Bidding vs Waterfall: What Changes When You Switch

Every mediation platform has to answer one question for each ad request: which network gets to fill it? For years the answer was a waterfall, a ranked list tried one network at a time. Today most platforms lean on bidding, where networks compete for each impression at the same moment.

The switch is often presented as a simple upgrade. It usually is an improvement, but it changes more than your eCPM. Fill, latency, reporting and how much control you have over pricing all move too, and knowing what to expect makes the change far less stressful.

This guide explains both models, the hybrid setups most apps actually run, and what shifts when you move from one to the other. Dollar figures are estimates from our benchmark tables, the same ones behind the revenue calculator.

How a waterfall works

In a waterfall, you rank network instances by the price you expect each to pay, usually based on its historical eCPM. Each instance has a floor. When your app requests an ad, the platform calls the top instance first. If it returns an ad at or above its floor, that ad shows. If not, the request falls to the next instance, and so on down the list.

The weakness is that the ranking uses averages, not live prices. A network lower down might have been willing to pay more for this particular user, but it never got the chance because a higher instance filled first at a lower price. To capture more value, publishers add several instances of the same network at different floors, which makes the waterfall longer, slower and harder to manage. Our guide to choosing floor prices covers how those rungs are set.

How in-app bidding works

With bidding, the platform sends the request to every bidding network at once. Each one returns a price for that specific impression, based on what it knows about the user, the placement and its advertisers. The highest bid wins.

Because every bidder sees every impression, the network that values a user most usually wins it. You no longer need to guess rankings or maintain dozens of instances, and the platform updates prices automatically. This is the main reason bidding tends to lift average eCPM for the same audience.

Bidding is not a free-for-all, though. Only networks that support bidding on your platform can take part, and each still needs its SDK or adapter in your app.

Hybrid setups

Most apps run a mix. Bidding networks compete in a real-time auction, and networks that do not bid sit in a waterfall alongside them. The platform compares the winning bid with the expected price of the waterfall instances and serves whichever is higher.

AppLovin MAX is built around bidding first. Unity LevelPlay and AdMob mediation both support bidding and waterfall sources side by side. The exact mechanics differ by platform, so check the current documentation, such as the AdMob mediation guides or the Unity LevelPlay documentation, before you design a stack.

A hybrid lets you keep a valuable non-bidding network while moving the rest of your demand to bidding. Over time, many studios find the waterfall shrinks to a few backstop instances.

Latency

A waterfall is sequential. Every instance that fails to fill adds a network round trip before the next one is tried. A long waterfall can mean a noticeable wait before an ad is ready, which matters most for interstitials and rewarded videos that users are waiting to see.

A bidding auction runs in parallel, so the time is set by a timeout rather than the length of a list. That usually means faster, more predictable loading. Pre-loading ads before they are needed helps in either model, and it is worth tracking the time from request to ad ready in your own analytics.

Transparency and control

Waterfalls give you direct control. You set each floor and can see exactly which instance filled. The cost is the work of maintaining it and the revenue lost to stale rankings.

Bidding gives you less to tune by hand. You see the winning network and price, and some platforms show losing bids too, but you cannot reorder the auction. Some publishers find this uncomfortable at first. The trade is that the auction usually does a better job than manual ranking, and your time moves to formats, placements and frequency instead.

What moves when you switch

Expect several changes at once:

• eCPM usually rises, as networks compete on each impression rather than on averages.
• Fill can move either way. It often stays similar, but removing a long tail of low waterfall instances can leave gaps, so keep a backstop. Our fill rate explainer helps you read the change.
• Network share shifts. A network that dominated your waterfall may win far less in an auction, and one you rarely saw may win more.
• Reporting looks different. Revenue per network will change, and some bidding revenue is reported with a delay before it is finalised.
• Payments stay split. In most mediation setups each network still pays you directly on its own terms, so check each threshold. Our guide to receiving ad payouts abroad covers the details.
• app-ads.txt matters more, since every bidder checks it. See app-ads.txt for Google Play.

An example using our benchmarks

Here is how the benchmark rows for three platforms compare for a Tier 1 Android audience. Revenue per 1,000 requests is eCPM multiplied by fill:

Take a game with 10,000 daily users, each making two interstitial requests and one rewarded request a day. On the AdMob row, that is 20 blocks of 1,000 interstitial requests at $7.04 plus 10 blocks of rewarded at $10.56, or $246.40 a day. The LevelPlay row gives $253.70, and the MAX row gives $306.90, about 25% more than AdMob.

Treat this as a direction, not a promise. The gap in our tables reflects the deeper bidding demand each platform typically brings, and your result depends on which networks you connect, your regions and your formats. Try your own mix in the AppLovin MAX calculator, the LevelPlay calculator and the AdMob calculator, or compare them side by side on the comparison page.

Switching without the drama

Move gradually. Most platforms let you A/B test a bidding setup against your current waterfall on a share of users. Run it for at least two weeks, compare ARPDAU and revenue per request rather than eCPM alone, and watch load times and crash rates as you add adapters. Keep one or two waterfall instances as a backstop until you are sure fill holds. Our guide to mediation or a single network covers the wider choice, the LevelPlay guide goes deeper on that platform, and the AdMob help centre explains how bidding shows up in AdMob reports.