One of the biggest challenges marketers face is to measure the performance of their brands, strategic business units or company as a whole.

Managers love easy to interpret measures, like average customer satisfaction and average rate of drop-out. However, research suggests that managers’ focus on averaging metrics might not always be the best way to track performance.

Managers typically rely on specific metrics, formulas used to summarize multiple indicators of performance into a single, informative and easy to interpret measure.

Average customer satisfaction, average rate of drop-out, average share of wallet and average growth in sales are only a few examples of the type of metrics companies use to measure the performance of their marketing actions.

One thing most metrics have in common is the fact they involve averaging raw performance numbers in order to reach the final summary statistics (for example, a company might compare the number of customers buying their products in different regions, or stores, in one year versus the previous year and compute the average attrition rate across regions).

Yet, despite the popularity of averages, companies can also use non-averaging metrics like the maximum, the minimum or the variance to summarize raw numbers. Marketing professors Steven Shugan and Debanjan Mitra, from the University of Florida, actually argue that managers’ focus on averaging metrics might not always be optimal.

Why do we tend to report averages?
In this article, the authors start by defending that ‘good metrics should capture the information in multiple observations’. In fact, as the authors acknowledge, averages are very easy to interpret, have strong mathematical properties and they are easy to communicate.

Moreover, managers often believe that non-averaging metrics fail to use all available information. Yet, this last conclusion, as the authors point out, is not correct. For instance, we can only determine the maximum attrition rate across regions after considering the attrition rate for each and every region. Therefore, the maximum actually uses all the raw information.

In fact, the authors show both mathematically and using real publicly available data that, under certain circumstances, nonaveraging metrics are actually better able to summarize raw information than the more often used average.

When should non-averaging be preferred?
The authors explain their finding by showing that non-averaging metrics, like the maximum, are able to 1. extract information from the sample size itself and 2. give more importance to rare pieces of information (e.g. the maximum metric gives more importance to the information contained in successful observations, which are rare in failure-rich environments).

Two examples help clarify this. Imagine that you want to quantify the potential of a new product and have access to data on the orders of different stores for this product. The number of stores ordering the product can, in itself, provide information about the potential of the new product.

Similarly, the number of price promotions used in a certain category in a year might provide information about the prospects for that category. Averages ignore this information. The authors note that while the expected value of a mean does not change with more observations, the expected value of a metric like the maximum increases. In fact, more observations increase the chance that we observe an even more favourable occurrence, changing the expected value of the maximum metric.

A second example. Many different reasons can independently lead a marketing campaign to fail (bad timing, unlucky use of media, non-catchy theme, co-occurrence of an even more catching campaign from a competitor). Conversely, for a campaign to succeed several factors, together, need to be aligned for the campaign to be effective and exert the desired effects in observable market outcomes (e.g. sales of a product).

The more likely occurrence of failures (in such a failure-rich environment) renders each failure less informative than each success. Therefore, when evaluating the performance of a certain creative agency, for instance, the maximum performance among its campaigns might tells us more about its innate capacity to craft good campaigns than the average performance.