This guest post from Dwight Ueda of Operating Principals, LLC explores the practice of forcing the distribution of performance ratings to fit a bell curve, explains why the practice is not statistically valid, and why it can in fact be harmful to organizational performance.
Have you ever been asked to re-rate the performance of a subordinate after you submitted a rating for approval? Despite having already communicated the original rating to the subordinate?
Have you ever been that subordinate? If not, consider yourself lucky. But one day it will probably happen to you if your organization uses forced distribution for its employee performance ratings.
In most cases, re-rating occurs when submitted performance ratings do not fit a bell curve. And usually, because ratings skew towards high performance.
Rather than accept the ratings, organizations (despite many having aspirations to being high performing) push back, asking managers to reduce their employee performance ratings so the distribution creates a bell curve (i.e., statistically normal).
The logic is: if you give one employee a high rating, you must give another employee a correspondingly low rating. It seems to make sense conceptually, until we look at it statistically.
There's nothing normal about the bell curve
Bell curve distributions, or normal distributions, are believed to be naturally occurring. They are what we expect when we measure natural phenomena like annual rainfall in New York City, women's shoes sizes, etc.
However, bell curve distributions are not the only statistical distributions that occur naturally. To presume that all distributions are bell curves is unnatural.
Take the following age-population distributions for instance.
In West Africa, the distribution looks like a steep ski slope with a demographic peak at ages 0-4 years. In contrast, Western Europe's age-population distribution resembles an uneven wall that trails off to the right. The peak occurs at ages 30-45, followed by a smaller, youth demographic, reflecting declining birth rates.
Neither of these regions has a "normal", bell-shaped distribution. In fact, to have one would be a national calamity as it would mean no one is having children!
Now, look at credit scores in the U.S.
Here too, the shape is not a bell curve. Instead, the distribution climbs, peaking near the end where it then drops precipitously.
Finally, let's look at the distribution of iron man triathlon times, where the trend is the mirror image of the credit score distribution.
Again, the distribution of these performances does not fall into a symmetrical bell curve.
Clearly, other distributions also occur in nature, frequently too. A bell curve is just one possibility among many.
Aspirations can drive outcomes
There's something else interesting to note about our last two examples. They appear to show that aspirations drive performance towards a desired outcome, whether it's credit worthiness or a winning time.
When people aspire to a goal, they modify their behaviors to pursue it.
So, why do many organizations demand that performance distributions be bell-curved?
We need to go back to school
For many of us, our familiarity with the bell curve goes back to our school days. But even there, the model can break down.
An educator's primary duty is to help as many students as possible master the subject matter. It is not to assess student mastery of said subject matter.
As a result, educators should work to drive student scores towards higher marks, resulting in distributions that look like the FICO scores.
Source: Dr. Rick Yount
That is, the earned grades skew toward higher marks, rather than being limited by bell curve-derived quotas. An educator's success is often measured by the degree to which their interventions raise scores beyond the constraints of the normal, bell curve.
How does all this apply to performance rating distributions at work?
In their quest to identify and retain top performers, organizations want to be certain that top performers are indeed top performers. They don't want good or much improved performers to be mistaken for top performers. So it's sometimes good to question ratings to make sure supervisors are properly applying the rating standard.
But workplaces are no different than schools when it comes to expending effort to raise performance. In fact, organizations often use more interventions than schools do to increase performance levels.
Workplace performance management and reward systems are designed to define, communicate, and reward desired performance. With these systems, the workplace provides guidance, coaching, and incentive for employees to perform well.
Consequently, it would be expected and natural for performance ratings to skew toward desired performance like with FICO scores.
When it comes to organizational health, the bell curve just doesn't fit
Despite this, many organizations still insist on distributing performance ratings along a bell curve. This insistence is not based on real world observation but on an inappropriately chosen model.
If the purpose of performance management is to raise performance levels, then we need to expect and desire skewed rating curves.
Re-rating and downgrading employee efforts does the exact opposite of motivating and rewarding employees. Employee commitment to personal and organizational improvement is likely to weaken when the organization fails to demonstrate appreciation for employee efforts.
Worse, if organizations demand that performance ratings must range in the middle, then they are likely to get middling performance.
Organizations should eliminate the practice of forced distribution because it is statistically unsound and disengaging for employees.
The best way to build a high performing workforce is to leverage the pride that people have in their work and to recognize the effort and improvements that come from it. Performing normally means skewed, not "normal", performance rating distributions.
Your Turn: How have you seen the forced distribution of performance ratings discourage performance?