Home / Blog / Article
Backend Engineering · Aug 13, 2026
Ad Mediation Waterfalls: A Technical Guide for Hybrid-Casual Titles
For most hybrid-casual titles, ad revenue is not a secondary monetization channel, it is the primary one, and the mediation layer sitting between your game and a dozen ad networks is doing more to determine that revenue than almost any other single technical decision in the codebase. A poorly configured waterfall leaves real money on the table every single day at scale, quietly, without ever showing up as a bug. This is how we structure mediation to actually capture that revenue.
Waterfall Versus Bidding: Why The Distinction Still Matters
A classic waterfall calls ad networks in a fixed priority order, each network getting a chance to fill the request before falling through to the next, with priority manually set based on historical average eCPM. In-app bidding, by contrast, lets every integrated network bid in real time for each individual impression, with the highest bid winning regardless of a manually configured priority order. Bidding captures more revenue per impression because it reacts to real-time demand rather than a stale historical average, but it requires more integration work per network and is not universally supported. Most mature mediation setups run a hybrid: bidding networks compete in real time, and traditional waterfall networks fill in below the bidding floor, capturing both models’ strengths.
Floor Pricing And The Cost Of Getting It Wrong
Setting a floor price too high means impressions go unfilled entirely, and unfilled impressions generate zero revenue regardless of what a filled one might have earned. Setting it too low means networks fill every request at a price below what real demand would have paid, leaving revenue on the table on every single impression at scale. We do not set floor prices once and leave them; we review fill rate against eCPM weekly per ad unit, since demand shifts by region, by season, and by the specific ad unit placement, and a floor price that was optimal in March is frequently leaving money on the table by June without ever throwing an error that would flag the problem.
Placement-Specific Waterfall Configuration
Treating rewarded video, interstitial, and banner placements with one shared waterfall configuration ignores that these ad formats have fundamentally different demand curves, different eCPMs, and different player tolerance for latency before the ad actually shows. Rewarded video, where the player has opted in and expects genuine value in return, supports a longer fill-attempt window and can prioritize networks with the highest eCPM even at the cost of a slightly longer load time. Interstitials, shown at natural pause points but without explicit opt-in, need a tighter latency budget since a slow-loading interstitial reads as the game hanging. We configure waterfalls per placement type independently rather than sharing one configuration across every ad format in the game.
Latency Budgets And The Revenue-Versus-Experience Tradeoff
Every additional network a waterfall tries before finding a fill adds latency, and latency past a few hundred milliseconds starts measurably hurting both ad viewability, since a slow load increases the chance a player has navigated away before the ad renders, and player experience, since a visible loading delay before a rewarded video reads as broken rather than intentional. We set an explicit timeout per waterfall tier, not per individual network call, so a single slow-responding network cannot stall the entire fill attempt and cost revenue on every network below it in priority order that never got a chance to respond within the available time budget.
Regional And Segment-Based Waterfall Variants
A single global waterfall configuration significantly underperforms because ad demand, and therefore optimal network priority and floor pricing, varies meaningfully by region, and increasingly by player value segment as networks build more sophisticated targeting into their bidding. We run distinct waterfall configurations for at minimum tier-one Western markets, where demand and eCPM are highest and network competition is most aggressive, versus emerging markets where fill rate matters more than per-impression price. Segmenting further by player LTV tier, showing higher-value players ad experiences tuned for engagement over raw fill rate, is the next layer of sophistication once the regional split is already capturing its available gains.
Testing Waterfall Changes Without Guessing
A waterfall reconfiguration is, functionally, an A/B test on real revenue, and treating it as anything less risks a bad change silently costing money for weeks before anyone notices the eCPM trend line moved. We run every meaningful waterfall change, a new network integration, a floor price adjustment, a priority reorder, through the same server-side experiment infrastructure used for other LiveOps tests, with revenue per user and fill rate as the primary metrics and a guardrail on ad load latency, rather than pushing a config change to one hundred percent of players and hoping the historical eCPM data justified the decision.
An Ad Mediation Health Checklist
Before considering a mediation setup optimized, confirm: the configuration uses a hybrid of real-time bidding and waterfall fill rather than pure waterfall alone; floor prices are reviewed against fill rate and eCPM on a recurring weekly cadence, not set once at launch; rewarded, interstitial, and banner placements each have independently tuned waterfall configurations; explicit per-tier latency timeouts prevent one slow network from stalling the entire fill attempt; regional and, ideally, player-value-segment variants exist rather than one global configuration; and waterfall changes are tested through proper experiment infrastructure rather than shipped to full traffic on assumption. Setups that check all six are typically capturing revenue that studios running a single unreviewed global waterfall configuration are leaving on the table every single day, invisibly, without a single error log to flag it.
Keep Reading
More from the blog
Want these insights on your build?
The engineers writing these articles are the ones who would work on your game.