Mediation Strategy Guide

How to choose between waterfall, in-app bidding, and hybrid mediation — and how to configure each for maximum revenue.

What Mediation Actually Does

When a user triggers an ad placement in your app, someone has to decide which ad network serves that impression. That’s mediation.

Without mediation, you’re locked to a single ad network. Whatever price they offer is what you get. With mediation, multiple networks compete for every impression, and the highest payer wins.

Mediation handles three things:

1. Network selection — which demand source serves each impression
2. Auction mechanics — how networks compete (waterfall, bidding, or hybrid)
3. Adapter management — keeping multiple network SDKs working together

The mediation strategy you choose determines how efficiently your inventory is sold. Get it right and you maximize revenue from the traffic you already have. Get it wrong and you leave money on the table every day.

Waterfall Mediation

How It Works

Networks are ranked in a fixed order based on historical eCPM performance. When an ad request fires, the mediation layer calls the top-ranked network first. If that network fills, the impression is served. If it doesn’t, the request falls to the next network in line.

Impression available
  → Call Network A (historical eCPM: $12)
    → No fill
  → Call Network B (historical eCPM: $8)
    → Fills at $8
  → Impression served at $8

The Problem

Network B served the ad at $8. But what if Network C — ranked third with a historical eCPM of $6 — would have bid $15 for this particular impression? In a waterfall, Network C never gets the chance. The request stopped at B.

Waterfalls use historical averages to order networks. But eCPMs fluctuate constantly based on advertiser demand, user demographics, time of day, and dozens of other signals. A static ordering based on last week’s averages is always out of date.

When Waterfalls Still Make Sense

• Networks that don’t support in-app bidding (some smaller networks still operate on a waterfall-only basis)
• Testing a new network where you don’t have historical performance data yet
• Legacy setups where migrating to bidding would require significant re-integration

Waterfall Optimization Tips

If you’re running a waterfall (even as part of a hybrid setup):

1. Re-order weekly based on actual performance, not setup-time assumptions
2. Use segment-level ordering — the best network in the US may not be the best in Brazil
3. Set floor prices per tier — the first network in the waterfall should have the highest floor
4. Limit waterfall depth — 3–5 tiers is sufficient. Each additional tier adds latency.
5. Monitor passback rates — if a network is passing back (not filling) more than 30% of requests, move it down or remove it

In-App Bidding

How It Works

All participating networks bid simultaneously in a real-time auction. The mediation layer sends a bid request to every network at once, collects bids, and awards the impression to the highest bidder.

Impression available
  → Request sent to all networks simultaneously
    → Network A bids $8
    → Network B bids $15  ← highest bid, wins
    → Network C bids $6
    → Network D: no bid
  → Impression served at $15

Why Bidding Wins

True price discovery. Every network sees every impression and bids its actual value. No network gets priority because of its position in a list. The result: the publisher always gets the highest price the market will pay.

No manual tuning. There’s no waterfall to order, no floor prices to set per tier, no weekly re-ordering. The auction handles optimization automatically.

Lower latency. Parallel bid requests are faster than sequential waterfall calls. One round-trip vs N sequential round-trips.

Fairer competition. In a waterfall, the top network has an unfair advantage — it sees every impression first. In bidding, every network competes equally.

Bidding Considerations

• Not all networks support it. Some smaller networks only work in a waterfall.
• Bid caching matters. Networks need to respond within a timeout window (typically 2–5 seconds). Slow responders may miss auctions.
• You lose some manual control. In a waterfall, you can force specific networks into specific positions. In bidding, the market decides.

Hybrid Mediation

How It Works

In-app bidding runs first as the primary auction. Networks that support bidding compete in real time. The winning bid then competes against a traditional waterfall of networks that don’t support bidding.

