What is the programmatic waterfall?
The programmatic waterfall offers a single ad impression to demand sources one at a time, in a pre-set order ranked by their historical average price, until one accepts it. Also called daisy-chaining or waterfalling, it was the dominant way publishers sold programmatic inventory before header bidding replaced it around 2014.
The name is literal: the impression “falls” down a ranked list of networks. The first one willing to meet the price floor takes it, and the chain stops.
How it works, step by step
- Networks are ranked by their historical average CPM — the network that has paid the most on average sits at the top.
- The top network gets first look. If it fills the impression at or above the floor, the auction ends.
- If it passes, the impression drops to the next network in line.
- This repeats down the chain until a network accepts — or the impression goes unsold.
Historically, the waterfall grew out of remnant inventory: publishers sold premium inventory through direct deals first, then waterfalled whatever was left through a chain of networks to recover some value.
Why it fell out of favor
The waterfall has two structural flaws that cost publishers money:
- Pricing is estimated, not live. Networks are ordered by past averages, so a buyer sitting lower in the chain might have paid more for this specific impression — but it never got asked. The highest bidder doesn’t necessarily win; the highest-ranked one that clears the floor does.
- Latency and lost fill. Every pass down the chain adds a round trip. Long waterfalls add delay that hurts both revenue and page speed, and impressions can go unsold even when real demand existed further down.
Header bidding fixed both by asking every demand partner for a real bid on the actual impression, all at once. For the full comparison, see header bidding vs waterfall.
Where the waterfall still exists
It hasn’t vanished entirely. You’ll still find waterfall logic:
- Inside mobile app mediation, where some SDKs fall back to a ranked chain for certain networks.
- As a fallback layer beneath header bidding, catching impressions the unified auction doesn’t fill.
- For exclusive first-look deals, giving a specific direct buyer priority before open competition.
- In legacy setups that simply haven’t migrated yet.
The modern best practice is header bidding as the primary auction, with — at most — a thin waterfall fallback rather than a deep chain.
The bottom line
The programmatic waterfall was a reasonable answer to a harder technical problem than publishers face today: it organized demand into a ranked sequence when simultaneous auctions weren’t yet possible. But because it prices on history and sells sequentially, it systematically leaves money on the table. For most publishers, replacing a deep waterfall with header bidding — and letting a managed ad monetization platform run the competition — is one of the clearest revenue upgrades available.