In the corporate world, at the end of every financial year comes a phase which we look upon with excitement, hate and fear. And that is the process of Performance Appraisals. The feelings would vary depending upon how an individual would have performed against his/her targets. But the most intriguing part of the appraisal process is the Normalisation of performance.
While the assessment of an individual’s performance usually happens between the boss and sub-ordinate, the overall rating of individual’s potential to grow left to the Human Resource (HR) function. Using property of a statistical probability distribution, the HR folks try various permutation/ combinations to classify the performances of all the employees in the organisation into Good, Average and Not-so-good. And if the overall business performance is subdued or if there are pressures for manpower rationalisation, the classification brutally becomes – the Good, the Average and the Bad. One cannot help but compare the situation to the title of the 1966 Hollywood Cowboy classic – The Good, the Bad and the Ugly.
When confronted the HR folks give justification that most of the natural processes and outcomes follow Normal/Gaussian distribution. Hence all performance of the employees in the organisation must also follow Normal distribution. But the process of Bell-curve fitting usually undergo further mutations. The proportion of the Good, Average and Not-so-Good is varied arbitrary. And there are further classes added – Excellent, Superior, Good … so on.
The use of Bell curve in Performance Appraisal raises several questions about Applied Statistics, which needs to be answered if the process of Performance Appraisals is to be implemented as a scientific management process.
Q1. The Normal distribution is a probability distribution for continuous variable. How can it be applied to Attribute data?
The Normal distribution is a probability distribution applicable to a random continuous variable. For example, height of all the students in a particular class would be normally distributed.
In contrast, during the Performance Appraisal, the employees are binned into 3 or more discrete classes (attributes). Forcing the properties of continuous variable probability distribution on Attribute data is not appropriate.
Q2. The probability distributions are used to make inference about the population based on a sample. During Performance Appraisals, the entire population data is at disposal of the Management. What is the need of Statistical probability distributions?
In Statistics, you study a sample – its distribution – superimpose it on a probability distribution – if a fit is established, using properties of probability distribution make inferences about the population. During Performance Appraisal data of the entire population is available for analysis. The management can study the population census and take informed decisions.
Q3. Before using the properties of Normal distribution, one needs to ascertain whether the sample fits the probability distribution. Is this done before fitting the Bell curve?
Before using the properties of Normal probability distribution, one needs to test whether the data follows Normal distribution. There are various tests to ascertain this.
In cases where the team sizes are small or all the employees are highly skilled or qualified the performance may not be Normally distributed. It could be skewed towards one end. In such cases force fitting the Bell curve would be inappropriate.
Answers to the above questions need to be found before segregation of employees could be labelled as Normalisation. Else, the entire exercise becomes inappropriate.
The Management might declare that the segregation would be done arbitrary or do away with entire appraisal process, as initiated by some forward thinking organisation.
