Definition
The percentage of competitor price changes correctly classified as strategic (requiring response) versus noise (clearance, errors, tests), measuring whether the repricing system's signal detection is accurate enough to avoid wasting margin on false alarms.
The Measurement Gap
The Gap
A competitor's price drops by 8%. The repricing AI sees the signal and responds within minutes, matching or undercutting. Three days later, the competitor's price reverts to its original level. It was a clearance event. Or an error. Or a price test. And your repricing AI just gave away margin responding to something that was never a real competitive threat.
Legacy competitive monitoring tracks competitor price changes as events, but it does not distinguish between types of events. A strategic price reduction (intended to hold and capture market share) demands a response. A clearance markdown (clearing slow inventory before discontinuation) does not. A pricing error (reverted within hours) certainly does not. A price test (exploring elasticity, reverted within days) is noise that looks exactly like signal until the reversion happens.
Conversion rate and win rate are both downstream of this problem. They improve when you respond to competitor price drops regardless of whether those drops are strategic or noise. Competitive intelligence dashboards count price movements but do not classify them. Revenue per competitive response blends valuable responses with margin-destroying reactions to noise. No legacy metric separates the signal from the noise before the repricing AI acts on it.
The structural issue is that repricing systems are built to respond fast, not to respond accurately. Speed rewards reactivity. Accuracy requires patience -- waiting to see if the competitor's move holds before committing margin to a response.
The Decision
We propose CPSA to answer: Is our repricing AI responding to real competitive threats or reacting to noise? This metric would retrospectively classify competitor price signals and measure the accuracy of the system's classification, enabling teams to add reversion filters and minimum-hold-time requirements before competitive responses trigger.
Formula
Plain Language
Sample competitor price signals from the period. For each signal, wait 7 days and check: did the competitor's price hold, partially revert, or fully revert? Classify each into three categories: strategic (price held), noise (price reverted), or indeterminate (partial revert or ambiguous pattern). Then check how the repricing AI classified the signal at the time it responded. Divide correct classifications by total sampled signals. Separately, calculate the total margin spent responding to signals that turned out to be noise.
Notation
CPSA = Correctly_classified_signals / Total_sampled_signals x 100
Misclassification cost:
Misclass_cost = Sum(Margin_lost_on_noise_responses)
Variables
| Variable | Definition | Units | Source System |
|---|---|---|---|
| Correctly_classified_signals | Number of competitor price signals where the system's classification (strategic vs. noise) matched retrospective verification | Count | Retrospective Classification (Derived) |
| Total_sampled_signals | Total number of competitor price signals sampled for classification verification | Count | Competitive Monitoring Tool |
| Margin_lost_on_noise_responses | Margin difference between the response price and the hold price for signals retrospectively classified as noise | Dollars | Repricing Platform / Financial System (Derived) |
Proposed formula, pending validation pilots.