Testing Your Prices Without Losing Customers: The Method

The fear that prevents any price changes

The reasoning that prevents people from testing a price can be summed up in a phrase heard in almost every organization: “If it goes wrong, we’ll see it right away—it’s better not to take the risk.” This phrase contains both a truth and a fallacy.

The truth: A price change that is poorly received quickly shows up in sales figures—and sometimes in customer returns. The mistake: Never testing a price comes at a cost that is just as real, though invisible—the one described in the first two articles in this series: the lost profit margin resulting from a price that was never challenged, and the cost of a price that has remained fixed since its launch.

This asymmetry in perception—a visible risk versus an invisible cost—explains why so many companies prefer to do nothing, even though the available data show that fear is almost always exaggerated compared to the actual risk.

This fear is also fueled by a classic availability bias: the only example of a price test that sticks in people’s minds is almost always a resounding failure—a price increase that was poorly received, widely discussed, and has become a classic internal case study. Successful tests, on the other hand, leave no memorable trace: they blend into normal operations, precisely because they did not provoke any notable reaction. This asymmetry in memory distorts the collective perception of risk far more than the actual data would justify.

What Customers Really Notice

The implicit assumption behind the fear of testing a price is that customers compare, remember, and penalize every price change. The reality, as documented by research, is much more nuanced.

This finding does not mean that price is unimportant—it is, and very much so. It means that accurately recalling a past price—a necessary condition for a customer to “notice” a price increase in the strict sense—is much rarer than we assume internally, where every price is constantly scrutinized.

This discrepancy between the feared elasticity and the actual elasticity is precisely what a test allows us to measure, category by category, rather than assuming it once and for all.

These two findings reinforce each other. If a significant proportion of customers do not recall a price precisely, and if actual price sensitivity is often lower than feared, then the perceived risk of a well-conducted test is structurally overestimated—without, however, falling into the opposite extreme: certain categories remain genuinely price-sensitive, and only a rigorous test can determine which category a product falls into.

The Five-Step Testing Protocol

Test with a small amount of data or traffic

The most common objection to price testing comes from companies that believe they lack the volume needed for a statistically reliable test—a single store, a limited catalog, or low traffic. This reasoning isn’t entirely wrong: a traditional A/B test, in which two groups are compared simultaneously, does indeed lose statistical power when volumes are low.

But the lack of large volumes doesn't mean we can't test—it just requires us to adapt our approach:

• The sequential test compares a period before and a period after a price change, within the same population, rather than two simultaneous groups—which is useful when it is not possible to isolate two comparable segments in parallel.
• A price sensitivity survey directly asks a sample of customers about their perception of a price before any actual change is made—a method that requires only a small sales volume to be useful.
• The deferred reference test adjusts the price of a handful of representative products rather than an entire category, to limit exposure while maintaining a usable signal.

In any case, the rule remains the same: it is better to conduct an imperfect test methodically than to conduct no test at all—the same rule already stated in the article on price as a dynamic variable in this issue.

In fact, a lack of traffic isn’t always the obstacle one might imagine. A company with few SKUs but a clean sales history can often learn more from a well-conducted A/B test than a large retailer with thousands of SKUs but data scattered across several unconnected systems. Volume helps, but the quality of the method matters more than the size of the sample.

What to Measure After a Test

This table is sufficient for evaluating a single test. To go further—to precisely calculate a price elasticity coefficient, choose between various measurement methods (before/after, A/B testing, modeling), and avoid statistical biases (data gaps, overlapping promotions, seasonality)—Booper has published a dedicated guide: How to Measure Price Elasticity? Method. This guide deliberately focuses on the early stages—the decision to test and the minimum protocol; the other guide covers the calculations and statistical rigor.

Errors That Skew a Price Test

• Testing the entire range all at once. This eliminates the possibility of comparison and turns the test into a blanket guess—exactly what a test is supposed to avoid.
• Stopping the test too early. An immediate reaction is not always a lasting one; drawing conclusions before the end of the buying cycle almost always skews the results.
• Ignore external factors. A competing promotion during the test period may cause the price to be attributed an effect that is not its own.
• Never make a decision. A test whose result leads to no decision—no generalization, no abandonment, no adjustment—serves no purpose other than to postpone the issue yet again.

FAQ

‍

Also in this series

• Why Do So Few Companies Actually Manage Their Prices?
• The price isn't set in stone: why it's a dynamic variable

Sources: NielsenIQ, Everyday Analytics, “Busting Pricing Myths,” 2021 · Journal of Retailing, Jensen & Grunert, “Price Knowledge During Grocery Shopping: What We Learn and What We Forget,” 2014