From Gut Feeling to Informed Decision: Journey of HR Analytics
From Gut Feeling to Informed Decision: Journey of HR Analytics
Using the right modelling system along with the correct data can go a long way toward decreasing your company's attrition rate.
Join the DZone community and get the full member experience.Join For Free
The Agile Zone is brought to you in partnership with Techtown Training. Learn how DevOps and SAFe® can be used either separately or in unison as a way to make your organization more efficient, more effective, and more successful in our SAFe® vs DevOps eBook.
“Your number one customers are your people. Look after employees first and then customers last.” – Ian Hutchinson, author of People Glue
Rewind to a couple of decades ago. Collecting data to help design and restructure policies to keep employees happy was an uphill task. But, now with technology enabling easy data collection, the task is even tougher to analyze mountains of data and draw actionable insights.
Human Resource analytics is rapidly evolving and opening new opportunities in areas of employee engagement, boosting productivity, recruitment, closer collaboration with other departments, etc.
The operational definition, “Systematic identification and quantification of the people drivers of business outcomes” articulates the power of HR Analytics.
While happy employees turn out to be the backbone of happy customers, “building projects around motivated individuals and giving them the environment and support they need, and trusting them to get the job done should be the mantra."
While it is agreed that regular pulse conversations, rewards, and recognitions, along with challenging opportunities, play a critical role in employee engagement, it is equally important to understand the intrinsic motivation of employees. Prevention is better than a cure, hence it is vital to forecast attrition and proactively mitigate the risks through various measures.
While HR analytics addresses quite a lot of challenges associated with people, it primarily tackles the below top two questions every organization is interested in:
- The attrition trend in my organization in next fiscal year
- Likelihood of key performers leaving my organization
While qualitative assessment is important, the below listed data-driven approach enhances the prediction and helps the organic growth of an organization.
Before we proceed further with our forecasting and prediction, it is important to understand the business problem, data source and critical factors that influence the employee attrition. The article ‘Forecasting the Future: Let’s Rewind to the Basics’ covers the basic aspects to be considered.
Root cause Analysis
Cause-and-effect diagrams always come handy to identify all potential problems for any challenge. It gives a visual representation of all data to be collected for further analysis & analytics:
Data Collection & Data Engineering
Human Resource departments typically gather a lot of information such as demographics, skill sets, performance, learning, growth, and pay. Based on the cause and above diagram, the data must be aggregated and grouped to be able to do correlation (relationship among various variables) and covariance (variance of random variables).
At times the simple plots and trends are good enough to convey important key takeaways. Hence it is important to represent the data visually prior to applying different models.
Time series helps in addressing the first problem, finding the attrition trend in an organization in next fiscal year. Given an observed sequence based on historical data, time series analysis helps in building a model that can predict what comes next. Though our use case here is on attrition for most of the scenarios where we want to forecast based on historical data, time series analytics comes handy. To learn more about Time Series and implementing it in R, check out one of my favorite little book series.
Time series analysis decomposes the historical attrition data into various components: observed (given data set), trends (long-term progression of the series), cyclical (repeated fluctuations), seasonal (seasonal fluctuations), and random (irregular). Less randomness in time series helps in better forecasting accuracy.
Various models such as Holt-Winters, Auto-Regressive Integrated Moving Average (ARIMA), automated forecasting using exponential model, or Auto ARIMA with simple moving average can be applied and the optimal one can be chosen through minimal forecasting errors (Root Mean Square Error, Mean Absolute Error, Mean Percentage Error, Mean Absolute Scaled Error, to name a few).
The model which we have chosen based on minimal forecasting errors can be applied to predict next data points. In the below example, the data points 1 to 110 are from historical and data points 111 to 118 are predicted outcomes for next 8 periods.
To solve the next challenge “likelihood of key performers leaving the organization”, we can leverage both explicable models (linear/logistic regression, decision tree, etc.,) or non-explicable models (random forest, neural networks, etc.,). While we may get better accuracy leveraging non-explicable models, the explicable models are suggested to be able to communicate the outcome better.
Like what we did in time series analysis, we need to choose appropriate error metrics before applying the model for better prediction. In our problem, probably the higher sensitivity (when it’s attrition, how often does it predict attrition?) is more important than lower false positive rate (when it’s not an attrition, how often does it predict attrition?).
More than regression, a classifier such as decision tree will be more powerful in building better representation as this will also help in actionable mitigations (better growth opportunities, work environment and rewards and recognition).
In summary, Human Resource analytics, especially attrition forecasting and prediction, helps HRD by providing actionable insights towards strategic decisions and optimizing the business outcomes.
“The time is now for organizations to take the utilization of data to the next level.” – Ayman Sayed, President and CPO, CA Technologies
Opinions expressed by DZone contributors are their own.