Most firms report (and advertise) customer satisfaction in percentage terms: “85 percent of our customers are satisfied,” “our customer satisfaction score is 90 percent,” and the like. This is mostly nonsense. Such assertions are the same as measuring intelligence by asking “Are you dim or smart”? the answers will not correlate to scholastic performance or to anything else for that matter.

What intelligence and satisfaction have in common is that they are both unobservable and they are both matters of degree. Sure, you can be satisfied or dissatisfied, smart of dim, but there is an underlying continuous property in both cases. Once we express something as the percentage of people expressing “satisfaction or “dissatisfaction,” we are not capturing the continuum. Information is discarded for the sake of black-and-white minimalism.

That’s a costly mistake. First, it leads to imprecision. Why would anybody want imprecise information? In statistical terms, the resulting number (estimate) has a large margin of error. In other words, noise looms large. The reason (or perhaps it’s an excuse) for expressing customer satisfaction as a binary “either-or-proposition” is that it is “simple.” But that is not the case. It’s not simple. It’s simplistic. And it is costly.

The imprecision of the information—which shows up in inexplicable random movements over time and make it impossible to compare across products, regions, outlets, etc.—is contradictory to the purpose of measurement. Measurement is about precision. Imprecision is easy enough to get in other ways.

Random noise is the opposite of precision. Obviously, random noise cannot predict future customer behavior. It cannot be linked to operations, either. Both links are critical.

If the customer satisfaction measure cannot be tied to future financial performance, its economic relevance is lost. If it can’t be tied to operations, managers can’t execute.