Header Bidding vs Waterfall: What's the Difference?

Header bidding vs waterfall, in one line

Header bidding runs one auction where every demand partner bids on an impression at the same time. The waterfall offers that impression to networks one at a time, in a pre-set order, until one accepts. That single difference — simultaneous competition versus sequential fallback — is why header bidding almost always earns publishers more per impression.

If you only remember one thing: the waterfall guesses who will pay the most based on the past, while header bidding asks everyone what they’ll pay right now.

How the waterfall works

The waterfall (sometimes called a daisy chain) is the older method. Ad networks are ranked by their historical average CPM — the network that has paid the most on average sits at the top. Each impression is offered down the chain:

1. The top network gets first look. If it fills the impression, the auction ends.
2. If it passes, the impression drops to the second network.
3. This continues until a network accepts — or the impression goes unsold.

The problem is baked into the design:

• Pricing is estimated, not live. A buyer sitting third in the chain might have paid more for this specific user than the network at the top — but it never got asked.
• Latency stacks up. Every pass adds a round trip. A long waterfall can add hundreds of milliseconds, hurting both revenue and Core Web Vitals.
• Fill is lost. Impressions can go unsold even when real demand existed lower in the chain.

How header bidding works

Header bidding flips the model. Before the ad server is even called, every connected demand partner is asked for a real bid on the actual impression, simultaneously. The highest bid is passed into the ad server, where it competes with direct deals and Google’s own demand on equal footing.

The gains come from three places:

• Real bids on real impressions. Buyers price the exact user in front of them, not a historical average.
• Unified competition. Everyone competes at once, so the true market-clearing price surfaces.
• Less waste. There’s no sequential drop-off; the whole pool sees the impression together.

Side by side

• Auction type — Waterfall is sequential; header bidding is simultaneous.
• Pricing basis — Waterfall uses historical averages; header bidding uses live, per-impression bids.
• Partner competition — Waterfall calls one partner at a time; header bidding calls all at once.
• Latency — Waterfall grows with chain length; header bidding is one timed auction.
• Revenue outcome — Waterfall leaves money on the table; header bidding lifts yield per impression.

So is the waterfall dead?

Not entirely. You’ll still see waterfall logic in a few places:

• Inside mediation on mobile apps, where some SDKs still fall back to a chain for certain networks.
• As a backup layer beneath header bidding, catching impressions that the unified auction doesn’t fill.
• In legacy setups that haven’t migrated yet.

The modern best practice is header bidding as the primary auction, with a thin fallback rather than a deep chain.

What actually matters when you switch

Header bidding wins on paper, but a sloppy implementation can erase the gains:

1. Timeout tuning. Too short and you cut off good bids; too long and you hurt page speed. This is the single most important dial.
2. Partner count. More partners is not automatically more money — each one adds page weight. The goal is the right partners, not the most.
3. Page performance. Asynchronous loading, lazy loading below-the-fold slots, and careful script management keep the auction from dragging down the user experience.
4. Governance. As you add partners, you need visibility into who’s bidding, what’s rendering, and whether any partner is introducing quality or policy risk.

That last point is where most publishers under-invest. Running a live auction across many partners is a revenue win and an operational surface that needs monitoring. This is exactly the layer AdGhost is built to manage — competition on the front end, governance and quality signals behind it.

The bottom line

The waterfall was a reasonable answer to a harder technical problem than we have today. Header bidding replaced it because simultaneous, real-time competition simply prices inventory better than a fixed order based on the past. For most web publishers, moving from waterfall to header bidding — implemented with disciplined timeouts and clean page performance — is one of the highest-leverage monetization changes available.

If you’re weighing a migration, start with a website monetization audit of your current stack: how deep your waterfall runs, how many partners you’re calling, and what it’s costing you in latency and unfilled impressions.