For years, the waterfall was how apps decided which ad network filled each impression. It is a mediation method that ranks demand sources by their historical earnings and calls them in order, one after another, until one accepts the impression. The name captures the shape, since the request cascades down a list of networks and stops at the first that fills.

How it works

Under ad mediation, a waterfall sorts demand sources from highest expected eCPM to lowest, using each network's past average. When an impression is available, the top network gets first refusal. If it declines or has no matching demand, the request drops to the next, and so on down the list. The impression is filled by the first network that says yes, not necessarily the one willing to pay the most for that specific impression.

Waterfall vs. in-app bidding

That last point is the waterfall's core weakness, and it is what in-app bidding fixes. A waterfall ranks networks on stale historical averages, so a network that would pay a high price for one particular user might sit lower in the order and never get asked, while a higher-ranked network fills the impression for less. In-app bidding instead has all demand sources bid on each impression in a single live auction, so the impression goes to whoever values it most right now. The waterfall guesses from the past, while bidding prices the present.

Why publishers are moving on

The practical cost of a waterfall is money left on the table. Because it fills on order rather than on true value, higher bids are routinely skipped, and the sequential calls also add latency as each network is asked in turn. In-app bidding recovers that lost revenue and speeds up the decision, which is why it has become the default for most modern mediation. Many stacks still run a hybrid, with a bidding layer competing against a few remaining waterfall lines.

The offerwall in a waterfall setup

An offerwall is a distinct, high-eCPM source that fits either arrangement. It is not a passive banner or video line, it is a rewarded placement users open on purpose, and it earns on completed offers. In a waterfall, an offerwall can sit as its own high-value line, and in a bidding setup it can compete for the users it suits. Because its earnings come from engagement rather than auctioned impressions, an offerwall adds revenue that does not depend on how the rest of the waterfall is ordered, and it tends to lift both eCPM and overall fill by giving motivated users something to do.

Common mistakes to avoid

• Leaving waterfall positions on old eCPM estimates that no longer reflect what each network actually pays.

• Treating a waterfall and in-app bidding as mutually exclusive. Many publishers run both in a hybrid setup during the transition.

• Chasing fill rate alone. A line that always fills at a low eCPM can earn less than one that fills less often at a higher price.