Waterfall or Bidding — How Do You Actually Maximize Ad Revenue?
Most teams wire up an ad SDK, turn on a few networks, and assume the mediation layer is a solved problem. It isn't. The mediation layer — the thing that decides which network serves each impression — is where a surprising amount of ad money is left on the table, and almost nobody looks at it after launch. Here is what the layer actually does and how to stop losing impressions to a static spreadsheet.
Q: What is the actual difference between a waterfall and in-app bidding?
A waterfall is you guessing. You rank ad networks into tiers by what you think their eCPM is, set a floor price on each tier, and ask them top to bottom: whoever clears their floor wins the impression. The "price" is a number you typed in last Tuesday. In-app bidding flips that — networks return a real bid in milliseconds, and the mediation SDK picks the highest. The difference is not cosmetic: a waterfall prices your inventory off a stale assumption, bidding prices it off what buyers will pay right now. Bidding has eaten most of the waterfall over the last few years precisely because static floors can't track a buyers' market that moves by the hour.
Q: Why does a pure waterfall quietly leak money today?
Three ways, and they compound. Set floors too high and the top tier doesn't fill, so the impression falls through to a lower tier that clears cheaply — you just sold premium traffic at a discount. Set them too low and you hand high-value impressions to a network that would have paid more. And because you re-tune manually, you're always half a week behind: buyer budgets spike and crash within a day, but your waterfall was last touched on Monday. The tell isn't a line on a dashboard — it's fill rate on your top tier sitting at 40% while your blended eCPM quietly bleeds. A waterfall turns pricing power over to your calendar, not the market.
Q: How do you decide which your game should actually run?
You almost never pick one. The real setup is hybrid: bidding is the price-discovery floor that catches every impression, and a thin 1–2 tier waterfall sits on top for direct deals or networks that don't bid. The decision factors are scale, region, and team. Tiny traffic doesn't justify a person maintaining floors. Some emerging markets only fill through waterfall networks. And the counterintuitive part: bidding won't always win every single impression, but it kills the leak-and-discount pattern, so small and mid teams usually net more, not less. If a vendor tells you bidding always beats waterfall on every view, they're selling you the dashboard, not the truth.
Q: What do you actually evaluate when picking a mediation SDK?
Stop staring at the eCPM number — a mediation SDK can flatter its own eCPM by filling with house ads or its own network. Look at fill-rate honesty, SDK size (a fat SDK drags load time and quietly drops retention), data compliance (where does impression data leave the device — COPPA and GDPR care), payout terms, and lock-in. The only real audit is independent: instrument impression-level revenue yourself and reconcile it against what the SDK reports. If the two drift, you're not maximizing revenue, you're maximizing the SDK's reported number.
| Axis | Waterfall | In-app bidding |
|---|---|---|
| Pricing basis | your typed floor | real-time bid |
| Tuning cost | constant manual | near-zero |
| Leak risk | high (mis-set floors) | low |
| Best for | direct deals, thin regions | price discovery, scale |
| Honest eCPM | easy to fool yourself | harder to hide |
