When Does Data-Driven Pricing Create a False Sense of Confidence And How Do You Know When the Data Is Lying to You?

Data-driven pricing has transformed the way companies make pricing decisions. Historical sales, elasticity estimates, competitor prices and demand forecasts can provide valuable evidence for understanding the market and selecting the right action.

But more data does not automatically mean a better decision. Sometimes it can create the opposite problem: a level of confidence that the underlying evidence does not actually support.

This is one of the topics we plan to explore at the Professional Pricing Society profitABLE26 Conference in Atlanta, and there are three questions in particular that we believe deserve discussion.

When does “more data” actually make pricing decisions worse?

The problem is not necessarily the amount of data, but what that data represents.

A retailer may have years of historical sales but very little meaningful price variation. Competitor monitoring may provide thousands of observations without showing whether customers actually respond to those competitors. An elasticity estimate may look precise while being heavily influenced by promotions, seasonality or limited historical experience.

In such cases, adding more observations can strengthen the model statistically without necessarily improving the commercial decision.

The important question is therefore not simply how much data we have, but whether that data contains the right information for the decision we are trying to make.

How can you tell whether the data reflects customer behaviour or your own past pricing decisions?

Historical data is partly the result of decisions the retailer has already made. Sales at a particular price may reflect genuine price sensitivity, but they may also reflect promotions, stock-outs, seasonality, competitor actions or changes in assortment and availability. If prices have historically moved within a narrow range, the data may also tell us very little about customer behaviour outside that range.

This creates a difficult analytical problem: are we learning how customers behave, or simply identifying patterns created by our own previous pricing decisions?

Understanding the difference is critical before historical relationships are used to determine future prices.

What should a pricing system do when the data is weak, contradictory or outside its experience?

Pricing systems are generally expected to provide an answer. But should they always? When evidence is limited or conflicting, producing another precise “optimal” price may create more confidence than the situation deserves.

A more intelligent pricing system should be able to recognise uncertainty and respond accordingly. Depending on the situation, that might mean providing a recommendation with lower confidence, keeping the decision within tighter guardrails, escalating it for human review or running a controlled experiment to generate new evidence.

And sometimes the correct answer may simply be: We don't know yet.

That is not necessarily a weakness of a pricing system. Knowing the limits of the available evidence may be just as important as knowing how to optimize when the evidence is strong.

These are some of the questions we look forward to discussing at Professional Pricing Society profitABLE26 in Atlanta, October 27–30, 2026, as part of the panel:

“Where Does Data-Driven Pricing Create a False Sense of Confidence—And How Do You Know When the Data Is Lying to You?”

‍