A perpetual obstacle for any call center manager is the revolving door of agent attrition. Each lost employee represents recruitment and training dollars flying out your front door. If only there was some way to detect that an employee was about to leave so that management could intervene, attempt to address the employee’s concerns and keep him or her a part of the team. Where can one buy such a crystal ball to detect the future?
Recently, here at West, we began investing in predictive analytics to identify the behaviors consistent with employees who have left the company to see whether those behaviors represent a potential risk of churn from current employees. How did we do it? We pulled three years’ worth of data across our thousands of agents — attendance trends, efficiencies, key performance indicator (KPI) adherence — and we looked at the employees who left to find patterns of behavior.
Some of the nuggets we uncovered include the following:
- Higher wait times between calls lead to lower attrition
- Agents who can flex their schedules up or down are less likely to leave
- Drastic changes in schedule (uptimes/downtimes) represent a risk of increased attrition
How well can we predict future churn? The model used evaluated more than 400 variables, and over 75 percent of future churn could be isolated to 20 percent of the workforce population.
So, imagine you have a workforce of 100 agents. The model we’re using can isolate 75 percent of your future churn (churn that would occur within the next 30 days) to just 20 agents. Wouldn’t you want to know which 20 of your employees were most likely to leave in the next 30 days?
The predictive analytics can also identify potential reasons for attrition. What we see trending includes on-boarding, KPI performance, schedule changes being denied and recent disciplinary action. There are two ways we’re using this information. First, we schedule an intervention with employees identified as likely to churn to try and determine what management can do to help. In the first 90 days of our study, an employee who had an intervention was three times more likely to stay than an at-risk employee who was not provided with an intervention.
The second way we’re using this information is to evaluate our HR practices globally to identify changes that can reduce attrition. As an example, we identified that we needed to make changes to our disciplinary action process. Of all terminated employees from our three-year study, 41 percent of employees who received a mild form of disciplinary action left within three days of receiving it. That accounted for 9.5 percent of all churn. Those leaving within one month of the disciplinary action represented approximately 20 percent of the churn. With numbers this shocking, we revamped our disciplinary action procedures, and over the last 90 days we have seen a significant reduction in churn. Our annualized attrition has been projected to drop by 20 percent — leading to a savings of $2.2 million in training expenses.
You don’t need a crystal ball to see that your call center needs to begin using predictive analytics to reduce attrition.