who actually monetise. You buy it as a black box and you trust the algorithm to find your buyers inside it.
For a 2-person UA team running $80K a month against four other publishers in the same auction, that is not a workable starting point.
Behavioural data starts somewhere different
Casual gaming on Onmo Signal is not a category derived from clicks or stated interests. It is 98 million US device IDs that have demonstrated active casual game behaviour in the last 90 days, scored against billions of in-app events from 170,000 apps. Each device carries a behavioural signature: which casual genres they play, how often, on what hardware, and whether they fall in the top spending percentile of their cohort.
That changes what you can do in the seed and exclusion layer.
You can take the top 20% of casual game spenders by recorded behaviour and use them as a Meta lookalike seed instead of your own installer file. Your install file is small, biased toward the people your old creative converted, and tells Meta to find more of those same people. A 500K seed of behaviourally scored casual spenders tells Meta a different story: find more people who look like real spenders across the casual category, not just people who looked like your last cohort of installs.
You can also exclude the lapsed segment. Most platforms cannot tell you which casual gaming users were active 18 months ago and silent since. Behavioural data can. Suppression of dead users is one of the highest-leverage moves in casual UA, and it is unavailable inside Meta's native interest tools.
What changes in your test cycle
Three things shift when your audience floor is behavioural rather than interest-based.
Day 1 CPI usually rises. This catches people off guard. Behavioural audiences are smaller, denser, and more competitive than the cheap broad interest pool, so the auction price goes up. That is the right trade. CPI is not the metric. CPA at day 30 is.
Day 30 CPA tends to fall. Because the cohort is biased toward people who actually monetise in casual games, payback curves tighten. The same creative spends slightly more per install but recovers faster. For finance, this is the conversation worth having with your CMO before you switch.
Lookalike quality stops degrading. First-party install audiences plateau because they keep teaching the algorithm to find more of the same converters you already had. A behaviourally scored seed broadens the pool the algorithm searches without losing density.
What this looks like in practice
A casual game publisher running across Meta, TikTok, and Snapchat does not need three different audience strategies. The same behavioural cohort can be uploaded as a Custom Audience to all three platforms in one click, used as a seed for lookalikes on each, and used as a suppression list against retargeting waste.
For a small UA team, that is the practical win. You stop building three different lookalike strategies, three different exclusion lists, and three different match-rate problems. You upload one cohort once, and the same audience definition runs across your channels.
Where to start
If you run casual UA in the US and you want to test this, you can browse the catalog and upload a cohort to your ad accounts in under 10 minutes. Currently free, no upload limits, no minimum spend.
Comments
Join the discussionNo comments yet. Be the first to comment.