In-App Bidding vs Waterfall Mediation: How Ad Mediation Works

When a mobile game is about to show an ad, a decision happens in a fraction of a second: which of many ad networks gets to fill that slot, and at what price. The software that makes the decision is a mediation platform, and the way it decides — a waterfall, an auction, or a mix of the two — has a direct effect on how much the game earns and on which advertiser's creative the player ends up seeing.

This guide explains both methods from the publisher's side, then turns around and looks at the same auction from the advertiser's side, because that is where a playable ad's chance of being served is decided.

What ad mediation is

A game that relies on a single ad network leaves money on the table: that network will not always have the highest-paying ad for every player, country and placement, and sometimes it will have no ad at all. Mediation connects the game to several demand sources through one SDK and decides, impression by impression, which of them fills the slot.

Common mediation platforms include AppLovin MAX, Unity LevelPlay (formerly ironSource mediation) and Google AdMob mediation. Each lets a publisher add demand partners, set rules, and report revenue across all of them in one place. The AppLovin MAX guide covers one of them from the advertiser's angle.

How waterfall mediation works

A waterfall is an ordered list. The publisher ranks networks — usually by their historical average eCPM, the revenue per thousand impressions — and the mediation layer asks each in turn whether it has an ad at or above a set price.

1. The game requests an ad for a placement.
2. Mediation calls the first network in the list with a floor price attached.
3. If that network returns an ad at or above the floor, it wins and the ad is shown.
4. If not, mediation moves to the next network or the next, lower price tier, and asks again.
5. This repeats until an ad fills or the list runs out.

The same network often appears several times at different floors, so a waterfall for a busy game can have many tiers and needs constant tuning.

The weaknesses of a waterfall

• It ranks by the past, not the present. A network ranked third might be willing to pay more for this particular impression, but it never gets asked first.
• Each step adds latency. Asking networks one after another takes time, and a slow fill can mean a missed impression.
• It needs manual upkeep. Floors and order drift out of date as demand changes by country, season and placement.
• It hides true value. A network that would have paid above the floor only pays the floor, so the publisher earns less than the impression was worth.

How in-app bidding works

In-app bidding, sometimes called header bidding for mobile or a unified auction, asks every bidding network at the same time. Each one looks at the impression — the app, the placement, the country, what it knows about the device — and returns a real-time bid. The highest bid wins.

The benefits for the publisher follow directly:

• Each impression goes to whoever values it most right now, rather than whoever averaged best last month.
• Less manual work: there is no long list of price tiers to keep in order.
• Lower latency, because the networks are called in parallel.
• Fairer competition between networks, which tends to push bids up over time.

Hybrid setups

Most publishers today run a hybrid: a bidding auction for the networks that support it, with a smaller waterfall behind it for demand that only works on fixed prices. The mediation layer compares the winning bid with the waterfall instances and fills from whichever pays more. The trend across the industry has been steadily towards bidding, but the waterfall has not disappeared.

eCPM floors: what they do and when they hurt

A floor is the lowest price the publisher will accept. In a waterfall it sets each tier; in bidding it is a reserve price below which bids are rejected.

Floors protect value — they stop an impression selling cheaply when better demand might come along a moment later — but they cost fill. Set them too high and the slot stays empty, which earns nothing at all. The usual practice is to set floors per country and per placement, test changes on a share of traffic, and judge them on total revenue per user rather than on eCPM alone, because a higher eCPM on fewer impressions can be a loss.

What this means for advertisers

For anyone buying installs, the publisher's mediation setup is the market their ad competes in. Your bid on a network is not only competing with other advertisers on that network; through bidding, it is competing with every other network's best offer for the same impression.

Networks do not bid a flat price. They bid what they expect an impression to be worth to them, which for performance campaigns depends on how likely the ad is to produce the result the advertiser pays for. That expectation is built from the creative's history: how often it is watched, tapped, installed and, increasingly, how valuable the resulting users turned out to be.

So a creative with a stronger install rate lets its network bid more for the same impression without the advertiser paying more per install, which wins more auctions, which means more delivery. A weak creative loses auctions it would have won, and its spend stalls. How playable ads get delivered covers the learning phase that builds those predictions.

Practical consequences for creative

• Creative quality is a bidding input. Two campaigns with identical targets can win very different shares of the same inventory.
• Format eligibility matters. Rewarded and interstitial placements carry playables; if a placement or network does not accept your format, you are not in that auction at all.
• Technical failures are lost auctions. A playable that fails validation, loads slowly or breaks on a device type cannot be served there; check it against each spec with the playable ad spec checker.
• New creatives start uncertain. Until a network has data, its prediction is cautious, so give new variants enough budget to be learned.

Choosing a mediation approach as a publisher

1. Pick a mediation platform with bidding support from the networks that matter to your audience.
2. Move every network that supports bidding into the auction.
3. Keep a short waterfall only for demand that cannot bid, with floors set per country.
4. Test floor and network changes with a split of users, not all at once.
5. Judge everything on revenue per user and retention together, since ad load affects both.

Where playable creative fits

Bidding has made the auction more efficient, and an efficient auction rewards the creative that converts. That is the part an advertiser controls. A playable that loads instantly, works on every device and shows the real game gives its network a better prediction to bid with, in exactly the rewarded and interstitial slots where mediation is most competitive.

Playable Ads Maker produces those files without a developer: build the playable once in the browser and export a ready-to-upload version for each major network, each with that network's packaging and exit call. More valid formats in more auctions is the simplest way to win more of them; the ad networks overview shows what each one expects.