In app campaigns, structure isn't there to target — the system handles that — it is there to separate goals and budgets that must not mix, and to avoid splitting the volume the algorithm needs. With that idea, the decisions are few and clear.
The three mandatory separations
- By operating system. Android and iOS always in separate campaigns: they measure differently (lesson 3), cost differently and learn at different speeds.
- By goal. Installs, in-app actions, value, re-engagement… one goal per campaign. Mixing them asks the system to optimise two things at once, and it will do both by halves.
- By market, when the market really is different. Not by country for its own sake: split when language, purchasing power, competition or monetisation change. Similar countries can share a campaign and add up the volume neither would have alone.
How many campaigns: the volume criterion
The right question is not "how many do I want" but "how many can I feed". Each campaign needs enough events of its chosen goal to exit learning and stay stable. As a public reference, a daily budget of around 50 times the target cost per install (install campaigns) or 10 times the target CPA (in-app actions) is recommended. If your total budget doesn't cover two campaigns at those minimums, build one.
The symptom of having atomised: several campaigns spending little, all with erratic CPA, none consolidating. The cure is consolidation, not more splitting.
Asset groups: themes, not scatter
Inside the campaign, asset groups bundle creatives. The criterion is one theme per group: one angle, one proposition, one creative type. That way the system can learn which message works without mixing apples and oranges, and you can read the asset reports meaningfully (lesson 6).
Two or three well-stocked groups outperform eight starving ones: the logic is the same as with campaigns.
A deployment order that works
- One install campaign per operating system, in your main market, bidding on installs while there is no data.
- Once there is in-app event volume, move the goal up in that same campaign or open a dedicated one for actions (lesson 5).
- With an installed base and deep links ready, add engagement (ACe) to win users back.
- Then new markets: one at a time, each with its own time and budget.
💡 Ninja trick: treat each campaign like an "account" from the Smart Bidding module: it needs its daily minimum volume for the conservative engine to have patterns. And name a leader — the highest-volume campaign — because conversion history also accumulates at account level and pulls the others along.
⚠️ Pitfall: replicating your search campaign structure (one per theme, one per product). In apps that only splits the learning across campaigns that never get to know anything.
What you should remember
- Always separate by operating system and by goal.
- Split markets only when they are genuinely different.
- Fewer, better-fed campaigns: check the budget minimums.
- Asset groups by theme, not by whim.