Definition
The range of price architectures tested relative to the current price, measuring whether the pricing AI is exploring broadly enough to find the true optimal price or converging prematurely on a local maximum within a narrow band around the status quo.
The Measurement Gap
The Gap
Pricing AI finds the "optimal" price. But optimal within what range? Most pricing algorithms explore within a narrow band around the current price -- typically plus or minus 10-15%. Within that band, the algorithm tests efficiently and finds the best price. But it never tests whether a completely different price architecture (50% higher with different positioning, or 30% lower with volume strategy) would produce better total margin. The algorithm converges on a local maximum and reports it as the optimum.
Revenue at tested prices tells you which tested price produced the most revenue, but it cannot tell you whether untested prices would have produced more. Conversion rate by price point maps the response curve within the tested range but says nothing about the response curve outside it. Profit maximization models find the peak within the data they have but cannot identify peaks in regions they never explored. Elasticity estimates from narrow-range testing extrapolate poorly to wide-range pricing decisions because elasticity is not constant across the full price spectrum.
The structural problem is the explore-exploit tradeoff. Pricing AI is rewarded for exploiting what it knows (pricing within the proven range) and penalized for exploring what it does not know (testing prices that might produce short-term revenue decline). Over time, the algorithm's exploration range narrows as it becomes more confident in its local optimum. No standard metric measures whether the exploration range is too narrow to find the true optimum.
The Decision
We propose PEBI to answer: Is our pricing AI finding the best price, or just the best price within 15% of where we already are? This metric would measure the tested price range as a percentage of current price, and compare AI-optimal outcomes to deliberately-placed "wild card" tests outside the AI's self-selected range.
Formula
Plain Language
Look at the full range of prices the AI tested in a given cycle. Subtract the lowest tested price from the highest and divide by the current price. That percentage is PEBI -- the breadth of exploration. Then, separately, run a "wild card" test: have a human pricing analyst deliberately place test prices outside the AI's self-selected range. Compare the best-performing wild card to the AI's optimal price. If the wild card produces materially better margin, the AI was stuck in a local maximum.
Notation
PEBI = (Price_max_tested - Price_min_tested) / Price_current x 100
Exploration Gap:
EG = Margin_wild_card_best / Margin_AI_optimal - 1
Variables
| Variable | Definition | Units | Source System |
|---|---|---|---|
| Price_max_tested | Highest price point tested by the pricing AI in the measurement cycle | Dollars | Price Testing Platform |
| Price_min_tested | Lowest price point tested by the pricing AI in the measurement cycle | Dollars | Price Testing Platform |
| Price_current | Current standard price (baseline for range calculation) | Dollars | Pricing Engine |
| Margin_wild_card_best | Best contribution margin achieved by a human-selected wild card test price outside the AI's range | Dollars | Tagged Wild Card Tests (Integration) |
| Margin_AI_optimal | Contribution margin at the AI's recommended optimal price | Dollars | Pricing Engine |
Proposed formula, pending validation pilots.