Definition
Ratio of actual margin outcome when a merchant overrides an AI pricing recommendation to the projected margin outcome if the AI recommendation had been followed.
The Measurement Gap
Merchants override AI pricing recommendations regularly, but organizations rarely measure whether those overrides create or destroy value. The override rate (percentage of AI recommendations rejected) is commonly tracked but tells nothing about outcomes. A 30% override rate could represent experienced merchants correcting a flawed model, or it could represent merchants ignoring a superior system based on habit and gut feel. Revenue and margin metrics capture total performance but cannot decompose how much value was added or destroyed by override decisions versus AI-followed decisions. Price compliance metrics measure adherence to the AI's recommendations but frame overrides as failures rather than evaluating their economic merit. Conversion rate after price changes captures customer response but not the counterfactual: what would have happened if the AI's price had been used instead. The missing measurement is a direct comparison of override outcomes to the AI's projected outcomes, tracked per merchant and per override reason, so the organization can make evidence-based decisions about override authority rather than debating it based on opinion.
Decision Supported:
Are merchant overrides of AI pricing making the organization more or less money? Should the override authority be expanded, constrained, or retrained?
Formula
Plain language: When a merchant overrides the AI price, track the resulting sales volume and margin for that item over the following N days. Compare to the actual margin of similar items in the same category where the AI recommendation was followed during the same period. Did the human do better or worse than what the AI actually delivered on comparable items?
Notation:
POVS = Margin_override_actual / Margin_control_actual
Variables:
| Variable | Definition | Unit |
|---|---|---|
| POVS | Pricing Override Value Score | Ratio |
| Margin_override_actual | Actual margin realized on the item during the measurement window after the merchant set the override price | $ |
| Margin_control_actual | Actual margin realized on matched control items (same category, similar velocity, similar price tier) where the AI recommendation was followed during the same window | $ |
| Measurement window | 7 to 14 days post-override, matched to the pricing engine's optimization horizon | Days |
Secondary method (convenience check): POVS_quick = Margin_override_actual / Margin_AI_projected, where Margin_AI_projected is the pricing engine's forecasted margin at its recommended price. This is faster to compute but biases the denominator if the AI's projection model is systematically optimistic or pessimistic. Use the matched-control method for any analysis informing override policy decisions.
Proposed formula, pending validation pilots.