How often have you heard the phrase, “numbers don’t lie”?
For many, the mere presence of numbers in a measurement or statement will give it all the credence necessary to be the absolute truth. But when those numbers turn out to be no more credible than guesses, they lead to bad decisions.
How some numbers can lead you to the wrong decision.

How often have you heard the phrase, “numbers don’t lie”? For many, the mere presence of numbers in a measurement or statement will give it all the credence necessary to be the absolute truth. But when those numbers turn out to be no more credible than guesses, they lead to bad decisions.

In an environment that produces massive volumes of data, management teams have become ever more reliant on numbers to make important business decisions. This means that numbers not only must be accurate; they also must have a high degree of correlation.

The purpose of this article is not to discredit scientific data; rather, it is to expose a common problem in business decision making—abuse of numbers to bring confidence to statements that lack both scientific proof and empirical data. The article also highlights the dangers of using aggregate numbers to draw conclusions about discrete events.

Abusing numbers to lend credence to statements

Science, mathematics, and numbers are meant to provide exact values, free of bias and emotion. Accordingly, people develop confidence in using numbers as a basis for educated decisions.

But when numbers are used as the basis of a subjective assessment, they lose credibility as a scientific measurement. For instance, many surveys will ask you to rate your degree of satisfaction or agreement with a certain statement on a scale of one to ten.  Although the results of such surveys may serve a useful purpose, they also do not provide an absolute and accurate measurement. In any such surveys, an alphabetical (e.g., A to J) or other scale could easily replace the numeric rating system.

Management will often set hard boundaries (e.g., a number lower than 9, a score between 7 and 8) to draw conclusions and possibly initiate major changes that may be detrimental to the organization. The use of numerical value may give the appearance of a scientific approach, but the fundamental assessment still resides in a subjective interpretation of the questions.

Using aggregate numbers to assess discrete situations

The quest to simplify metrics and provide management with an overall aggregate number to assess performance and/or benchmark the organization or different business units within it can lead to erroneous conclusions and bad decisions. An aggregate metric may reflect overall performance, but it makes a very poor tool for diagnosing problems and initiating corrective measures. The following examples illustrate the pitfalls of aggregate measurements.

“Sales per labor hour” (“S/LH”) is a standard key performance indicator (“KPI”) used to assess and compare retail outlets. As the name suggests, it is simply a ratio of sales to labor hours. Unfortunately, most large retail chains use this KPI to rank store management and/or to allocate labor hours. But there are many variables that can affect S/LH, and using this KPI to benchmark stores or plan labor assumes that all of these variables are equal across all stores—a risky assumption! For example, one major variable that affects S/LH is demographics. A more affluent customer base will tend to buy more expensive products within a particular category (e.g., perfume, electronics, food), which translates to a higher S/LH. This is by no means a representation of labor force performance; rather, it is merely a reflection of higher sales. As we know, “sales will hide many sins,” and this single demographic variable can completely skew a store ranking. Using only S/LH as a labor-planning baseline, therefore, will result in overstaffing stores in affluent regions and understaffing stores in the lesser affluent regions, and that can lead to customer service issues.

In distribution, “cases per hour” (“CPH”) is a standard KPI used to benchmark operations and measure throughput. Again, this is an aggregate value driven by many variables, such as pick density (average number of cases picked from a given location), layout, slotting, weight of cases, length of the travel path, equipment (condition, age, speed), etc. Not only do these variables vary from distribution center to distribution center, they also vary on a daily basis within one given facility. Using such a metric for internal benchmarking may lead to a conclusion that one facility excels, when in reality it could be just average or worse. Using CPH for benchmarking purposes becomes even more problematic when applied across different industries; for example, a satisfactory CPH in the foodservice distribution industry may be a mediocre measure in the food retail distribution sector.

Discrete measurement for specific decisions

Understanding that aggregate measurements are not adequate for addressing discrete components of your business should lead you to seek other measurements that address specific needs. For instance, the distribution industry uses discrete engineered labor standards to measure the productivity of its workforce and to plan labor requirements. This form of measurement is specifically designed to account for the difficulties an employee encounters during each of his work assignments—thus, it can drive better and more confident decisions about workforce management, labor planning, process evaluation, and many others issues. Furthermore, advancements in business intelligence and dashboards make it easier to rely on multiple discrete measurements rather than a single aggregate KPI.