I have written posts in the past about the importance of developing analytics that use your data to understand where your pricing is effective and where it is not. See Pricing Analytics Can Improve Your Profitability and Just the Facts Ma’am. One important caveat is that you must measure the right things and avoid creating cool charts that simply confirm your biases. Using your data to mislead others is as bad or worse than not analyzing it at all.
In 2015, the Wall Street Journal published an article, How Not to be Misled by Data, in which the author gave examples of people citing statistics in an effort to end an argument, but the statistics were misleading. We have had employees of clients do that with us several times, but we are never satisfied with a single statistic, because it can mislead. As an example, one client had been running a test of price changes at a sample of customers. After four months, they compared their transaction volume and margins to the four months prior to the launch of the test. Sample customer volumes and margins had both decreased during the test period, and the client concluded the price changes were ineffective. We dug a little deeper.
We showed the client the test period had always been seasonally lower than the four preceding months, and their analysis did not account for that seasonality. Our data showed the volume and margin reductions in the test period were less severe than the seasonal declines of previous years. We also showed that the year-year seasonal reduction of customers who received the test prices, were two percentage points better than for customers who did not receive test pricing. On the surface, our measures showed the pricing test was successful. Together with the client, we continued to dig deeper.
For a few years prior to the test, our client had been tracking whether their customer requested a quote before ordering, either by online lookup or by phone. They knew that some customers checking pricing nearly all the time, while others simply ordered without comparing prices. They said the margin improvement at the sample customers primarily came from customers who seldom asked for quotes, and higher prices were driving away customers who frequently compare prices. If true, that would be an important insight. Once again, their statistic was misleading.
We took the group of customers who received quotes before ordering at least 85% of the time, and for each product purchased we measured prices by quintile. (Each quintile contained 20% of the unique price points observed.) In the four months prior to the pricing test, both sample customers and non-sample customers had transactions somewhat evenly distributed among the quintiles (see Figure 1). In both cases the middle quintile had the most transactions, and the lowest price quintile had the fewest.
More importantly, the sample customers saw a substantial increase in the percentage of transactions occurring in the top two quintiles, while the other customers stayed consistent. In other words, the pricing test had successfully moved a large number of transactions to higher price points. So perhaps these customers who receive price quotes all the time are not as price-sensitive as our clients believed.
Of course, there are many dimensions to price sensitivity, beyond how often the customer compares prices. Identifying those dimensions is part of the segmentation process. Together with our client, we continued to evaluate the price points at which business had been won and lost in each segment and price sensitivity dimension. If our client had simply accepted the high-level analysis that volumes had decreased after prices were changed, they would have reverted to their old prices and reduced their profitability.
Analytics remain important, but we should not use them as a weapon. We all have biases, and our analytics should be used to test those biases. We need to be open to accepting that our pre-conceived notions could be wrong and change those notions when appropriate.
Leave a Reply
Want to join the discussion?Feel free to contribute!