In this article, we will go step-by-step in this analysis of employee resignation. We suggest you login to your AnswerMiner account and follow the steps to explore the HR dataset. It is a free set that you can find in the data-market within the app.
Start!
Let’s say you want to understand why your best employees are leaving prematurely and know who is the most likely to leave next.
First to have this information, you need to collect some data about your employees. I am sure you have it, and it is something like this in this example.
Metrics
- Satisfaction Level
- Last Evaluation
- Number of Projects
- Average Monthly Hours
- Time Spent at the Company
- Work Accidents
- Promotions in the Last 5 Years
- Departments (column sales)
- Salary
- Employment Status (left or not)
First Stats
When you have this data, you first need some descriptive statistical information to understand the big picture.
You find something like
- Average Satisfaction Level: 61% (median 64%)
- Attrition rate: 24%
- Percentage of Employees in the Low Salary Category: 49%
- Average Monthly Hours: 201
- Number of Projects per Employee: minimum 2, maximum 7
- Performance: 72%
DataViz
Now that you have a quick understanding of the situation, let’s make some visualizations of a correlation matrix to find the relationships between the factors, such satisfaction level and leaving the company.
On average, people who leave have a low satisfaction level, work more, and didn’t get promoted within the last five years.
Based on this information you can filter and see only the people that have left the company so that we can visualize the distribution of each feature.
Based on this data, you did not want to retain everybody. Some people did not work well, as you can see from their evaluation, but there are also many good workers who leave and this is the problem.
Know Your Employees
Out of the15,000 employees that compose our database, 3,571 people have left the company.
More problematic, the total of employees who received an above-average evaluation (72%) spent at least four years with the company, or were working on more than five projects at the same time and still left the company is 2,014. These are the people the company should have retained.
By now, you probably have a picture in your head of your own company and which factors you should pay attention. Now you are trying to figure out the answer to next question:
I see who left the company and that there were many good workers we should have retained, but how can I predict when someone will leave, and who will that person be?
Outcome
To answer this question, there is a tool you can use: the prediction tree.
Based on the tree, you can conclude three important outcomes:
- There is an 86.57% chance that an employee will leave the company if he or she works more than 2,016 hours per month, his or here valuation is above 80%, and he or she has worked for the company for more than five years.
- There is a 94.93% chance that an employee will leave the company if he or she only two projects and the satisfaction level is below 45.6%
- There is a 100% chance that an employee will leave the company if his or her satisfaction level is below 11.5% and has more than three projects and the evaluation level is below 46.5% -it is probably not a problem if this employee leaves.
As mentioned above, the main metrics to pay attention are satisfaction level, working hours, number of projects, and the evaluation level.
Filtering the dataset based on this information will help you find out exactly who will leave next so you can take care of them and prevent their resignation.
Afterlife
Take action based on your analysis.
- Contact the employee personally and find out the reason why he or she is planning to leave.
- Contact the department manager and tell him or her the information you have found so that he or she can pay attention to the identified employee and prevent the resignation.
This example was one that showed how you could perform better in employer branding and keep people at your company with the help of data-driven management. As mentioned a beginning of the article, you can find the dataset in the AnswerMiner data market and analyze it or integrate your own data.
