Waterfalling is one of the traditional ways publishers manage programmatic ad inventory. Instead of letting several demand sources bid for an impression at the same time, the ad server approaches them one after another. Each demand source gets its turn based on a predefined priority order.
The name comes from the way the process works. Demand sources are arranged in a sequence, and the ad request moves down that sequence until a partner can fill the impression. A publisher might place a high-paying ad network first, followed by another network, an SSP, and finally a lower-paying source. If the first partner cannot fill the impression, the request moves to the next one.
This approach was widely used before real-time auction-based methods such as header bidding became common. It is relatively straightforward to understand and gives publishers direct control over which demand sources are called and in what order.
Imagine a publisher has an ad slot available on a page. The publisher’s ad server first checks the highest-priority demand source. If that source can return an ad, the process stops, and the impression is filled. If it cannot, the ad server calls the next source in the sequence.
The order usually reflects historical performance. A publisher might rank a demand partner that typically generates a high CPM above one that has delivered lower prices in the past. The underlying assumption is simple. If a partner has paid well before, it should get the opportunity to fill the next impression before a lower-paying partner.
A simplified waterfall could look like this
Demand source A → Demand source B → Demand source C → Demand source D
If A fills the impression, B, C, and D are never asked. If A cannot fill it, the request continues to B. The same happens further down the chain until an ad is returned or no demand source remains.
This makes waterfalling fundamentally sequential. The publisher decides who gets access to the impression first, instead of allowing multiple buyers to compete for it simultaneously.
Waterfalling solved a practical problem for publishers. Ad inventory had to be monetized across multiple demand sources, but there was no simple way to let them all compete for every impression in real time.
A priority list provided a manageable alternative. Publishers could connect several networks and adjust their order based on observed revenue. If one partner consistently generated better results, it could move higher in the waterfall. Lower-performing sources sat further down and mainly served as a backup when higher-priority demand couldn’t fill the slot.
The model also gave publishers clear operational control. They knew which partner to contact first and could build different waterfalls for different ad units, formats, or audiences.
There was a catch, though. But historical performance is not necessarily a good indicator of what a particular impression is worth right now.
The biggest weakness is that demand sources do not compete against each other for the same impression. A partner can win simply because it sits higher in the sequence, even if another partner further down the waterfall would have paid more.
Suppose demand source A has historically delivered a $3 CPM and demand source B has delivered $2.50. A publisher therefore puts A first. On a particular impression, however, B might be willing to pay $5 while A would only pay $2.50. Under a traditional waterfall, B does not get the opportunity to compete because A has already filled the impression.
This can result in opportunity cost for the publisher. The ad slot is sold, but it may not have been sold to the demand source willing to pay the most for that particular impression.
Another issue is latency. When a demand source fails to fill the impression, the system must call the next source. A long waterfall can therefore involve several sequential requests before an ad returns. Depending on the implementation, this can add delay to ad delivery and page rendering.
There is also a management problem. Publishers need reliable performance data to keep the waterfall in a sensible order. Changes in demand, seasonality, audience composition, geography, or advertiser budgets can make an old priority structure less effective over time.
The logic behind waterfalling becomes easier to understand when looking at how pricing is determined. In a traditional setup, publishers often use historical CPMs or negotiated rates to establish the order. A demand source with a higher expected yield receives a higher position. The ad server then works through the sequence according to those predefined priorities.
This creates a relatively static decision process. The impression’s value is estimated based on previous results rather than discovered through competition for the individual impression. That distinction is important. A waterfall can still generate revenue efficiently, particularly when demand sources are well understood and carefully managed. But it does not provide the same auction dynamics as a system in which multiple buyers can bid on the impression at the same time.
Header bidding was developed largely to address this limitation. Instead of waiting for one demand source to respond before contacting another, header bidding allows multiple demand partners to submit bids for an impression before the publisher’s ad server makes the final decision.
The publisher can then compare the bids and select the winning demand based on the configured auction logic.
| Waterfalling | Header Bidding | |
|---|---|---|
| Demand access | Sequential | Multiple demand sources compete |
| Decision basis | Usually historical yield or predefined priority | Current bid for the impression |
| Competition | Limited | Direct competition between participating bidders |
| Publisher control | High control over priority order | High control over participating demand and auction rules |
| Potential yield | Can be limited by fixed ordering | Can increase competition for valuable impressions |
| Request flow | One partner after another | Multiple partners can be contacted in parallel |
| Latency considerations | Can increase with long waterfalls | Parallel bidding can also create latency if poorly configured |
| Management | Requires ongoing priority adjustments | Requires bidder and auction management |
| Typical role today | Legacy or fallback setups | Common approach for competitive programmatic demand |
Header bidding does not automatically make every setup faster or more profitable. It adds technical complexity, including bidder configuration, timeout management, auction mechanics, and consent requirements. But its central advantage is clear. Demand sources can compete for the same impression instead of waiting for their position in a fixed sequence.
Waterfalling hasn’t disappeared just because header bidding exists. Publishers may still encounter waterfall structures within ad networks, SSP configurations, or hybrid monetization setups. Some systems also use sequential demand as a fallback when a primary auction does not produce a suitable result.
For smaller publishers, a simple priority-based setup can sometimes be easier to operate than a large header bidding stack. The trade-off is that simplicity comes with less direct competition between demand sources.
The key question, then, is not whether waterfalling is inherently good or bad. It is whether the chosen demand setup gives each impression a reasonable opportunity to reach the demand source that values it most.
For publishers managing their monetization through WordPress, the distinction also helps explain why ad management is more than simply placing an ad code on a page. The delivery logic behind each placement influences which demand sources get access to the inventory, when they are called, and what happens when the primary demand does not fill the slot.
Waterfalling remains a useful concept to understand because many modern ad stacks have evolved from this model. Once the sequential logic is clear, it becomes much easier to understand the rise of header bidding, unified auctions, and other demand-optimization approaches.