Phase 1: Bidding auction
  → Network A bids $12
  → Network B bids $8
  → Winning bid: $12

Phase 2: Waterfall (with $12 floor)
  → Network C (waterfall, historical eCPM $15)
    → Fills at $14 (above $12 floor) ← wins overall
  → Impression served at $14

OR

Phase 2: Waterfall (with $12 floor)
  → Network C (waterfall, historical eCPM $15)
    → No fill above $12 floor
  → Impression served at $12 (bidding winner)

Why Most Publishers Use Hybrid

Pure bidding would be ideal, but not every network supports it yet. Hybrid captures the benefits of bidding for networks that support it while still accessing demand from waterfall-only networks.

The bidding auction sets a dynamic floor for the waterfall. This means waterfall networks only win if they can beat the best real-time bid — a significant improvement over a static floor.

Hybrid Configuration Tips

1. Put all bidding-capable networks in the bidding auction — don’t keep them in the waterfall
2. Keep waterfall-only networks in the waterfall but ranked by recent performance
3. Let the bidding winner set the waterfall floor — this is the key mechanism that makes hybrid work
4. Review quarterly which networks have added bidding support and move them out of the waterfall

Comparing the Three Approaches

Floor Price Strategy

Floor prices are the minimum eCPM you’ll accept for an impression. Set them too high and you lose fill. Set them too low and you sell premium inventory cheap.

When to Use Floors

• In waterfall mediation — floors prevent lower-tier networks from filling at rock-bottom prices
• As a revenue protection mechanism — ensures you don’t serve impressions below a minimum acceptable value
• In geo segments where eCPMs are predictable — US/UK floors can be set more aggressively than emerging market floors

When NOT to Use Floors

• In pure bidding — the auction itself is the price discovery mechanism. Adding floors on top can reduce fill without improving price.
• When you don’t have enough data — setting floors before you understand your eCPM distribution is guessing
• Globally without segmentation — a $5 floor makes sense in the US but kills fill in Southeast Asia

How to Set Floors

1. Start with no floors. Run for 2 weeks and collect eCPM data by geo, format, and network.
2. Analyze the distribution. For each geo/format combination, look at the 25th percentile eCPM — this is the low end of what your inventory typically earns.
3. Set floors at or slightly below the 25th percentile. This cuts the bottom without losing meaningful volume.
4. Monitor fill rate. If fill drops more than 5%, your floor is too high.
5. Adjust monthly. eCPMs shift seasonally (Q4 high, Q1 low) and with market changes.

Floor Price by Geo (Starting Points)

These are rough starting points — adjust based on your actual data:

Network Selection

How Many Networks Do You Need?

Choosing Your First 3–5 Networks

Start with networks that cover your primary geos and formats:

1. Google AdMob — broadest demand, works everywhere
2. AppLovin — strong in gaming, high eCPMs in tier-1 markets
3. Meta Audience Network — precise targeting, high CPMs for certain demographics
4. One regional specialist — Mintegral for APAC, InMobi for India/SEA, Amazon APS for US e-commerce
5. One video specialist — Liftoff Monetize or Unity Ads for high-quality video demand

When to Add More Networks

Add a network when:

• Your fill rate is below 95% in a significant geo
• A new network offers strong demand in a geo or format where your existing networks are weak
• You’re seeing eCPM plateau — more competition can push prices up

Remove a network when:

• It’s filling less than 5% of impressions (overhead isn’t worth it)
• Its eCPM is consistently below your floor in every geo
• It’s causing latency issues or SDK conflicts

Geo-Based Configuration

Ad monetization performance varies dramatically by geography. US impressions can be worth 10–20x what the same impression earns in a lower-tier market. Your mediation configuration should reflect this.

What to Segment by Geo

• Floor prices — higher in tier-1, lower or none in tier-3/4
• Network priority (waterfall) — different networks perform best in different regions
• Ad frequency — consider showing more ads in low-eCPM geos (each impression earns less, so you need more volume)
• Format mix — rewarded ads work globally; interstitial tolerance varies by market

Geo Tier Framework

Network Strengths by Geo

Monitoring and Iteration

Mediation isn’t set-and-forget. Performance shifts constantly and your configuration should evolve with it.

