Corporate life · Regression to the mean
The Employee of the Month Curse
The short version: any monthly result is part skill, part luck. The person at the top of the list usually had the most luck, and luck doesn't repeat. In our simulation the winner did worse the next month 92% of the time, and the person in last place improved 92% of the time. That's regression to the mean, not a curse, and not proof that criticism works.
Cousin Kevin's sales floor has a wall. On the wall is a frame. In the frame is the Employee of the Month, smiling the smile of someone who has just been given a parking space near the entrance.
Management noticed a pattern. Whoever went into the frame did worse the following month. Almost every time. It was discussed at a leadership offsite and given a name: "award complacency". The award was cancelled.
Instead, the bottom performer of each month is now called in for a "development conversation". Management noticed that almost every one of them improved the following month. It was presented at the next offsite as evidence that tough love works. Everyone agreed. Nobody checked.
Every month is skill plus luck
A month of sales depends on how good you are, and on things nobody controls: which customers walked in, who was on holiday, whether a big deal closed on the 31st or the 1st. So the person at the top of a list of 20 is rarely just the best. They're usually good and lucky that month. Next month, the skill stays. The luck goes back to average.
We simulated it: 20 salespeople, 200,000 months. Each person's monthly score is their skill plus a random amount of luck. Skills differ a bit between people; luck differs more in any single month. Then we looked at what happened to the winner and to the person in last place the following month.
| Following month | Result |
|---|---|
| Employee of the Month does worse | 92% of the time |
| Average drop for the winner | 16.7 points |
| Average luck the winner had in the winning month | 16.7 points |
| Winner wins again | 10% (pure chance would be 5%) |
| Winner's typical place next month | 8th of 20 |
| Person in last place improves | 92% of the time |
Look at the second and third rows. The winner drops, on average, by exactly the amount of luck they had. The "curse" is just the luck leaving. And the 10% repeat rate, twice what pure chance would give, is the part that is real skill. It exists. It's just much smaller than the frame suggests.
Nobody in this simulation was praised, criticised, motivated or developed. The numbers are random draws. The "curse" and the "tough love effect" appear anyway, every time, because they're built into the arithmetic. This is regression to the mean, and it's the same reason every hangover cure seems to work.
Average score: the winning month and the month after
The closed circle
Follow management's logic and every step is reasonable. Praise is followed by decline, so praise is harmful. Criticism is followed by improvement, so criticism works. Therefore, the company should criticise everyone and praise no one. Therefore, results will keep moving up and down exactly as before, and every movement will confirm the policy. There is no exit from this circle, because every possible outcome is evidence that the circle is right.
How to check whether anything actually worked
1. Find people with the same score who weren't treated. If last month's bottom performers improve just as much without a "development conversation", the conversation did nothing. That's a control group, and it's the only way to separate an effect from the bounce.
2. Judge on many months, not one. Skill shows up in the average over a year. Luck shows up in any single month.
3. Be suspicious of anything aimed at the extremes. Any treatment given only to the best or only to the worst will look like it changed them, because the extremes move back towards the middle on their own.
Where it costs you money
"Fund of the year" lists are Employee of the Month for money. The top fund of any single year usually had the luckiest year, and the same arithmetic says most will fall back towards average. Investors who move their money to last year's winner are buying the frame, not the skill.
The moral
The frame on the wall is still there. It's empty now. Management says it's a reminder of the dangers of complacency. Kevin came last in September, had his development conversation, and came fourth in October. He has been asked to share his journey at the next offsite. He has been told the slides should start at 97.
The simulation assumes 20 salespeople whose skill varies with a standard deviation of 5 points around 100, plus monthly luck with a standard deviation of 10 points, independent from month to month. Other assumptions change the exact percentages but not the direction. The company is an illustration. Educational content, not financial advice.