Media agencies built their margin around a service that is now being automated. For two decades, the complexity of buying and placing advertising across channels justified significant fees and opacity. That complexity is no longer a barrier — AI platforms handle it natively and at cost.

The disruption is not coming. It has already moved into agency economics. What it exposes is a gap: the operational layer generated revenue without necessarily generating brand value. Plumb was built to fill that gap — to score what the machines can't price, and make the capital decision defensible before the IO is signed.

The holdcos' accountability gap is not a payment-model question. Every compensation structure produces a version of the same conflict: the agency's economic incentive is not fully aligned with the brand's media outcome. Plumb is structurally upstream of all three.

Every audit carries a number. Every number resolves at flight close against the KPI targets named at the time of the verdict. The figures below are illustrative of the calibration record the system builds over time — each closed audit logged against its verdict, outcomes visible to the brand. Calibration is falsifiable and improves with every audit added to the corpus. The holdco shows you the wins. Plumb shows you the wins, the misses, and the process breaches.

Plumb's scoring treats each metric as a signal with two properties: amplitude — how strong the signal is — and a confidence weight based on the quality of the data source behind it. The five metrics in each layer produce a weighted composite score. The three layer scores are not averaged. They operate as compound gates: all three must clear their thresholds simultaneously for a COMMIT verdict to issue. A strong Optimization score does not override a failed Authenticity score. Each gate clears independently.

The core question Plumb is answering is a probability question: given every signal measurable at the time of the brief, what is the likelihood this specific buy produces the outcome the brand declared? That is a different question from "is this buy efficient?" Efficiency is what the bid stream measures. Probability of outcome is what Plumb scores.

The weighting construct is explicit. Optimization metrics are sourced from third-party data and weighted at full confidence where attribution is closed-loop. Authenticity metrics carry lower automatic confidence on the two human-review cases — Cultural Truth and Moment Respect — and those weights are not finalized until a named buyer has reviewed the data and logged their reasoning. Pricing metrics are arithmetic once the brief fields are complete. All weights and thresholds are documented and available to the brand at audit close.