You don't need a tool to do a basic ad-revenue audit. You need 45 minutes, the last 30 days of network reports, and this list. Run it quarterly and you'll catch the 80% of leaks that show up at every studio we've audited.
Section A — Floors (5 items)
- Floors set within the last 90 days? If the last edit on your top ad unit is older than 90 days, treat the floor as stale until proven otherwise.
- Country-level floors set for Tier 1 vs Tier 2/3? If one floor covers all geos, you're almost certainly leaving 10–25% on the table.
- Format-level floors not copied across formats? Banner and rewarded should not share a floor. Interstitial sits between them.
- Floor near the 80th-percentile clearing price? Pull a 30-day eCPM distribution for the ad unit. Your floor should be near the 80th percentile, not the 50th.
- No "experimental" $0 or $999 floors lingering? Common after A/B tests. Trivial to fix once spotted.
Section B — Bidders / waterfall (3 items)
- Every bidder enabled is currently bidding? Pull a 7-day report showing impression counts per bidder. Any bidder with sub-1% impression share on its enabled units is either misconfigured or dead.
- No paused bidders that no one remembers pausing? Re-check the dashboard for "paused" or "off" toggles, especially on Meta, TikTok, and Pangle. Every paused day is 5–15% lost demand on that ad unit.
- Network count between 4 and 8? Below 4 means you're leaving demand uncaptured. Above 8 means you're paying operational tax with no marginal revenue.
Section C — Demand quality (2 items)
- JP / US / KR eCPM each tracked separately? These three are disproportionately high-value. If your reporting blends them with ROW, you're flying half-blind.
- Direct deals (if any) actually filling? If you sold a $10 CPM line and it's filling at $3 effective, either the deal collapsed or the ad unit mapping is wrong.
Section D — Ad format mix (2 items)
- Rewarded video impression share matches design intent? If you designed for 1.5 rewarded video views per session and the data says 0.3, players aren't engaging with the placement. That's a UX problem with revenue consequences.
- Interstitial frequency capped sanely? Frequency caps drift after updates. Pull average interstitials per session and compare to your published spec.
How to act on what you find
Each item that fails is one row in a spreadsheet with three columns: What(the symptom), Estimated impact ($/day, even rough), and Fix difficulty(10 min / 1 hour / blocked on network rep). Sort descending by impact, divide by difficulty, and you have your priority list for the next 14 days.
What this audit doesn't catch
Cross-app patterns, geo-mix drift inside a single country, and bidder-versus-bidder cannibalization need either a multi-app dashboard or an LLM that's seen thousands of these patterns. That's what we built Mediation One for. But you don't need it to run this 12-point check this quarter — and you should, regardless of whether you ever talk to us again.