Use Case

Employee Attrition Analysis Made Easier with Tellius

Employee turnover hurts a firm’s productivity and drives costs related to recruitment, hiring and training. Qualified replacements are hard to find in today’s competitive talent market.

Employee attrition analysis is useful for identifying:

  • What attributes/ teams/ regions are experiencing high attrition (descriptive)
  • Why top talent leaves (diagnostic)
  • Which employees are likely to churn (predictive)
  • How to possibly change outcomes via targeted interventions/investments (prescriptive)

This post will showcase an intuitive and data-driven approach to performing employee attrition analysis in Tellius — the leading AI-Powered Decision Intelligence platform — in order to help HR departments reduce attrition, maintain productivity, and avoid unnecessary recruitment/training costs.

Why Employee Attrition Analysis is Hard

Most firms understandably spend more time and energy backfilling positions than seeking answers to why the employee left in the first place. That’s because employee attrition analysis is hard:

  • Poor data. Exit interviews are lagging indicators (too late to take action). Satisfaction surveys might be overly positive or misleading.
  • Hard to bring it all together. It can be time-consuming to bring together multiple data points and generate meaningful insights about attrition.
  • Multivariate Problem (No single explainer). No single driver can typically explain someone’s voluntary departure — the reality is messier since each firm has its own unique/complex ecosystem of drivers.

What is the Potential Solution?

A data-driven approach is critical to analyzing employee turnover, utilizing multiple data points to identify hidden connections and key predictors of employee turnover.

Classification models are the main approaches applied today to measure and predict the risk of employee attrition. The problem with these predictive models is that they are typically complex, hard to interpret, and lack concise explanations of what factors affect attrition for HR leadership to actually act upon.

Tellius is an AI-powered analytics platform that helps business and analytics teams across a variety of industries — including HR professionals — quickly parse multiple datasources for ad hoc exploration (querying) and automated analysis such as the main drivers and predictors of employee turnover — so HR and Business leaders can course correct faster, introduce corrective action sooner, and retain the top talent to drive productivity and reduce costs.

People Analytics / HR Analytics in Tellius

Let’s dive deeper into how an HR department can use Tellius to quickly perform employee analytics, identify insights from employee data to identify the main drivers of voluntary departure as well as key drivers of employee retention.

Step 1: Load the Data

Schema:

  • Employee Personal Information
  • Employee Attrition and Job Satisfaction Survey
  • Department and Manager Information

We start by loading data from internal and external sources.

Once the data is connected, Tellius helps us evaluate the quality of our data by inferring the data types of each column and automatically creating descriptive statistics for each column to help users streamline initial exploratory data analysis in a simple point and click fashion.

Data Preparation

Tellius has a very robust data prep layer that provides full flexibility for business users and advanced users alike to clean up and manipulate their data prior to, during, and after analysis.

  • Business users can use a simple point and click interface to join data, deduplicate records, create new columns, and other data preparation tasks.
  • Code comfortable users can prep data in Tellius using Python or SQL.

For example, we can create a dynamic value for turnover percent. In order to do that we utilize the Calculated Column feature which seamlessly recalculates the value depending on the set of dimensions a user is looking at.

Additionally, more advanced users can utilize Python or SQL editors to enrich the table with more complex calculations.

Below is an example of applying a transformation to the BusinessTravel variable in order to clean the values from extra characters using the REPLACE function as well as creating a decile for monthly income values across all employees.

Data Exploration via Natural Language Search

Let’s look at how an HR analytics team can utilize Tellius’ Search functionality to easily explore all the relevant data without lengthy Python or SQL scripts or reliance on technical members of the team by performing calculations and generating visualizations on the fly for quick insights that can be shared with the team.

For example, we can ask Tellius to show us the attrition rate by department by simply asking the following question: “What is the attrition rate by department?”. Tellius returns the result in the form of the following chart:

We can explore if job role or income correlates with higher attrition rates by asking the question “Show me attrition rate compared to income by job role”, resulting in the following chart where the size of the bubble represents the average salary:

We can quickly assess that lower income is one of the main factors associated with higher turnover. We also see that Sales suffers from the highest turnover rate, while the R&D department tends to retain its talent longer.

The possibilities are endless when it comes to getting quick ad-hoc answers from data whether it is for the purpose of performing data exploration, reporting, or creating a data story.

 Digging Deeper with Segment Driver Insights

Let’s identify the main drivers of attrition in our organization with Tellius.

First, we ask Tellius to show the break down of employees by attrition flag. We see that the overall attrition rate is about 16% in our company.

This is typically where traditional BI tools end (answering the ‘What’ question by displaying a graph). With Tellius, however, it is possible to take attrition analysis to the next level with the power of augmented analytics Insights engine.

Simply click on the Driver button on the employees who churned (Churn = “Yes”), which automatically performs ML powered analysis that identifies the key factors contributing to the turnover.

