Leveraging a custom Employee Satisfaction (ESAT) Analyzer, the client was able to reduce attrition by 5% by targeting at-risk employees and conducting “Stay Interviews”
Using an automated dashboard to track progress, the HR team was able to conduct more than 100 “Stay Interviews” within a month
A global beverage manufacturer was grappling with employee attrition across multiple branches, with an annual attrition rate between 5% – 15%, across zones, for a total of approximately 20,000 employees. The HR team were unable to gauge and effectively act upon cases of employee attrition ahead of time as existing feedback loops were inefficient and risk assessments were highly subjective. As a result, any interventions to retain employees were undertaken too late & were often ineffective in making a difference and preventing attrition.
The client wanted to plan timely employee interventions to prevent attrition. To enable the same, an ESAT Analyzer was created, using a combination of ML techniques and technologies, to achieve the following:
1. Predict risk of attrition for each employee
2. Highlight the employees that need immediate attention by calculating the level of attrition risk
3. Identify the key drivers of attrition for each employee
This would enable proactive interventions from the HR team, through an in-depth understanding of the pain points of at-risk employees.
The ML-powered custom AI application created for the client used a basket of ML classification techniques, which assessed pertinent factors including historical staff turnover data, performance and appraisal information, manager attributes, employee attributes, and employee goals and attainment.
To set the solutioning in motion, the team first put together the required data. The data was pre-processed before being run through the ML classification framework.
Fig 1: Overview of the ML Classification Attributes
Data Pre-processing: TheMathCompany worked with the client to collate the needed data from the respective stakeholders. Then, existing, disparate data was analyzed to identify areas that required corrective measures and an improved quality of data and data collection.
The data was cleaned thoroughly, in adherence with data compliance regulations, and thereafter used for feature engineering.
ML Classification Framework: A custom ML framework was set up, which utilized the Basket of Algorithms approach to automatically custom-select the best algorithm for every model build.
Attrition Analyses: By using a Champion Algorithm, results were generated to identify at-risk employees, key drivers of attrition, and each driver’s respective contribution to attrition rates. For instance, analyses revealed that peer compensation contributed 19% to attrition, while career progression in comparison to peers, manager performance, and goal achievement contributed 17%, 11%, and 8% each.