I Turned Off My AdMob Floors and Revenue Collapsed — The Mediation Rules That Actually Worked Solo

Make 80% match rate the investigation boundary

For the daily health check, I watch match rate as one indicator. In my operation, any group whose match rate falls below 80% becomes an investigation target.

• If the floor is causing the low match rate, lower the floor
• If the floor is off and match rate is still low, that is not a floor problem but a demand shortage (a region or format problem)

Being able to make this split avoids the reflexive "just lower the floor" response. For a group at 94% match rate with a high floor ratio, I sometimes deliberately hold. Groups with fewer than 20 impressions a month have thin data, so it is also important to skip them rather than be jerked around by the numbers.

Cap each change at +/-30%

When I move a floor, I keep each change within +/-30% of real eCPM. Doubling or halving in one step disrupts mediation's learning and makes it impossible to evaluate whether the change was good.

# Floor adjustment decision flow (pseudocode)
ratio = floor / ecpm_30d          # floor ratio
if match_rate < 0.80:
    if floor_is_on and ratio > 0.60:
        floor = ecpm_30d * 0.55   # too high -> 55% of real
    else:
        investigate("demand")     # suspect demand shortage
elif ratio < 0.40:
    floor = min(floor * 1.30, ecpm_30d * 0.55)  # raise, capped at +30%
else:
    hold()                        # ratio 50-60% -> HOLD

Just running this loop — "look at the ratio, stay within +/-30%, converge toward 50-60%" — quietly keeps floors from drifting. Operations that avoid chaos beat flashy optimization.

Handle the end of Unity Ads waterfall

The big one entering 2026 was that Unity Ads waterfall support in AdMob mediation ended on January 31, 2026. This pushes a move to bidding; the waterfall slots structurally stop serving.

In my case, I was slow to notice and kept Unity Ads waterfall rows across all 20 Android groups. I only realized something was wrong when about 85,000 requests over seven days returned almost zero fill, then investigated and learned support had ended. As a fix, I stopped the waterfall rows in every group and switched to monitoring Unity Ads on bidding only.

The lesson: mediation is not "set it once and you are done"; you have to keep tracking changes to each network's delivery model. Zero delivery means zero revenue, yet it is hard to notice when you only look at dashboard totals. You need a habit of regularly viewing fill by network and by format. The details are in the official AdMob Unity Ads mediation integration docs.

Reduce it to a twice-monthly, 15-minute review

Finally, I turn these judgment rules into a recurring task. Twice a month, I review floors with this procedure — about a minute per group, roughly 15 minutes total.

1. Open "ad activity (mediation)" in AdMob reports over the trailing 30 days
2. Set dimensions to app + mediation group, with impressions, match rate, and estimated revenue as metrics
3. Extract groups below 80% match rate
4. Check each group's floor ratio (floor / real eCPM)
5. If the ratio is over 60% with falling match rate, adjust the floor toward 55% of real, within +/-30%

Just running these five steps keeps the floors of 42 groups from diverging from reality. Rather than launching one flashy optimization, never skipping the plain monthly inspection keeps indie ad revenue far more stable. Some months I resist the urge to touch the floors at all and just read the ratios, and that restraint is often what works best.

At the start of next month, try simply extracting only the groups below 80% match rate. That alone makes it clear where you need to act right now.