Can somebody explain me what is a Bell-Curve, How is the curve derived and How it is used in HR parlance?
I have put this question to many persons and got contradicting answers. So Im looking for something concise and clear.
Answers (5)
Manish B
Senior Manager, Technology at Sapient
Best Answers in: Computers and Software (1), Information Storage (1)
"Bell curve" is usually used as another name for "Normal Distribution graph", a statistical tool used to plot measures values of phenomena that can be measured (or approximated) quantitatively. For a detailed discussion on the Normal Distribution graph I would suggest reading the detailed article on the English Wikipedia: http://en.wikipedia.org/wiki/Normal_distribution. The Normal Distribution graph is very commonly called the Bell Curve since it resembles a bell.
The Bell Curve is an important statistical tool due to an important mathematical theorem known as the Central Limit Theorem which postulates that many physical (and psychological) phenomena can be approximated (with varying degrees of error) using the normal distribution.
The Bell Curve is useful for analysis as it helps identify several statistically important parameters, including:
- Average or mean value of the sample data set;
- Standard deviation of the sample data set - this helps in determining how "consistent" various values are with each other;
- Variance - this is proportional to the square of the standard deviation and usually points at the degree of error in measurement inherent within the sample data set.
If "m" is the (numerical value of) "mean" and "s" the (numerical value of) "standard deviation", usually the following ranges are used:
- the range of values from m-s to m+s are said to be within the variance band and represent stable values
- all other values usually represent unstable values
HR teams can utilize Bell Curves in many different ways provided the data point under study has a sufficiently large number of sample values or observations. For example, if number of hours spent by all people within an organization are plotted on a graph, a Bell Curve will be obtained. Employees having spent more than m+s hours can be thought of as the best trained employees, while those having spent less than m-s hours can be thought of as least trained employees. HR can then focus on providing additional trainings to folks who spent less than m-s hours on training. The concept can be applied very easily to any other similar statistical exercise.
Some common pre-requisites to performing Bell Curve analysis are:
- All samples in the data set must be for the same parameter. For example, mixing room temperature with wind velocities will not help
- Number of samples in the data set should be equal to or greater than 50 (or 60 as many others may claim). A smaller data set will be prone to greater inaccuracies on account of experimental errors
- The process under observation must be conducive to numerical measurement
- The possible number of numerical values for each observation must be discrete and sufficiently large. For example, if a process can have observation values of 1 or 2 only, a Bell Curve will not yield much information.
Maurice W
Currently Looking for a new position
Best Answers in: Blogging (3), Web Development (2), Compensation and Benefits (1), Internet Marketing (1), Search Marketing (1), Computers and Software (1), Using LinkedIn (1)
I think Manish has a fairly comprehensive explanation –though he missed out skew and kurtosis. The main trouble is that bell curves are abused by HR and Other sectors by applying them to areas where they may not be valid.
Whilst you can say use bell curves and statistical theory in production engineering – if I could remember my prod eng courses you can say if I by a new machine tool which improves tolerances by x that will have Y results in term of defects which you can assign a direct roi.
Its us in say trying to remove say the bottom 10% every year only really works for the first year as after that you not working with the same curve.
Also a company like Google will have a radically different distribution of say IQ when compared to the Population as a whole.
Nel Stephenson’s Cryptomicon has a nice explanation of bell curves and manipulating them.
The answers above are great and right on target. Great job!
As to how used in HR: EEO -- used to chart the actual population of employees against community population regarding various attributes (e.g., race, gender, etc.); Quality -- Six Sigma process analysis; Benefits -- experience rating for benefits programs; Compensation - salary levels of exempt population (i.e. do they "fit" curve or is there an abnormality that needs to be addressed?)..other applications too, but you get the idea
Michael
All the answers above are correct BUT THEY MISS A VERY IMPORTANT POINT - a Normal or Bell Curve assumes that the sample of the population was RANDOMLY SELECTED.
You could evaluate a company's employee base using the properties of the Normal Distribution ONLY if the firm's selection, training, and performance managment & coaching processes are non-existent or completely RANDOM. This is why Maurice Walshe's statement "The main trouble is that bell curves are abused by HR and Other sectors by applying them to areas where they may not be valid" is true.
The concept of "forced ranking" is a good example - the bottom 10% of employees MUST be "C-Players" because "you always have some "deadwood." What happens 4 years into your forced ranking program when you have pruned the deadwood and only hired A-Players? It seems to me that you wind up forcing perfectly capable and effective employees into a category driven by the curve rather than by their performance. (In my humble opinion, of course.)
just to complement the above answers:
http://www.agilemanagement.net/Articles/Weblog/HRMyths3-PerformanceBucke.html
http://www.coolavenues.com/know/hr/amit_bell_1.php3
Rgrds,
Marco
