What Do You Do When a Supplier Underperforms? Remember “reversion to the mean”

Getting to Yes is for me the most important non-fiction book of my lifetime, and we've talked about it here several times. Daniel Kahneman is a behavioural psychologist who won the Nobel Prize for Economics for pointing out that people don't behave logically, which somewhat messed up the theoretical models economists  used.

He talked about concepts such as priming, anchoring, and our somewhat illogical (but understandable) attitudes to risk. But another one of the chapters in the book discusses a point that is very relevant to our business lives in a number of ways, from how we motivate staff to how we contract manage suppliers.

Kahneman tells the story of talking to an air force instructor in the Israeli military. He told Kahneman that the "stick" worked much better than the "carrot" as far as he was concerned. So if a trainee had a bad flight and the instructor gave him or her a good talking to (or worse), then the student tended to perform better the next time. However, said the instructor, praising people after a good performance didn't seem to work - they tended to perform worse the next time.

It took Kahneman a while, he says, to work out what was happening, and come up with the logical explanation. Much of what we do in life is subject to some natural variation around an average level or a trend level. So perhaps I can run 10K in 40 minutes (I wish), but on a good day that might be 38, on a bad day 43 minutes. There might be an improving trend if I train hard enough, but the variation will remain.

So think about it. If I have a 38 minute run, and you praise me, the likelihood is still that next time it will be around 40. It looks like your praise did not work. If I do 43 minutes and you scream at me, then hey presto! Next time I do 40. But that is nothing to do with the carrot or the stick - it is what the statisticians call reversion to the mean.

Think about this for staff performance. Sue does a really poor presentation at the team meeting. Sure, a good manager might want to ask if everything is OK - but it might just be that off-day, and next time will be fine. It is more important to look for trends, so if performance is declining once the variation is taken into account, then that is significant.

And it is the same for suppliers. An outsourced service provider has a bad month on the customer service metric. If you kick them it probably will improve next month, but that's nothing to do with your kicking. But if you don't kick them, it probably will improve next month too as we see that reversion to the mean come into play.

Equally, don't get excited by a single good month, but look for the trends. Don't take too much notice of any single observation of measures, good or bad. Be consistent in your messaging, look for continuous improvement and clear trends (good or bad) by all means, but don't be too fulsome in your praise or indeed too scathing in your criticism based on one bad result.

Share on Procurious

First Voice

  1. RJ:

    Thanks Peter for reminding me about one of the most interesting chapters in the book.

    Another factor that buyers often fail to understand and that often goes hand-in-hand with regression, especially in monitoring SLA performance, is the small numbers fallacy. We will often set targets such as 98% on-time/accurate delivery, with a Service Credit applying for each 1% (or even 0.1%) below target achieved. which sound great and work well if a supplier is making hundreds or thousands of deliveries each month. However, if a supplier only makes daily deliveries (5 a week) then one single failure in a month will immediately push the performance down to 95% and potentially trigger significant compensation, even if this is the first failure in a year.

    Even such targets as “99.8% systems availability Mon-Fri during working hours” need to be carefully considered – this might translate to less than 20 minutes of downtime in a month.

Discuss this:

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.