How to predict the likelihood of an employee quitting and get a million more insights from one graph
Hello, my name is Evgeniy Kiriyok, I am a mentor on the course on HR analytics in Yandex Workshop, author of the book “HR Analytics. A Guide to Personnel Analysis and Channel AnalysisAnalytics in HR“
What if I told you that there is an HR analysis method that can predict the likelihood of an employee quitting during their probationary period? Moreover, this method will allow you to evaluate the quality of different sources of candidates, and will also identify problem areas in the adaptation process, evaluate the effectiveness of selection and give many other conclusions about the life cycle of an employee in the company. And you don't need to have a degree in math or understand data analysis to use it. It is available to everyone. Intrigued and want to know more?
In that case, I am pleased to present to you His Majesty's Survival Chart.
The survival plot, or Kaplan-Meier plot, comes from medical research and was developed to assess the survival of patients during treatment. Yes, yes, everything is serious, doctors wanted to understand how this or that treatment (or lack thereof) affects the survival of patients over time. Later, this method began to be used in research as a fairly universal tool for analyzing the final states of processes and the probability of their occurrence.
Below we will learn how to build a survival graph using Excel and interpret the observation results. We will deliberately not use R, Python and more complex stats. calculations in order to make the tool accessible to practitioners without knowledge of data analysis.
So why do I need it?
This method is used for very different tasks, here are a few of them:
Assessing the likelihood of an employee quitting over time
Assessing the effectiveness of the company's adaptation process
Evaluating the effectiveness of the selection function
Assessing the quality of selection sources
Calculation of average employee life cycle indicators in a company and much more
Okay, let's tell you
Let's start with something simple. Let's say you want to evaluate the quality of the selection and onboarding function. To do this, you collected information about the company's hiring for a period of 3 months from one source of candidates. It does not matter how long the employee actually worked – all three months or one day. Let's present the resulting data array in the form of a table:
Nothing complicated yet: there is the date when the employee started working for the company, the date when he left it, and the number of weeks he worked. If the employee continues to work, he will not have a termination date or number of weeks worked.
Now let's start the calculations. To do this, we will build another table next to it, which will have the following fields:
employee is the name of the employee who left in time period t;
time
