I'm a strong believer in simulating and evaluating business decisions whenever possible. Companies should consider many factors when designing employee review policies, and using simple models, one can evaluate combinations of these factors using simulation.

This program simulates a review process where employees in the bottom X percent of a company are fired and replaced for several quarters.

This tool might be used to determine the combination of review process attributes that lead to increased company production.

In the simulation, employees in each review period contribute a total amount of production to the company's total. Managers then estimate (with assessment error) how much each employee produced. The employees are sorted based on the manager's rating, and the bottom X percent of the employees are put in a misses bucket.

If employees are grouped in the misses bucket for a specified number of quarters within a larger window of quarters, the employee is fired.

When an employee is fired, that employee's slot remains empty for a specified number of quarters. When the employee slot is filled with a new hire, he or she produces at a specified lower level for a quarter. The simulation can also simulate drawing new hires from a population with improved average productivity over the existing population (subject to manager assessment error).

Output values are total average company production and a histogram of the production of fired employees for each policy configuration. Presumably, one would select the policy that maximized productivity while minimizing firing.

To run:

python simulate_qpr.py | python -mjson.tool