How do you know someone is still running a manual waterfall? Nobody has to ask, they’ll tell you every Monday morning when they’re updating eCPM floors.

Logging into the dashboard, shifting network positions, guessing at Tier 2 pricing, hoping the week plays out favorably. For years, that was just the job.

Advertisers moved. Budgets shifted into performance-based, real-time environments. Networks optimized their algorithms for publishers running automated setups. And the manual waterfall quietly became the thing costing publishers 15 to 30 percent of recoverable ad revenue every month.

UndrAds ran the math and a publisher at $50K monthly in ad revenue is likely missing out on $7,500 to $15,000 through undersold impressions and stale eCPM floors alone.

This guide covers what the migration actually looks like, what to watch for, and how to structure the transition without destabilizing existing revenue. If the broader monetization picture needs a refresher first, this guide to monetizing mobile apps is a solid starting point.

Why Manual Mediation is a Structural Problem

The waterfall ranks ad networks sequentially based on historical eCPMs. A network at position three might be willing to pay more than position one on any given impression but never gets the chance to bid. Publishers accept a lower price because of how a queue was configured months ago.

The operational cost compounds the revenue problem:

A single title with geo-segmentation can generate hundreds of waterfall line items requiring active management
eCPM floors go stale in days during volatile demand windows
Every new game added to a portfolio multiplies the manual workload proportionally

Ad ops becomes a bottleneck the moment a studio crosses 100K DAU. Most studios feel this as a resourcing problem. The real diagnosis is an infrastructure one.

Manual Mediation vs. Automated Ad Operations

Dimension	Manual Mediation (Waterfall)	Automated Ad Operations
Pricing mechanism	Historical eCPM floors set by the publisher	Real-time auction with networks bidding per impression
Competition	Sequential, one network at a time	Simultaneous, all demand sources bid at once
Revenue yield	Frequently undersells available inventory	Maximizes yield through live competition
Latency	High, cascades through multiple calls before fill	Low, single auction resolves in milliseconds
Operational overhead	Constant manual tuning required	Algorithm manages placements automatically
Geo/segment complexity	Multiplies workload with every new market	Handled dynamically by the platform
Fill rate management	Requires manual balancing across networks	Automated, with configurable fallback layers
Transparency	Limited, pricing opaque across networks	Real-time win/loss signals with full reporting
Adaptability	Responds to weekly or biweekly adjustments	Adapts to market conditions continuously

If most of the left column describes the current setup, the revenue drag is already active.

What “Automated” Actually Covers

Three distinct models sit under the umbrella:

In-App Bidding (IAB): All demand partners receive the ad request simultaneously and compete in a real-time auction. No sequential calls, no cascading delays. In-app bidding now accounts for roughly 80% of the typical publisher’s monetization setup. Full breakdown here.

Automated Waterfall Optimization: Algorithms dynamically reorder and reprice waterfall line items based on live performance data. Tools like Bidlogic handle this end-to-end, covering eCPM floors, mediation groups, and A/B test automation.

Hybrid Models: The industry standard for 2026. Strong bidders (Meta, AppLovin, Mintegral, Unity) compete in a real-time auction first. When the winning bid lands below a set threshold, the system routes to a waterfall fallback. Captures bidding upside while protecting fill rates in lower-tier geos.

For most publishers migrating from manual setups, the hybrid model is the right destination architecture.

Migration Checklist

Phase 1: Audit and Baseline (Weeks 1-2)

Know exactly what you have and what “better” means before touching anything.

Document waterfall configuration across all ad units: rewarded video, interstitial, banner
Export 90 days of eCPM, fill rate, and ARPDAU data by geo, network, and ad format
Identify which networks in the current stack support in-app bidding
Map revenue by network to surface dependency concentration risk
Define success criteria before migrating (example: 5% ARPDAU improvement within 30 days)

Phase 2: Platform and Partner Selection (Weeks 2-3)

Choose the stack that fits the UA strategy and geo footprint before writing a single line of config.

Select a primary mediation platform aligned to the UA strategy and game engine
Confirm bidding adapter availability for every top-revenue network
Generate API and reporting keys for each bidding partner
Assess whether the platform supports automated waterfall tools natively, or if a third-party layer is needed
Verify SDK compatibility with the game engine

Phase 3: Technical Setup (Weeks 3-4)

Build the bidding layer and hybrid fallback correctly the first time.