The resulting Insight is an easy-to-read report:

 

 

Breaking down the report, we see:

 

3. The middle section provides more details about each segment. For example, employees matching Segment 2 — i.e. who work over time, have lower income, are not Research Scientists, and have worked less than or equal six years — have a turnover rate of 4.4 time that of typical attrition (i.e. 70% attrition vs 16%).

Taking Action on Employee Attrition Insights

Now that we know the attributes of higher attrition employees, we can act on them by applying the segment labels to our current employee list to identify at-risk employee groups to then consider corrective measures such as incentives (e.g. more competitive pay, overtime flexibility or reassign employees to different projects to improve engagement).

Tellius has a reporting layer that makes creating this targeted list of employees simple to get. In the example below we created a simple graph with the Driver Insights a list of matching the main attrition drivers.

We can also use the Driver Insight to look at what attributes to lead to employee retention (vs attrition).

 

In the output above, the key factors contributing to employee retentions are job role, work experience, recent salary increase, and education level. Looking at the list of factors, leadership may decide to invest more in employee education, employee promotion paths, or offer exsisting employees more opportunities switch to other internal opportunities to help them try out job roles that might suit their interests and skills sets better.

Comparison Analysis Using Insights

During the data exploration phase, we identified that the Sales team suffers from the highest turnover whereas the R&D team enjoys the highest retention rates in the company.This sparks a few questions:

  • What factors are different between the two departments with respect to the attrition rate?
  • Why do sales employees tend to leave the company more frequently than R&D?

To answer those questions, we can use Tellius’ Comparison Cohort Insight. This insight type compares two cohorts with respect to a KPI (in this case, attrition) and identifies the biggest differences. Below is the output of Tellius’ Comparison Cohort Insight:

If we compare the Sales and R&D departments with respect to the attrition rate, we can see that some of the factors with the biggest differences are stock option indicators, work life balance, and travel frequency. From this analysis, we can discern that granting stock options to the employee in the sales department, reducing the amount of travel, and improving the work-life balance may help reduce the turnover rate in the Sales department.

 Train an ML Model to Identify Employee Attrition Risk

We have heretofore performed descriptive (“what happened”?) and diagnostic (“why did it happen?”) analytics, but Tellius also allows users to easily perform predictive analytics via a robust machine learning modeling capability built on Apache spark using open-source Spark ML to train, assess, and apply predictive models. ML modeling in Tellius is no/low/full code:

  • AutoML (i.e. no-code ML modeling), where users selects a target variable (i.e. attrition) and Tellius takes care of everything else (i.e. feature transformation, algorithm selection, hyperparameter optimization).
  • Point-n-Click (i.e. visual ML modeling), which offers users more control over model selection and hyperparameter tuning approach.
  • Code (Python), where users can bring their own Python ML models to bear on the data.

For this use case we’ll use AutoML to train multiple models at once and assess which one fits our organization the best, automatically, to train a classification model that identifies the probability of attrition for each employee.

Step 1. Select the target variable, Attrition.

Step 2. Select the columns to assess and click on Predict.

Step 3. Review the model.

After the models are finished training, Tellius surfaces a model leaderboard with the most performant model’s list of input features, algorithm documentation, and model parameters.

The results of the model can be saved to a Tellius Vizpad (dashboard) to easily visualize, identify, and share to relevant parties (i.e. HR, departments leads, business unit managers, etc.)  employees at risk of leaving the firm. An example of such a Vizpad looks like this:

Conclusion

This article provided an overview of how an HR analytics team can utilize Tellius to perform a holistic analysis of employee attrition, including:

  • Code-free exploration for ad hoc answers to questions via NLQ.
  • Key drivers of employee turnover, differences between retained vs quit employees,
  • Likelihood of employee turnover using a predictive model

…as well as how to make any necessary changes to the data and visualize/share results with key stakeholders in the form of an actionable dashboard.

Each functionality covered in this article is an integral piece of the complete People Analytics solution focused on analyzing and reducing employee turnover, de-risking employees before they leave the company, and improving the overall performance of the organization.

Learn more about other use cases for Tellius or take it for a free 14-day spin today!

Try Tellius with Your Data

Start for free

Learn More

  • CLTV 125
    Use Case

    Analyzing and Predicting Customer Lifetime Value (CLTV)

    Calculate, analyze, and predict customer lifetime value (CLTV) faster and easier with AI-Driven Analytics

    Read Now
  • Tellius announces live insights
    Blog

    Tellius Announces Live Insights for Cloud Data Warehouses

    This new capability leverages the compute power of cloud data platforms for automated generation of advanced insights without requiring data to be extracted.

    Read Now
  • Web Promo Social Cards (6)
    Webinar Replay

    Trends in Data Analytics & AI-Driven Intelligence for 2022

    2021 has been a year of innovation and evolution in data analytics, and 2022 will only see more dramatic shifts ahead.

    Watch Webinar