HR Analytics in the Era of the Great Resignation

Employee Turnover and Why You Should Care

        Attrition or employee turnover refers to the percentage of employees who leave your company during a specific period. In general, turnovers include resignations, terminations, or retirements but don't include the internal movement of jobs within an organization.

Why is it important to maintain a low rate of turnover from an organizational standpoint? It comes down to three main reasons:

1. It costs money to hire and train new talent.
2. High attrition impacts your organization's ability to complete your work.
3. High attrition makes it challenging to acquire new employees.
   

It costs money to hire and train talent

Finding the right talent for your own business is difficult and expensive. It's hard to associate an exact cost for recruiting. The recruiting process consists of expenses related to advertising a job posting and any time spent building that job posting, sorting through potential candidates, and then interviewing them.

Suppose we put a price tag on the cost of attrition. In that case, it can range from anywhere from 30% to almost two times an employee's annual salary, depending on the industry.

High attrition makes it hard to get work done

This one needs little explanation, when skilled people leave the company, work is often interrupted or stopped until a replacement is found.  Even then, it can take a while for your new hire to get back to your prior employee’s level of productivity.  It can be months, sometimes years, for that new employee to achieve the same level of productivity as your prior employee.

High attrition makes it challenging to acquire new employees

It's normal to have employees leave the company. However, when you have a high rate of employees leaving within a short period, you ponder "why?", and so does your team and those on the outside looking in.

A high attrition rate is often indicative of a work environment and a company culture that is not sustainable in the long run. With proper analytics, you may be able to link attrition to inadequate work-life balance, poor working conditions, a non-inclusive work environment, etc.

Employees who leave voluntarily due to poor company culture tend not to refer those in their network to their previous workplace. This makes it difficult to recruit talent, which contributes to the cycle of making it more expensive to find new talent and much more expensive to lose existing talent.

The HR Analytics & Attrition Dashboard

So, how do we get ahead of all this?  The domain of HR analytics can give us insight into a broad spectrum of potential issues and solutions, whether it's understanding your organization's workforce distribution, gauging diversity/inclusion goals, or forecasting budgets.

To show you what I mean, we've built a Tableau dashboard to demonstrate the value of HR analytics and attrition analytics. The dashboard shown below is an example of how an organization can evolve its HR Analytics, regardless of where they currently fall on Gartner's analytics maturity model  (Figure 1).

About the Dashboard

         This dashboard was built using this dataset here. With this dashboard, we'll start off with a general overview, followed by an in-depth analysis of attrition.

High turnover presents a number of issues for an organization. Minimizing turnover helps retain experienced talent and positions a business to better attract top talent. This dashboard is intended to help organizations understand their current employee landscape and how they can use their data for actionable insight with regard to attrition. Let’s start at the beginning of the Gartner Analytics Maturity Model…

Demonstrate descriptive HR analytics and attrition (What happened?)

The first page of the dashboard gives us an overview of our organization. Using this page of our dashboards some of the insights we can quickly uncover are:

• Effectively gather the characteristics & demographics of our organization such as a total number of employees, age, gender, backgrounds etc:
• Our KPI’s tell us that our organization has 1470 employees with an average age of 37. We can also immediately observe our attrition rate is at 16.12%. What happened?

• Understand results from recent surveys of satisfaction by department:
• We can observe that our average satisfaction is low (2.72/4) and is different by the type of satisfaction and their respective department.

• Identify individuals who have been in a position for a substantial amount of time.
• The average time a person spends in a role in our organization is around 7 years. From our dashboard, we can see that there are some individuals who have been in the same role for more than 15, presenting an opportunity for further investigation.

Perform diagnostic analytics on Attrition

(Why it Happened?)

We’ve noticed that our attrition rate is high, and we want to dive in further. Our second page of the dashboard allows us to start diagnosing why that may be. Some insights we can uncover from this view include:

• Overtime appears to be a significant driver of why employees leave
• A large number of employees quit within the first year of joining.
• Naturally, employees with low satisfaction responses on our survey had more turnover.

Identify High-risk employees for turnover and showcase model findings on employee turnover

(What will happen? And How can we make it not happen?)

Our third page of the dashboard is an interactive way to showcase our ‘data science’ findings within Tableau. A few highlights from this dashboard are:

• Why you should use this model:
• Attrition costs can skyrocket depending on the employee and can pile over time. By using a model to predict attrition, you can prevent up to 3.3x turnovers by identifying high-risk employees and having conversations/implementing changes before it’s too late. This can save up to $4M in cost.
• Statistically significant influential factors:
• Our model identified working overtime, frequent traveling and job satisfaction as the most important factors for employee turnover. Employees who work overtime are 5x likely to quit. Employees who work in HR and frequently travel are highly likely to quit as well. Actionable insights from these findings can include improving morale within the HR department, initiatives to improve work-life balance and evaluating frequent travel occupations (How can we make it not happen)
• Turnover Calculator:
• Our turnover calculator quickly allows us to calculate the probability of a potential turnover by entering the statistically significant characteristics of an employee. (What will happen?)
• Common Characteristics & Crosstab:
• The crosstab allows us to download a list of high turnover employees based on probability. By identifying common characteristics within the high-risk employees, we can build out programs that will help improve retention across the board. We can also reach out to individual employees in an effort to get ahead of issues and improve employee-employer relationships. (What will happen & How can we make it not happen?)

Getting Buy In on Your Models from Your Team:  Visualizing Model Performance with Tableau

This page of the dashboard is intended for data scientists. As a data science professional, I find it a bit of a nuisance to manage and keep track of models and their health. If tableau is a part of our tech-stack, we can build a dashboard that can help us understand, organize and manage our model lifecycles better. With this dashboard we can:

• Flip between hundreds of different iterations of models we’ve built for our dataset and compare important evaluation metrics.
• By using the top level filters, we can flip through different training splits, variable selection procedures, algorithms and sampling methods used to compare ROC/AUC.
• Dynamically sort through all models with respect to key-metrics and thresholds:
• By using the bottom portion of the dashboard, we can sort through different key-metrics relative to our business context with respect to threshold. For example, if we want to set our thresholds anywhere from 0.01-0.99 and sort by f-1 score to observe the best performing model we can now do that effortlessly.
• Observe K-S Charts and Confusion Matrices:
• We can hover over the bars to get additional information like their respective confusion matrices and K-S statistics.

Conclusion

Hopefully, you now have some ideas on how you can apply analytics to your own Human Resources datasets in order to improve employee retention and understand key trends while you still have time to act on them.