Field notes · 21 Apr 2026
Fill rate and eCPM are not a trade-off you “pick”
Product reviews still ask teams to “choose” fill or eCPM, as if they were menu items. They are not. They are two readings of the same week, joined by inventory quality, geo mix, format, and whether the SDK finished its work before the user left. Fill Rate Diagnostics exists because that fake choice keeps shipping.
Price without volume is a costume
eCPM can rise because fewer, better-matched impressions remain. That can be healthy. It can also be a timeout dressed as premium. If fill fell in the same hours, you do not have a strategy. You have a narrower window. Write both numbers on the same line of the brief or do not write the brief.
Volume without price is not automatically “user friendly”
High fill with collapsing price often means you opened the filter too wide or you are filling with inventory that does not match the session. Users still saw ads. Finance still saw a weaker day. Calling that a win for “experience” is a story, not a measurement.
Geo mix will fake a trade-off
A week that adds more Thailand evening traffic next to more US lunch traffic will move fill and eCPM in directions that look like a choice. They are not. They are two audiences. Split the rows before anyone “optimizes.” SEA operators at the Bang Kapi desk are tired of being told their fill is a setting they forgot to pick.
A diagnostic order we actually use
- Confirm the session graph did not move first.
- Split fill by country, format, and hour.
- Check SDK timeouts and crash-free rate in the same window.
- Only then talk about floors or bidding density.
If step one already explains the week, stop. Silence is an acceptable conclusion in Ad Monetization App Analytics. It is also the one that most slide templates refuse to allow.