Ad mediation is quietly moving from waterfall to hybrid

Most studios set up ad mediation once, during first integration, and never look at it again. The waterfall gets ordered by average historical yield, the networks get plugged in, and the whole thing runs on autopilot for years. That used to be fine. It isn't anymore.

Why a static waterfall leaves money on the table

A waterfall asks each network, one at a time, "will you pay at least this floor price?" The order is fixed ahead of time, based on which network paid best on average last month. The problem is that "on average" hides a lot. A network might pay well for a US player on a Tuesday and badly for the same player on a Saturday. A pure waterfall can't react to that — it just works down its fixed list until someone says yes, missing whichever network would have paid the most for this exact impression, right now.

That gap used to be small enough to ignore. It isn't now that more networks bid programmatically in real time. When several networks can name their actual price for an impression instantly, a fixed order stops being a ranking and starts being a guess.

What the hybrid model actually does

The setup studios are moving to combines two mechanisms instead of picking one. A small set of networks that support real-time bidding compete against each other instantly for every impression — whoever bids highest wins that specific impression. Below that bidding layer, a traditional waterfall still runs for networks that don't support real-time bidding, checked in the old fixed order as a backup.

In practice this means your top networks are fighting for every single impression on its own merits, while your smaller or legacy networks still get a fair shot lower down the stack. Neither mechanism replaces the other completely — the hybrid setup keeps whichever one produces more revenue for a given impression, without you having to guess in advance which network that will be.

The shift matters because it changes where the yield actually comes from. In a pure waterfall, yield comes from picking a good fixed order and re-ordering it periodically by hand. In a hybrid setup, yield comes from letting competition do that ordering automatically, impression by impression, without anyone touching a dashboard.

What changes in your setup

Moving to a hybrid model is not a rip-and-replace project. Most mediation SDKs already support bidding alongside waterfall networks — the work is mostly configuration, not a new integration. Three things are worth checking before you touch anything.

First, confirm which of your current ad networks actually support real-time bidding through your mediation SDK. Not every network does, and the ones that don't will simply sit in the waterfall layer as before — nothing breaks, they just don't get the speed benefit.

Second, check your floor prices. A pure waterfall relies on floors to protect against underselling an impression. A bidding layer competes on price directly, so an aggressive floor that made sense for a fixed-order waterfall can actually suppress bids in a hybrid setup and cost you fill rate instead of protecting revenue.

Third, look at where the bidding layer sits relative to your rewarded video networks specifically. Rewarded ads carry the highest player tolerance and usually the highest eCPM in a mobile game, so they're the format where the yield difference between a stale waterfall order and live bidding shows up fastest. If you only have time to reconfigure one placement first, start there rather than interstitials or banners — the upside is bigger and the risk of a visible player-facing regression is lower, since a rewarded placement only fires when the player asks for it.

What to measure before you call it done

The only honest way to know whether a hybrid setup earned its complexity is to run it against your existing waterfall as a real holdout, not a before-and-after comparison across different weeks. Ad revenue moves with seasonality, player mix, and campaign timing on its own — comparing "last month" to "this month" will tell you a story that has nothing to do with mediation.

Split your traffic, run both setups in parallel for long enough to smooth out day-to-day noise, and compare eCPM and fill rate on the same cohort of players over the same window. If the hybrid setup isn't beating the waterfall on that basis, the added complexity isn't earning its place yet — leave the waterfall alone until it is.

Watch fill rate as closely as eCPM. A hybrid setup that raises average price per impression but drops fill rate for lower-value inventory can end up flat or worse on total revenue, especially in markets where your eCPM was already thin. The two numbers move independently, and only tracking one of them is how a "successful" migration quietly loses money in the segments you weren't watching.

This is the same discipline that matters anywhere you're testing a change to your revenue stack: a held-out comparison group beats a timeline comparison every time, whether you're testing an ad setup or an AI-driven offer. Guessing from the trend line is how studios end up chasing noise instead of real gains.

Ad mediation isn't the flashiest line item in a monetization stack, but it's one of the few places where a configuration change — not a redesign, not new content — can move revenue without touching the player experience at all. Worth the hour it takes to check.