App marketing budget
Most budget documents allocate by channel. Allocating by what each line teaches you produces better decisions, particularly when the budget is small.
9 August 2026 ยท PAD team
Allocate by learning, not by channel
Early on, money buys information rather than growth. A budget that produces ten thousand installs and no idea why is worse than one that produces two thousand and a reliable cost per retained user.
| Line | Small budget | Scaling budget | Purpose |
|---|---|---|---|
| Paid tests | 50% | 60% | find what works |
| Creative production | 25% | 20% | the input that most changes cost |
| Creators | 10% | 10% | reach you cannot buy directly |
| Tools and measurement | 10% | 5% | makes the other lines legible |
| Reserve | 5% | 5% | for the thing that works unexpectedly |
Creative is not a cost, it is the lever
Cost per install responds to creative more than to bidding, targeting or budget size. A team spending everything on media and nothing on production runs out of working ads within weeks and then blames the auction. Plan for a steady rate of new concepts rather than occasional refreshes.
The line that gets cut first
Measurement. It looks like overhead, produces no installs of its own, and is the reason every other line can be judged. Cutting it does not save money, it removes your ability to know which of the remaining lines to cut.
Delivery belongs in this category and costs almost nothing. Measured on 21,000 taps routed between 3 and 9 August 2026, 19,381 came from inside social apps and 95.6% were iOS. If a share of that never arrives, every other line in the budget is being evaluated on partly fictional numbers.
How much to spend in total
There is no ratio that survives contact with a specific app. The usable rule is different: spend enough in one channel to reach several hundred installs, because below that the numbers swing too widely to act on. If the budget cannot reach that in one channel, it certainly cannot reach it in three.
When to increase it
- Cost per retained user is below the value of a retained user, with two weeks of stable data behind it.
- Creative is still producing new winners rather than repeating.
- The channel has not shown frequency saturation, which appears as rising cost at flat volume.
If all three hold, raise budget by roughly 20% at a time and re-read. Larger jumps disturb the auction and make the result uninterpretable.
What to stop funding
Any line that has run for a month without a number attached to it. This usually means brand campaigns, secondary markets started out of curiosity, and tools nobody opens.
The smallest line in the budget is the one the rest depends on.