28 April 2026 · Desk editorial

Scoring channels when ATT consent stays uneven

Several Thai iOS mixes we review still sit between the mid-teens and high twenties on ATT consent, depending on category. Android CPI, meanwhile, looks calm enough to dominate the weekly meeting. If your channel score follows CPI, iOS becomes a rounding error. That is how leaked budgets hide.

Datatracebase treats coverage as a quality defect. A network that cannot return SKAN postbacks for a meaningful share of iOS traffic is not “unmeasurable”; it is lower quality until coverage improves. Unknown is a score, not an invitation to scale.

A practical weight, not a sermon

In Scoring Studio we do not prescribe a universal ATT multiplier. We ask you to publish a rule: below a stated consent threshold, the iOS portion of a source cannot outrank a better-covered source on CPI alone. The threshold is yours. The requirement is that it is written, dated, and visible to finance.

When consent jumps after a prompt redesign, the score can move up. When it falls after an OS change, the score moves down the same week — not after the quarter closes.

What this does not fix

Better weighting does not restore lost identifiers. It also does not make modeled conversions into ground truth. If your MMP’s modeled row is the only iOS story you have, the ledger should say “modeled” in the same cell as the score. Quiet footnotes are how blended ROAS returns through the back door.

Read the method overview on the Channel Score page, then the coverage module inside the Studio syllabus.