Method

The Datatracebase Channel Score

A four-part judgement, written so a UA lead and a finance partner can disagree in public without hiding behind a composite index.

Operators working together at screens during an evening review
Critique night — scores are argued out loud, not averaged in a black box
RRetain

Day-7 and Day-14 retained value versus source cost.

CCover

ATT consent mix and postback completeness as quality defects.

DDecay

Week-three quality drop separated from creative fatigue.

SSurface

Store listing, referral, and brand search on the same ledger.

What the score is not

It is not a 0–100 vanity index. It is not a replacement for your MMP. It does not invent incrementality where you have no holdout. When coverage is weak, the score says “unknown,” which is more honest than a false precision CPI rank.

Thai iOS mixes in particular often sit well below global averages. The method treats that as a channel property, not as a reason to ignore iOS when Android looks tidy.

How a weekly review uses it

Each source gets a short narrative: what moved, which weight changed, and whether the change is coverage, creative, or audience. If two sources share an audience, the ledger says so instead of double-counting retained users.

Teams that finish Scoring Studio leave with this operating rhythm, not with a slide template to ignore.

See programmes that teach the score Ask the desk about fit