The expensive failures on mobile user acquisition platforms are rarely the loud ones. A campaign that spends nothing is annoying but it announces itself. The campaigns that hurt are the ones that run smoothly for weeks, report an acceptable cost per install, and deliver users who never come back. Most ironSource ads mistakes belong to that second category. Everything on the dashboard looks fine and the damage only becomes visible when someone finally compares day thirty revenue against what was spent to acquire the cohort.
We manage in-app acquisition alongside the rest of a client’s media mix, which means we usually see the same errors from the outside before we inherit them. They repeat across accounts and across studios of very different sizes, which suggests they are structural rather than careless. Below are the ones that cost the most, with what to do instead. None of them require exotic tooling to fix. They require deciding what a good user is before you start buying, and then refusing to optimise toward anything easier.
Optimising for installs when you are paid by retention
Cost per install is the most available number and the least useful one. If the campaign is told to find cheap installs, it will find them, and cheap installs on incentivised or low-quality inventory are extremely available. The account then looks efficient while the cohort decays within days. Point optimisation at an event that correlates with value: tutorial completion, first session depth, a purchase, an ad revenue threshold. Even an imperfect proxy beats an install count.
The prerequisite is that the event is actually being sent. Plenty of accounts select a deep event in the interface while the SDK or the measurement partner never fires it in volume. The campaign then optimises against a signal that is too sparse to learn from, which produces erratic delivery that gets blamed on the platform. Verify events with live traffic and check the counts, not the integration status page.
The ironSource ads mistakes hiding in source-level data
Publisher and app-level performance varies enormously. Treating the campaign as a single unit means a handful of poor sources quietly absorb budget behind a tolerable blended average. Pull source-level reporting regularly and act on it, but act on cohort quality rather than install cost. A source with a high install price and strong retention is usually your best source, and it is the one a cost-led cleanup deletes first.
These are the patterns worth checking every week before anything else.
- Sources with an unusual concentration of installs in a short window, which often signals low-quality or automated traffic.
- Cohorts where day one retention looks normal but day seven collapses, pointing at mismatched creative promise rather than a bad product.
- Creatives whose install rate holds while cost per deep event climbs, which is fatigue arriving before the obvious metric shows it.
- Geographies bundled into one campaign despite very different monetisation, so a cheap market masks a failing expensive one.
- Bid changes made the same day as creative changes, which makes both untestable.
Creative treated as an asset instead of a consumable
Playables and video creatives decay. They decay faster than most teams plan for, and the decay is usually gradual enough that it gets attributed to seasonality or competition. If your creative pipeline produces a new batch occasionally rather than continuously, you will spend a predictable portion of every month running fatigued assets at deteriorating efficiency.
The other half of this mistake is creative that oversells. An ad showing gameplay or functionality that the app does not deliver will win the auction and lose the user, and it shows up as a retention problem rather than a creative problem. When day seven numbers are weak across an otherwise healthy account, look at what the winning creative promised before you look at onboarding.
Impatient bidding and over-segmentation
Changing bids daily keeps campaigns permanently unstable. Every meaningful change restarts learning, and an account edited every morning never completes a cycle, so it never produces the stable data the editor is looking for. Set a bid, give it several days, judge it on a full cohort window, then move it in measured steps.
Over-segmentation compounds this. Splitting a modest budget across many narrow campaigns leaves each one below the volume needed to optimise, so all of them behave erratically and the account looks like a platform problem. Consolidate until each live campaign can clear a reasonable number of target events per day, then split only when volume genuinely justifies it.
Measuring on a window shorter than your payback
The last of the common ironSource ads mistakes is structural rather than tactical. If your business recovers acquisition cost over weeks, judging campaigns on a few days of data means you are systematically cutting the cohorts that pay back slowly and scaling the ones that spike early and die. Define your payback window first, build a simple cohort view that matches it, and hold decisions until the window closes. It feels slow. It is considerably cheaper than the alternative, which is a year of confident decisions made on data that was never capable of supporting them.
Keep reading: Ironsource Ads · Werbung