- For its key operations, the company relies on workers who require extensive training which takes the course of a year, starting with classroom training, on-the-job training, and then a probationary period until they are fully certified to engage in operations. During an extended period of business growth, the company hired thousands of workers, but determined that one-third of new hires attrited during the first three years, with the majority resigning or being terminated within the first 12 months. The financial impact was quantified as tens of millions of dollars, and it hindered their ability to build sustainable capacity.
- We gathered data from dozens of sources, including operations, human resources, training records, test scores, employee surveys, scheduling, weather, and so on. Over 100 potential factors in driving attrition were identified in conjunction with subject matter experts. These were organized into a taxonomy, with categories such as employee profile, training and qualification, working conditions, operations, teams and supervision, economic, etc.
- Each of these factors were tested with models to quantify the impact of each. Besides traditional statistical modeling, machine learning was employed to guide the analysis, and text mining was used to extract information from training evaluations, employee surveys, etc.
- The model was used in two senses. At a strategic level, it directed senior management to those factors within their control which could be addressed by changes in programs or policies. At a tactical level, scoring logic was developed to provide a risk score of attrition for the current workforce, at the employee level.
- The model was able to predict 28% of resignations and 43% of terminations of the riskiest 10% of employees within the first three months of their tenure.