When to Switch From Waterfall to In-App Bidding
Most games should already run bidding-first mediation. The real questions are what to keep in the waterfall, how to migrate without a revenue dip, and how to tell whether it worked.
Should my game still use a waterfall?
For most mobile games, no — not as the primary structure. In-app bidding runs a simultaneous auction where each network values the specific impression, which removes the guesswork of waterfall ordering and the latency of sequential calls. The major mediation platforms — AppLovin MAX, Unity LevelPlay and Google AdMob — are all built around bidding, with waterfall lines kept for demand sources that do not bid.
A manual waterfall still makes sense in narrow cases: a network with strong demand in a specific geo or format that does not offer bidding, or a direct deal you want to protect. Those should be deliberate exceptions, not the default.
When is the right time to migrate?
When any of these are true:
- Your waterfall has many instances and you see load latency or timeouts before ads are ready — the Mediation Latency calculator estimates the leakage.
- Your top networks all support bidding in your mediation platform.
- Your team spends more time re-ordering floors than analysing placements.
I rebuilt mediation on both AppLovin MAX and ironSource LevelPlay while moving from waterfall to bidding. The migration is rarely the hard part. Reading it correctly is.
How do you migrate without a revenue dip?
- Baseline first. Record revenue per eligible opportunity, fill, show rate and latency by placement, geo and platform for at least two weeks.
- Move one placement or geo at a time, or use your platform’s A/B tooling to split traffic.
- Keep a small number of waterfall lines for non-bidding demand that genuinely adds value.
- Remove dead instances. Every network SDK you keep adds weight, start-up time and crash surface.
- Wait for settled data. Ad revenue is revised for days; read results on complete data, not on day one.
How do you know it worked?
Not by fill rate. When we moved from waterfall to bidding, fill improved first while revenue per eligible opportunity did not move at the same rate — more opportunities were clearing, but the additional fills were not creating proportional revenue. Read revenue per eligible opportunity by geo and placement, alongside bid rate, win rate and show rate for each network. Bidding improves competition; it does not remove the need to inspect delivery.
Related: Fill Rate Is Not a Strategy · Network Revenue Quality Is More Than eCPM · AppLovin MAX vs Unity LevelPlay vs AdMob