A waterfall sends each ad request to Network A first. If its bid clears your priority threshold, it wins. Otherwise, the request waits for Network B, then moves to Network C. That sequence takes 300 to 500 milliseconds, while hiding whether Network C would have beaten Network A by fifty cents on that impression.
At scale, this missing bid information is the waterfall's structural flaw. It is not a tuning error. Sequential auctions work this way.
The Sequential Problem
Waterfalls assess demand partners individually, following a priority order based on historical CPM, usually averaged across a few weeks. Mobile demand does not follow monthly averages: CPM changes with time of day, geography, advertiser budgets, and day of week. A ranking based on last month's data trails the live market by design.
A waterfall cannot identify the highest price for each impression. Network A wins whenever it fills, even if B or C would pay more. Lower-ranked partners are reached only after higher-ranked ones pass, so their potential bids remain unknown and invisible.
This inefficiency is measurable. A game serving 400,000 rewarded video impressions per month through one top-of-waterfall network effectively prices every impression at that network's decision, subject only to your floor. Partners that might bid aggressively in peak hours or valuable geos never compete.
How Concurrent Bidding Changes the Auction
Real-time bidding sends each impression opportunity to every configured demand partner at once. Each gets a response window, typically 100 to 150 milliseconds, to submit a bid. Mediation gathers the responses, chooses the highest valid bid above your floor, and serves its creative. Latency is limited by your timeout, not by adding network response times together.
Concurrent bidding improves yield in two ways. Every impression becomes a real auction instead of a first-look sequence. Also, partners in a competitive auction have reason to bid what they are actually willing to pay because other partners bid at the same time. Across many impressions, clearing prices tend toward the inventory's true market rate rather than the top waterfall partner's offer.
Non-winning bids also inform later decisions. Lost bid prices can support floor optimization and help estimate the market value of an impression type. Waterfalls lose this signal because lower-priority partners bid only after higher-priority partners pass.
Why Mobile Games Are a Special Case
Game ad placements can magnify waterfall inefficiency. Rewarded video, the highest-eCPM format in most game categories, appears between sessions or after level completions. Interstitials occur at natural transitions. Both create bursts of requests instead of an even flow.
A casual puzzle game producing 300,000 rewarded video impressions monthly does not spread them evenly. Friday evenings in North America, Saturday mornings in South Korea, and Tuesday afternoons in Southeast Asia carry different advertiser demand. A waterfall ranked by one network's monthly average CPM treats them alike. RTB mediation auctions every request again, capturing stronger prices during high demand and using appropriate floors otherwise.
Rewarded video also offers a short opportunity to fill. The user accepted the ad and is briefly paused. Serving the highest available bid instead of the first available bid captures revenue without changing the experience.
RTB Mechanics in Practice
The mediation layer collects parallel bids behind the scenes. Your app calls the SDK once for each ad slot, supplies configured context such as format, placement ID, and available geo or user-segment signals, then receives a filled creative. The SDK call is unchanged between waterfall and RTB setups; the complexity sits in mediation.
Floor prices are the main RTB control. The floor is an auction reserve: bids below it are rejected, and the impression goes unfilled instead of selling below the price you consider worthwhile. Correct floor setting is a separate discipline covered in the eCPM Floors Guide.
Some demand partners do not support RTB. Others use reserved-price contracts or require a fixed waterfall position. A hybrid setup runs RTB partners in a concurrent first-price auction, then falls back to a traditional waterfall of non-RTB partners when no RTB bid clears the floor. This protects fill rate while adding yield where RTB competition is available.
The Waterfall Suits Simple Setups, Not Growing Inventory
Waterfalls still fit some small setups. With low impression volume and one or two demand partners, the yield difference from RTB narrows, and integration work may not justify switching. The break-even point depends on impression volume, current eCPM, and the number of partners you can maintain.
Concurrent bidding becomes more valuable as volume rises. At 500,000 impressions per month across multiple geos and formats, even a modest CPM gain accumulates. A studio using Adrop's beta saw rewarded video eCPM stabilize above earlier single-network averages after competing demand was added, within two weeks of switching, based on aggregate data from that early-access cohort. The game did not change. The auction did.
Maintenance is another consideration. Five networks in a waterfall need regular re-ranking as CPMs move. RTB needs floor management instead, which is generally less time-intensive because floors change more slowly than network rankings.
What Changes When You Switch
For developers, moving from waterfall mediation to RTB is operationally simpler than it may sound. Set floor prices instead of a priority order. Add partners by enabling them in the mediation dashboard, rather than placing them in a ranked list. The SDK call in your app remains the same.
The result is an auction that follows current demand instead of your latest manual update. A Friday evening rewarded video impression in Japan competes at the current rate for that inventory, not the price set during your last waterfall update. For studios managing several active titles with regular content updates, less waterfall maintenance is a practical benefit alongside higher yield.
RTB also makes floor strategy more important because the floor is your main yield lever. Floors set too high reduce fill rate; floors set too low give up part of the auction's market-clearing benefit. Finding the right balance requires experimentation; the eCPM floor guide covers floor calibration in detail, and the mediation layer includes tools such as floor A/B testing for that work.