Make-good Guarantee
The Dynamic Ad commitment that every 90-day engagement will outperform the client's trailing-90 baseline, or the underperformance is published with a written explanation.
Common questions
Common questions
- What does the make-good guarantee mean exactly?
- It means every 90-day engagement either beats the client's trailing-90 baseline on the primary engine metric or triggers a published explanation of why it did not. The threshold is a documented Bayesian posterior (greater than 80% probability of improvement, point estimate at least 5% better than the matched baseline).
- What happens when the make-good threshold is not met?
- A publish-why post ships. The post names the engagement (anonymized by default), the matched-baseline figure, the engagement-period figure, the confounders accounted for, and the hypothesis tested. The operator has the right to redact identifying detail; nobody has the right to block the post.
- Where is the make-good methodology defined?
- At /methodology. It defines baseline construction (trailing-90 frozen snapshot), channel-mix matching, three confounder corrections (seasonality, promotional cadence, attribution-model shifts), the threshold for better, the trigger for publish-why, and the two-operator countersignature gate.
The make-good guarantee ships on every engagement letter: better than your trailing 90 days, or we publish why. The threshold is not a soft promise; it is a documented Bayesian posterior on the primary engine metric (cost-per-qualified-acquisition for Lead OS, contribution-margin ROAS for Commerce OS), measured against a frozen baseline snapshot captured at engagement start.
When the threshold is not met, the post is not optional. The publish-why trigger fires when the posterior probability of improvement is under 50%, or the primary metric regresses by 10% or more versus the matched baseline. The post names the engagement (anonymized to industry plus spend tier unless the operator consents to identification), the matched-baseline figure, the engagement-period figure, the confounders accounted for, and the hypothesis tested. No fabrication, no excuse, no retroactive baseline tuning.
The methodology is published at /methodology and signed against every engagement. Two operators countersign every cycle-close number before publication. The make-good is structurally bound to the operating record by the approval gate; no published figure has ever been computed by an algorithm alone and shipped without two human sign-offs.
Example
A Commerce OS engagement closes its 90-day cycle. The matched-baseline contribution-margin ROAS was 2.4×; the engagement-period figure is 3.1×. Bayesian posterior on improvement is 87%, point estimate 29% better than baseline. Both thresholds clear. The cycle-close report ships with both numbers, the matching schema, and the names of the two operators who signed. If the posterior had landed at 45%, the publish-why post would have shipped instead.
Always on · Algorithm-led · Human-approved
Walk this operation against your account.
Thirty focused minutes with one of the founders. We map Make-good Guarantee against your stack and tell you which entry point fits your stage.