Definition
Percentage of AI-triggered merchant fund holds that were retrospectively justified by actual risk events within a 90-day validation window.
The Measurement Gap
The Measurement Gap
No metric tells you whether automated merchant fund holds are accurate. The number of holds placed is tracked as an operational volume metric. The financial impact of holds (funds withheld, duration, merchant communication effort) is tracked as a cost metric. But the accuracy of holds (whether the merchants subjected to fund holds were genuinely risky, and whether the hold was justified by subsequent events) is not measured by any standard payment metric.
This is the payments equivalent of a false arrest problem. A merchant fund hold freezes settlement funds based on an AI behavioral scoring model's prediction that the merchant poses elevated risk. If the prediction is wrong, the hold destroys the merchant relationship, blocks the merchant's cash flow, generates support escalations, and may drive the merchant to a competing acquirer. The merchant attrition rate captures the eventual outcome (merchants leave) but no metric captures the cause: that the automated hold was wrong.
The gap is economically significant because merchant acquisition costs are high (months of lost revenue during onboarding, underwriting effort, compliance review), and a single false hold can destroy a merchant relationship that took significant resources to establish. False holds also create reputational risk when merchants share their experiences with industry peers, potentially deterring future merchant acquisition.
Without measuring hold precision, organizations cannot determine whether their AI risk models are genuinely protecting the portfolio or generating a stream of false alarms that destroy merchant relationships faster than new merchants can be acquired.
Decision Supported
What percentage of AI-triggered merchant fund holds are justified by subsequent risk events, and is the hold system protecting the portfolio or destroying merchant relationships?
Formula
Plain Language
Count the number of AI-triggered holds where the merchant subsequently experienced a qualifying risk event (chargeback ratio exceeding card network thresholds, confirmed fraud, regulatory action, or involuntary closure) within 90 days. Divide by total automated holds to get the precision rate.
Notation
MHPR = Holds_with_subsequent_risk_event / Total_automated_holds × 100
Variables
| Variable | Definition | Unit |
|---|---|---|
| Holds_with_subsequent_risk_event | Holds where the merchant experienced a qualifying risk event within 90 days: chargeback ratio exceeding 1% of transactions (Visa/Mastercard threshold), confirmed fraud, regulatory action, or involuntary closure | Count |
| Total_automated_holds | All AI-triggered merchant fund holds in the measurement period | Count |
Proposed formula, pending validation pilots.