Boosting Productivity Through Statistical Experiments

By Justin Lahart

As human beings, we’re often not good at dealing with randomness, probability and cause and effect, and that can lead us astray. Take the “hot hand” in basketball. The idea that sinking a set of shots makes a player more likely to sink the next one has been shown to not be true. But try telling that to players and coaches.

The same is true for companies, too – as many people who have presented management with information that didn’t fit the bosses’ worldview can attest. One way to cut through such problems is to use the same sorts of statistical methods that economists use to tease out the lines of cause and effect in complex situations.

Polyester yarn-maker Unifi came up with 25 things that people there thought were surefire ways to cut down on how often yarn broke as it was being “textured” – a process of stretching and heating that gives it bulk and strength. A statistical analysis by consulting firm QualPro found that only five of those ideas really worked. Some of them actually hurt, like yanking on splices in the yarn to make sure they were sound – something that some workers routinely did in the belief that it helped.

Other companies have had similar experiences when Qualpro put the things they were doing to the statistical test. Toys “R” Us found that the glossy paper they were using for newspaper inserts was actually lowering sales – consumers responded better to cheaper paper. Salespeople at Southwestern Bell thought that increasing their office hours would increase equipment sales to business, but an analysis showed that would actually hurt because it made the sales force less immediately responsive to customer requests.

Doing statistical experiments on the company floor isn’t a new idea, of course, and having those experiments yield surprising answers isn’t either.

In the 20s and 30s, researchers conducted a number of studies at a Western Electric plant called the Hawthorne Works. One study looked at how higher and lower levels of light affected factory workers. It turned out both did, probably because they were responding to the increased interest in them. Now the “Hawthorne effect” refers to any experiment where subjects make positive changes to their behavior simply because they’re being studied.

Related article: Moment of Truth for U.S. Productivity Boom