1. Designing retention strategy
It's good to know employees who have a high risk of turnover, but in reality, we usually want to design focused retention strategies for a wide range of employees with different risk of turnover. In order to do this, we will classify employees into different risk buckets.
2. Know who may leave
You already calculated probabilities using the predict() function, but in this lesson, we will take a slightly different approach to do the same.
As retention strategy is designed for employees who are active, that is, who are still working in the organization, you will calculate the probability of turnover only for them. So first, let's filter the dataset to only active employees and then use the tidypredict_to_column() function from the tidypredict package. Along with the data, you pass the final logistic regression model to this function.
3. Know who may leave
tidypredict_to_column creates a column named fit which stores the probability of turnover for each employee. Let's see the top 5 rows with the highest probability.
Here we select the emp_id and fit columns and pick the top 5 rows with maximum probability.
4. Classification of employees in risk buckets
Now that we have calculated the probability of turnover for each employee, we will classify these employees in four different risk buckets.
5. Classification of employees in risk buckets
So far, we identified that by using 0.5 as the cut-off value, our model accuracy is best. Using 0.5 as the base we will classify all employees who have a turnover probability less than or equal to 0.5 as No-Risk. Now employees with turnover probability value greater than 0.5 can further be classified into different risk buckets such as;
6. Classification of employees in risk buckets
Low risk, employees who have turnover probability between 0.5 to 0.6
7. Classification of employees in risk buckets
Medium risk; employees who have turnover probability between 0.6 to 0.8.
8. Classification of employees in risk buckets
High Risk; employees who have turnover probability greater than 0.8.
9. Classify employees into risk buckets in R
To do this, you can use the cut() function. cut() can be used to classify a set of continuous numbers into categories.
The breaks argument specifies the points at which you want to cut the continuous set of numbers and the labels argument to assign names to these categories.
In practical scenario you may need to validate the risk buckets of these employees by having more organizational context & information about these employees.
You may again reclassify these employees in different risk buckets by using different turnover probability ranges as well.
10. Retention strategy
Once you have the risk buckets, you should prioritize the retention strategy for employees identified in the high-risk bucket.
As an HR professional, you will work with the management team to come up with an action plan and hold conversations to understand the engagement level of the employee,
identify issues and concerns which require immediate action.
Similarly, for medium-risk employees, you can track behavioral changes and schedule discussions which can be an open house or one-on-ones.
11. Retention strategy
For all employees in the low-risk bucket, you can understand employees concerns by an open house discussion, after that create an action plan and bring their concerns to closure.
For employees in the no-risk bucket, there is no action required on an immediate basis.
Of course, these were just some suggestions and depending on your organization, you can come up with several other strategies.
12. Let's practice!
Go ahead and classify the remaining employees into risk buckets.