Ironsource Ads Trends in 2026: What Changed and What Still Works

Unity and ironSource completed their merger in late 2022. Since then, the product names have shifted. ironSource’s mediation platform became Unity LevelPlay, and user acquisition tools from both sides have been brought closer together. Many advertisers still say “ironSource ads” when they mean buying installs across this network.

For a practical buyer, the change means one account relationship covers more inventory, but documentation, dashboards and API endpoints have moved. If your team built scripts against older reporting endpoints, check that they still return complete data.

Value-based bidding replaces install volume

Bidding for cheap installs is giving way to bidding for return on ad spend or for a deeper event such as a purchase or level completion. The networks have pushed ROAS campaign types because install volume alone tells them little about user quality. This only works when you send in-app revenue events through your measurement partner reliably.

Bidding mediation is the default

On the monetisation side, real-time bidding has largely replaced waterfall setups in LevelPlay. Publishers set fewer manual price floors and let networks bid per impression. For advertisers, it means competition for each impression is more direct, and creative quality affects how often you win.

Playables and interactive end cards

Playable ads let a user try a slice of the game before installing. They are more expensive to produce than a static video, but they filter out people who do not enjoy the core loop. Teams that test playables usually see fewer installs at a higher quality.

Privacy-first measurement

Apple’s App Tracking Transparency and SKAdNetwork changed iOS measurement years ago, and in 2026 most teams are working with SKAdNetwork 4 conversion values or AdAttributionKit. Android is moving in a similar direction through the Privacy Sandbox. Reporting is delayed and aggregated, so decisions rely more on modelled data and cohort trends than on individual user paths.

Plenty of fundamentals have not changed, and they are where most gains still come from.

• Creative testing with a clear control: one variable per test, enough impressions to judge, and a rule for when to stop.
• Country tiering: campaigns split by market value, because a user in one country can be worth many times a user in another.
• Source-level optimisation: reviewing performance by publisher app and lowering bids on sources with poor retention.
• Day-7 and day-30 cohort reporting rather than daily install counts.
• A tidy event map shared between product and marketing teams so everyone agrees what a “conversion” is.

Start with measurement. If your MMP is not sending revenue events to the network, ROAS campaigns have nothing to optimise toward. Check postback configuration in your MMP and confirm events appear in the campaign dashboard within a day.

Next, review creative production. If your library is mostly 30-second videos made for other networks, budget for at least a few playable concepts and shorter vertical cuts. Interactive end cards are a cheaper first step than full playables.

Then simplify campaign structure. Many accounts still carry dozens of campaigns from the install-bidding era, each with a small budget and its own learning history. Old campaigns also make reporting harder, since the same country appears in six rows with six different settings. Consolidating into fewer ROAS campaigns per country tier gives the algorithm more data per campaign.

An example budget split

For illustration, a casual game spending the equivalent of 20,000 US dollars a month might set aside 60 percent for proven ROAS campaigns in top markets, 25 percent for expansion markets on deeper-event bidding, and 15 percent for creative tests including one new playable a month. The exact split matters less than having a fixed share for testing that is protected from cuts.

The most common one is switching to ROAS bidding before there is enough revenue data. The campaign then struggles to spend, and the team switches back to install bidding, concluding that ROAS does not work. Wait until you have a steady flow of purchase events.

Another is trusting a single attribution source on iOS. SKAdNetwork data is delayed and coarse. Compare it with your MMP’s modelled data and your own revenue totals before cutting a campaign. A third mistake is ignoring the monetisation side: if your own game uses LevelPlay, changes there can affect the economics of your acquisition spend.

Because iOS postbacks arrive late and in aggregate, a test that looks weak on day 3 can look fine on day 10. Set a minimum test length before you start, usually two full weeks, and a minimum spend so each variant reaches a few hundred installs. Compare variants on the same cohort day, not on whatever the dashboard shows today.

Keep a simple test log with the hypothesis, start date, budget, metric and result. After a few months, the log becomes the most useful document the team owns, because it stops the same failed idea from being tested three times by three different people.

Signals worth watching weekly

• Cost per install by country tier, to catch auction pressure early.
• Day-7 ROAS by campaign, compared with the same week last month.
• Share of spend going to the top ten publisher apps.
• Creative age: how many weeks the leading video has been live.