Integrate bidding adapters for each participating network
Build mediation groups per ad unit with two layers: bidding sources and waterfall fallback
Set initial floor prices anchored to 90-day historical data
Configure separate geo-specific mediation groups for Tier 1, Tier 2, and Tier 3 markets
QA ad serving end-to-end in a staging environment before going live

Phase 4: A/B Test and Validate (Weeks 4-6)

Prove the setup on a controlled traffic slice before committing the full portfolio.

Route 20-30% of traffic to the new bidding setup, keep the existing waterfall for the remainder
Run for a minimum of two weeks to capture day-of-week variance
Monitor ARPDAU, eCPM, fill rate, and latency daily
Track ANR rates for any uptick pointing to SDK conflicts
Review win rates per network to validate genuine bidding competition

Phase 5: Full Rollout and Optimization (Week 6 onward)

Scale confidently and shift ad ops time from maintenance to strategy.

Shift 100% of traffic once A/B results confirm positive movement
Keep waterfall fallback active for Tier 3 geos through early weeks
Set floor prices to guard against eCPM erosion over time
Schedule monthly reviews of network performance and demand partner mix
Set automated alerts for fill rate drops below defined thresholds

Tools Worth Knowing

MAX by AppLovin: Market leader in 2025. AppLovin ties ROAS campaign access to MAX mediation, making it a practical requirement for studios running AppLovin UA. Setup guide here.

Unity LevelPlay: Strong auto-optimization algorithm, tightly integrated with Unity workflows. Credible option for publishers without AppLovin UA dependency. See full platform comparison.

Google AdMob: Essential for Google demand. Over 97% of AdAge’s top 100 advertisers buy through AdMob. Best used as a component within a broader stack. See the top AdMob alternatives worth considering alongside it.

Appodeal: Managed mediation with built-in bidding optimization. Good fit for teams that want to delegate stack management entirely.

Bidlogic: Purpose-built automation layer for waterfall and bidding management. Covers eCPM floors, mediation groups, and A/B test automation in one platform.

Priority bidding networks to activate first: Meta Audience Network, AppLovin, Mintegral, Unity Ads, ironSource.

Pitfalls Worth Preparing For

Going bidding-only without testing: Bidding outperforms waterfalls in Tier 1 markets. In Tier 3, cached waterfall ads frequently win on eCPM due to connectivity constraints and lower programmatic demand. A global bidding-only rollout often hurts the markets where the waterfall was working.

Fix: Default to hybrid. Reserve bidding-only for Tier 1 markets with validated competition.

Missing eCPM erosion: Demand partners bid aggressively when entering a new auction environment. eCPMs look strong in weeks one through four. As networks optimize to protect margins, bids decline. Publishers who skip ongoing monitoring mistake the early spike for a permanent gain.

Fix: Set floor prices from day one. Rotate demand partners regularly. Run a small bidding-only cohort as an early-warning signal for drift.

Migrating all traffic at once: A full cutover eliminates the control group. There is no way to attribute performance changes to the migration, and no clean recovery path if something breaks.

Fix: A/B test first. 20-30% of traffic, two weeks minimum.

Treating every geo the same: Tier 3 markets have lower programmatic budgets, different top networks, and higher reliance on cached inventory. A Tier 1-optimized setup applied globally underperforms a well-maintained waterfall in those markets.

Fix: Separate mediation groups by geo tier from day one.

Expecting automation to run itself: Floor price calibration, demand partner relationships, seasonal adjustments, and network quality controls still require active oversight. AI agents are improving how publishers handle ad ops but strategic decisions stay with the team.

Fix: Redirect ad ops time from waterfall updates to demand strategy and ARPDAU analysis.

Underestimating SDK weight: Multiple bidding adapters increase SDK size and initialization load. In performance-sensitive titles, the result can be ANR events that damage session metrics and store ratings.

Fix: Audit SDK weight before adding adapters. Test for ANRs explicitly in QA.

Where to Start

The mobile ad ecosystem is consolidating around programmatic, real-time buying. Publishers on fully manual waterfall setups are progressively less competitive for that demand regardless of inventory quality. The gap between what automated setups earn and what manual ones earn widens every quarter.

The migration is a sequencing question. A phased approach with a defined baseline, a hybrid target architecture, and active monitoring through the transition delivers the upside of automation without a high-stakes single bet.

Start with a 90-day data pull. Define the baseline. Pick the platform. Move 20% of traffic and measure for two weeks.

Not sure where the current setup is losing revenue? UndrAds can run the numbers for you.

How to Migrate from Manual Mediation to Automated Ad Operations

Rashmita Behera