Cohort quality versus volume: a scorecard for weekly UA reviews

Most weekly UA reviews still open with the same three numbers: spend, installs, blended CPI. All three are easy to pull and update in real time, and none of them tell you whether the cohort you just bought will pay back. Sensor Tower's State of Mobile 2026 put a figure on the shift teams already feel. Install volume is softening while IAP revenue keeps climbing, a signal the report reads as studios trading new-user volume for lifetime-value expansion, and a scorecard built around installs is measuring the wrong side of that trade.

Building a cohort quality scorecard

A cohort quality scorecard doesn't replace the spend and install numbers. It sits alongside them and forces the meeting to look at a cohort by age rather than by campaign name.

The first column is cohort size and source, split by network, then by paid versus organic. Sensor Tower's Feb 25 State of Gaming report found mobile IAP installs down roughly 7% year on year even as revenue rose about 1%, which means a smaller cohort can still be a better one; reviewing size alongside quality stops a shrinking top-line number from reading as failure on its own.

The second column is retention at D1, D7, D30, tracked against the cohort's own history rather than a single target. A network that delivers strong D1 and weak D7 is buying attention, not intent. Flag any cohort where D7 retention drops more than the trailing four-week average for that source, because that is the earliest place volume masks a quality problem.

The third column is blended ROAS at D7 and D30, not day zero. Day-zero ROAS rewards spend that front-loads low-value purchases and penalises campaigns that attract players who convert later, so put both windows on the same row and let the meeting see whether a channel is fast and shallow or slow and durable.

The fourth column is payer conversion rate by cohort, separated from average revenue per payer. A channel can hold flat ARPPU while conversion quietly erodes. That's a media-buying problem rather than a monetisation one, and both get missed when they're folded into a single blended number.

The fifth column is a creative fatigue signal, whatever the team already tracks for it: CTR decay, IPM trend, frequency against unique reach. AppsFlyer's State of Gaming for Marketers 2026 noted top advertisers now producing 2,400 to 2,600 creative variations a quarter, a 25 to 30% increase year on year, and that volume only pays off if the review catches which concepts are already decaying before spend follows them down.

The last column is a plain-language note: what changed since last week, and what the team is doing about it. Five quantitative columns with no narrative beside them is just another dashboard nobody reads.

Getting the room to use the scorecard

Getting a room to actually use the scorecard is a separate problem from building it. The instinct in most weekly reviews is to defend the headline number first (spend and installs against target) and reach cohort quality only if time allows. Reorder the agenda. Open with the scorecard and let spend and installs follow as context for why a cohort looks the way it does, because a channel that missed its install target but is holding D30 ROAS above the portfolio average is a channel worth defending rather than cutting, and a room that only sees the miss will make the wrong call before the quality data even gets a hearing.

Be explicit about what the scorecard won't do. It won't tell a team which creative concept caused a retention dip, only that one happened; the fatigue column is a flag rather than a diagnosis, and chasing the actual cause still needs a proper creative-level breakdown. It won't replace incrementality testing either. A cohort can look excellent on every column here and still represent spend that would have converted organically. The scorecard is a triage tool for the weekly cadence, built to catch problems and durable wins early, not a substitute for the slower and harder measurement work that confirms whether a channel is adding real value.

Don't try to backfill all five columns in the first week. Start with retention and blended ROAS at D7, since most stacks already expose both, and add the rest once the room is reading those two against each other without prompting.

None of this needs new tooling. Most MMPs and BI layers already expose retention and ROAS windows at the cohort level, along with payer conversion; the change is in what gets printed at the top of the weekly deck. Put the scorecard above the spend summary, not below it, and the review starts with the question that matters: is this cohort worth another dollar, not just the last one.

Related archive reading

These articles provide related context and remain subject to their stated review status.

• A Cohort Quality Scorecard for the Weekly UA Review
• ROAS targets versus payback windows