The Two Models That Determine How Your Ad Inventory Gets Sold
Every time your app requests an ad, a decision process runs behind the scenes to determine which advertiser wins the impression and how much they pay. For years, this process followed a sequential model known as the waterfall. More recently, unified auctions (also called header bidding or in-app bidding) have disrupted that model by letting all demand sources compete at the same time. Understanding the mechanics of each approach is essential for publishers who want to maximize revenue from their ad inventory.
How the Traditional Waterfall Works
The waterfall model operates on a simple sequential principle. When an ad request fires, it moves through a ranked list of demand sources one at a time. Each source gets a chance to fill the impression at or above a predetermined floor price. If it cannot fill, the request passes to the next source in the chain.
Here is the typical flow:
- Step 1: The ad request hits the top-ranked demand source with the highest floor price (say $15 eCPM)
- Step 2: If that source has no matching campaign at that price, the request falls to the next source at a lower floor ($10 eCPM)
- Step 3: This continues down the chain until a source fills the impression or the request exhausts all options and either serves a house ad or returns empty
The ranking is typically based on historical eCPM performance. Publishers set floor prices for each tier and periodically adjust them based on reporting data. This approach has a fundamental flaw: a demand source ranked fifth in the waterfall might be willing to pay $20 for a specific impression, but it never gets the chance because a source ranked second already filled it at $10.
How Unified Auction Works
Unified auction flips the model entirely. Instead of a sequential chain, all participating demand sources receive the ad request simultaneously and submit their bids in real time. The highest bidder wins the impression, and the publisher earns the maximum possible price for that specific request.
The mechanics look like this:
- Step 1: The ad request fires and is sent to all integrated bidding demand sources at once
- Step 2: Each demand source evaluates the impression (user data, context, advertiser budgets) and returns a bid within a defined timeout window (typically 1–3 seconds)
- Step 3: The ad server compares all bids and selects the highest one
- Step 4: The winning creative renders in the app
This is a true auction. Every demand source competes on equal footing for every impression, which eliminates the inefficiency of sequential passbacks.
Revenue Difference: 15–30% Uplift Is Typical
The revenue impact of switching from waterfall to unified auction is well-documented. Publishers consistently report a 15–30% increase in ad revenue after migrating. The reasons are structural:
- True price discovery: Every impression sells for its actual market value, not an arbitrary floor price set weeks ago based on averages
- Reduced passback loss: In waterfalls, each passback introduces latency and lost impressions. Unified auctions eliminate passbacks entirely.
- Increased competition: When all bidders see every impression, the effective bid density increases, pushing clearing prices higher
- Better fill rates: Because all sources bid simultaneously, the likelihood of finding a buyer for any given impression goes up
Latency Comparison
Latency is a common concern when evaluating auction models. Counterintuitively, unified auctions often have lower effective latency than deep waterfalls:
- Waterfall latency: Each passback in the chain adds network round-trip time. A waterfall with 8 tiers where the impression fills at tier 5 requires 5 sequential network calls. This can total 3–5 seconds.
- Unified auction latency: All bid requests fire in parallel. The total latency equals the slowest bidder’s response time (typically capped at 1–3 seconds by timeout). One parallel round instead of multiple sequential rounds.
For users, this means ads appear faster, which improves viewability rates and reduces the chance of users scrolling past an empty ad slot.
Transparency: Seeing What You Could Not See Before
Waterfalls are opaque by nature. Publishers see which tier filled the impression, but they do not see what lower-ranked sources would have bid. This makes optimization a guessing game. You adjust floor prices based on aggregate historical data, never knowing the true clearing price for individual impressions.
Unified auctions provide full bid-level transparency. Publishers can see every bid submitted for every impression, including losing bids. This data reveals:
- Which demand sources consistently bid highest for specific geographies or user segments
- How close the second-highest bid is to the winner (auction pressure)
- Whether certain demand sources are consistently losing by small margins (indicating potential for optimization)
When Waterfall Still Makes Sense
Despite the clear advantages of unified auction, waterfalls are not dead. There are legitimate scenarios where sequential prioritization is necessary:
- Direct-sold campaigns: When a publisher has sold inventory directly to an advertiser at a guaranteed CPM, that deal must take priority over programmatic demand. GAM handles this through priority-based line items that sit above the auction.
- Guaranteed campaigns: Sponsorships, takeovers, and other guaranteed-delivery deals require deterministic serving that auctions cannot provide
- SDK limitations: Some demand sources have not yet implemented bidding adapters. These sources can only participate through the traditional waterfall model.
Hybrid Approaches: The Practical Reality
Most publishers today run a hybrid model. Direct-sold and guaranteed campaigns are prioritized at the top. Below that, a unified auction runs among all bidding-enabled demand sources. Sources that do not support bidding are placed in a traditional waterfall that competes with the auction winner.
This hybrid approach looks like:
- Priority 1: Sponsorships and direct deals (guaranteed delivery)
- Priority 2: Unified auction among bidding partners (real-time competition)
- Priority 3: Waterfall of non-bidding demand sources (sequential fallback)
How Google Ad Manager Handles Both Models
GAM is uniquely positioned because it supports both waterfall and unified auction within the same ad serving decision. Open Bidding allows third-party exchanges to bid in real time alongside Google demand. Meanwhile, traditional mediation partners can be configured as waterfall entries that compete with the auction’s winning bid.
Key GAM features for publishers navigating this space:
- Open Bidding: Server-side bidding integration with third-party exchanges
- Mediation groups: Configure waterfall chains for non-bidding SDK networks
- Dynamic allocation: Lets Ad Exchange compete against every line item in the waterfall in real time
- Unified pricing rules: Set floor prices that apply across both auction and waterfall demand
Practical Migration Steps: Waterfall to Unified Auction
If you are running a pure waterfall today and want to move toward unified auction, follow a measured approach:
- Audit your demand sources: Identify which of your current mediation partners support in-app bidding. Most major networks (Meta, AppLovin, Unity, Pangle) now have bidding adapters.
- Enable bidding for supported partners: In your GAM mediation configuration, switch eligible partners from waterfall to bidding mode one at a time
- Run parallel comparison: Use A/B testing to compare revenue between your existing waterfall and the new bidding setup across a subset of traffic
- Monitor for two weeks minimum: Bidding algorithms need time to learn your inventory. Initial performance may not reflect steady-state results.
- Migrate remaining traffic: Once bidding performance is validated, shift all traffic to the unified auction model while keeping non-bidding partners in a fallback waterfall
The shift from waterfall to unified auction represents the single largest structural change in mobile ad serving in the past decade. Publishers who make the transition methodically, with proper A/B testing and parallel running, consistently see meaningful revenue gains without sacrificing user experience.
RevenueFlex manages GAM waterfalls and auction configurations on behalf of publishers, handling the complexity of hybrid setups so that every impression finds its highest-paying buyer. Whether you are running a pure waterfall, a pure auction, or something in between, optimizing the decision logic is where the revenue gains live.