Price Test Contamination Rate (PTCR)

GMI-MKT-PPO-012Marketing: Pricing and Promotional OptimizationIntegration-Required

Definition

The degree to which AI price test results are distorted by customer awareness of alternative prices, measuring whether the organization can trust the elasticity estimates its pricing experiments produce.

The Measurement Gap

The Gap

Price testing is the foundation of AI pricing optimization. The pricing model shows different prices to different users, measures demand response, and estimates elasticity. Those elasticity estimates drive every subsequent pricing decision. But there is a contamination problem that most pricing organizations do not measure.

Customers talk. They use price-tracking browser extensions. They compare prices across devices and sessions. They share screenshots. When a user-level A/B test shows different prices to different visitors, the existence of the price difference is discoverable. And when customers discover it, they do not behave like the isolated test subjects the experiment assumes. They comparison-shop more aggressively, they abandon at the higher price not because of the price itself but because they know a lower price exists, and they defer purchases to try for the lower price. The measured elasticity is contaminated by awareness effects that would not exist at a single, uniform price.

Standard A/B testing metrics report elasticity with confidence intervals, but those intervals measure statistical sampling error, not contamination error. Conversion rate by test variant measures response to the tested prices but cannot distinguish between genuine price sensitivity and awareness-driven behavior. Revenue per variant captures the outcome but not the mechanism. No standard price testing metric measures whether the test itself introduced behaviors that would not exist at a uniform price.

The Decision

We propose PTCR to answer: Can we trust our price test results, or is price awareness contaminating the data? This metric would compare elasticity estimates from standard user-level tests to elasticity estimates from contamination-resistant methods (geographic holdouts or sequential tests), measuring how much the standard method overstates sensitivity.

Formula

Plain Language

Run the same price test two ways. First, the standard way: user-level A/B test where different visitors see different prices. Second, a "clean" way: geographic holdout (different regions see different prices, so no individual customer encounters both) or sequential test (different prices at different times, so there is no simultaneous comparison to discover). Compare the elasticity estimates. If the standard test shows significantly higher price sensitivity than the clean test, the difference is contamination -- customers in the standard test were reacting to the existence of price differences, not just to the prices themselves.

Notation

PTCR = |Elasticity_standard_test - Elasticity_clean_test| / Elasticity_clean_test x 100

Variables

Variable Definition Units Source System
Elasticity_standard_test Price elasticity estimated from user-level A/B test Elasticity coefficient Price Testing Platform
Elasticity_clean_test Price elasticity estimated from geographic holdout or sequential test design Elasticity coefficient Custom Experimentation (Integration)

Proposed formula, pending validation pilots.

At a Glance

Unit
Percentage
Trend Direction
Lower
Reporting Frequency
Per test cycle
Category
Pricing Measurement Integrity
← Price Model Currency Score (PMCS)Discount Dependency Index (DDI) →

Assess Your Organization

The AI-GPS Assessment evaluates your AI governance maturity and recommends KPIs like this one for your implementation roadmap.

Take the Free AI-GPS Assessment