What to Monitor Weekly

Seasonal Patterns

eCPMs follow predictable cycles:

• Q4 (Oct–Dec): Highest eCPMs of the year. Holiday advertiser spending peaks. Don’t make major config changes — let the revenue flow.
• Q1 (Jan–Mar): eCPM dip after holidays. Normal. Don’t panic-adjust floors.
• Q2–Q3 (Apr–Sep): Stable mid-range. Good time to test new networks and configurations.

A/B Testing Mediation Configuration

Don’t change your mediation config for 100% of users at once. Test changes on a subset:

• Split users 50/50 between configurations
• Run for at least 1 week (ideally 2) to account for daily variation
• Compare ARPDAU, not just eCPM — fill rate and impression volume matter too
• Only promote the winning config to 100% when results are clear

Network Neutrality

A mediation platform that operates its own ad network has a structural conflict of interest. It makes money both from mediation fees and from its own ad network’s revenue share. This creates an incentive to favor its own demand — even when another network would pay more.

How the Conflict Manifests

• Information asymmetry — the mediator knows what other networks are bidding, giving its own network a competitive advantage
• Bid adjustments — the platform can shade competing bids or inflate its own
• Default configuration — the platform’s own network may be pre-configured with advantages (higher priority, lower latency)
• Reporting bias — dashboards may emphasize the platform’s own network performance

What to Look For

Ask your mediation provider:

1. Do you operate your own ad network?
2. What percentage of impressions does your network win?
3. Can I see true bid-level data for all networks including yours?
4. Can I remove your network from the auction entirely?

If the platform’s own network consistently wins 40–50%+ of impressions, that’s a red flag. In a truly neutral auction with 10+ bidders, no single network should dominate to that degree.

Common Mediation Mistakes

1. Using a Single Network Without Mediation

No competition = no price discovery. Even adding one more network through mediation typically increases revenue 30–50%.

2. Setting Global Floor Prices

A $5 floor makes sense in the US but kills fill in India. Always segment floors by geo at minimum.

3. Never Updating Waterfall Order

Networks perform differently over time. If you set your waterfall order 6 months ago and haven’t touched it, you’re leaving money on the table.

4. Keeping Underperforming Networks

If a network fills less than 5% of impressions and has a below-average eCPM, remove it. Dead weight adds latency and complexity.

5. Ignoring Bid Density

If you have 8 networks but only 2 are bidding-enabled and the rest are waterfall, your auctions aren’t competitive. Move networks to bidding as they add support.

6. Over-Optimizing One Metric

Chasing the highest eCPM can hurt fill rate. Chasing fill rate can lower eCPM. Optimize for ARPDAU — the metric that captures both.

7. Trusting the Platform’s Own Network Blindly

If your mediation platform tells you their network wins 50% of auctions, verify independently. Run a test with their network removed and compare total revenue.

8. Changing Everything at Once

Adjust one variable at a time. If you change floor prices, network order, and ad frequency simultaneously, you won’t know what caused the result.

Getting Started

If You’re Setting Up Mediation for the First Time

1. Choose a neutral mediation platform — one that doesn’t operate its own ad network
2. Integrate 3–5 demand sources — start with AdMob + AppLovin + one regional specialist
3. Enable in-app bidding for all networks that support it
4. Set no floor prices initially — gather 2 weeks of baseline data first
5. Monitor daily for the first 2 weeks — watch eCPM, fill rate, and ARPDAU
6. Add floors after 2 weeks based on actual eCPM distribution by geo
7. Add more networks one at a time, measuring the impact of each

If You’re Migrating from a Waterfall to Bidding

1. Identify which of your current networks support bidding
2. Move bidding-capable networks into the bidding auction — don’t run them in both
3. Keep waterfall-only networks in the waterfall with the bidding winner setting the floor
4. A/B test the hybrid setup against your current pure waterfall
5. Monitor for 2 weeks — you should see higher eCPMs and less manual maintenance
6. Roll out to 100% when results are clear

Further Reading