What is a retained outcome?

The portion of an engagement's measured improvement that survives seasonality, promotional, and attribution-model-shift confounder correction. The make-good guarantee runs on retained outcome, not on platform-reported lift.

Why correct for confounders at all?

Because platform-reported lifts often vanish under seasonality or promo confounding. Paying the make-good on an uncorrected lift would mean paying on an outcome that did not actually happen, which the operator cannot defend at cycle close.

Where is the correction methodology documented?

At /methodology. The six-section page defines the baseline construction, the mix matching, the three confounder corrections (seasonality, promotion, attribution-model shift), the threshold for better, and the two-operator countersignature gate.

Retained Outcome

The portion of an engagement's measured outcome that survives confounder correction and qualifies for the Dynamic Ad make-good comparison.

Common questions

What is a retained outcome?: The portion of an engagement's measured improvement that survives seasonality, promotional, and attribution-model-shift confounder correction. The make-good guarantee runs on retained outcome, not on platform-reported lift.
Why correct for confounders at all?: Because platform-reported lifts often vanish under seasonality or promo confounding. Paying the make-good on an uncorrected lift would mean paying on an outcome that did not actually happen, which the operator cannot defend at cycle close.
Where is the correction methodology documented?: At /methodology. The six-section page defines the baseline construction, the mix matching, the three confounder corrections (seasonality, promotion, attribution-model shift), the threshold for better, and the two-operator countersignature gate.

An engagement closes a 90-day cycle and the platform-reported number is X% better than the trailing-90 baseline. The make-good guarantee does not pay out on X% directly. It pays out on the portion of X% that survives confounder correction.

Three corrections are applied: seasonality (same-month-prior-year baseline, multi-year average where possible), promotional cadence (engagement-period promo windows matched to trailing-period promo windows), and attribution-model shift (deduplicated server-side conversion source where the trailing period has it). The surviving figure is the retained outcome.

Retained outcome is the only figure the make-good claim runs on. Platform-reported lifts that vanish after seasonality correction do not count; lifts that survive every correction do. The conservative posture is structural: the brand commits only to what it can defend.

Example

A Commerce OS engagement closes its first cycle. Platform-reported ROAS up 38%. Seasonality correction: -8 points (Q4 holiday confound). Promotional correction: -4 points (the engagement period ran more frequent promo than the trailing period). Attribution-model correction: -3 points (server-side deduplication recovered conversions the platform had double-counted). Retained outcome: +23% on the matched baseline. That is the figure the make-good runs against.

Always on · Algorithm-led · Human-approved

Walk this operation against your account.

Thirty focused minutes with one of the founders. We map Retained Outcome against your stack and tell you which entry point fits your stage.

Book a 30-minute call →Or write us