1to1 Magazine - November/December 2008 - (Page 47) EXPERT INSIGHT one-on-one with Kevin Zimmerman As the World Churns 1to1 speaks with Kerem Can Özkisacik, senior consultant at Peppers & Rogers Group Turkey, to discuss the value of churn analytics. For many people, the term “churn” is closely associated with the telecommunications and cable television industries, sectors where rampant customer turnover is too often a way of life. But as Özkisacik notes, the basics of churn analytics can be applied to most any industry. As the old saying goes, the cost of retaining an existing customer is less than that of acquiring a new one, and churn metrics can be a critical tool in making that axiom a reality. “Companies can retain 20 to 25 percent of their revenues by protecting just 5 percent of their most profitable customers.” You recently completed a churn analytics study for a major telecommunications company doing business in the Middle East. What was the situation? The company was experiencing a churn rate of 25 to 30 percent, which was enough to effectively mean no ROI on new customers. Its executives also noted that it was becoming increasingly difficult to capture new customers, because its competitors were becoming more effective in countering any new initiative, sometimes in a matter of hours. And this is a business where newcomers are less profitable than customers with higher tenures. into the future. Using this model it is now possible to take early measures to retain especially high-value customers. Ultimately we were able to combine the churn scores with each customer’s value to identify the high-risk customers, and to take proactive retention actions to counteract churn. What about a loyalty program as a way to reduce churn? In some cases [that would work], but not this one. When we were doing the initial analysis, we selected customers both from within the loyalty program and without, and compared their churn rates. There was no statistical difference between the two groups. In some instances the churn rate was even higher for the loyalty members, which was an astonishing result. The company is now redesigning its loyalty program to work better in this regard. It was not targeted and not very well structured. What other industries have you conducted churn analytics studies for? The automotive industry, supermarkets, and other retail entities—it’s quite common. It’s possible to predict churn behavior for almost all industries. And not just for B2C customers; business customers can be used as the main group to predict churn behavior. It’s also important to be able to predict churn behavior of small-business owners. We are developing some models to better understand those behaviors as well. Are there any general findings from this particular study that might be applicable to other industries? First, operators should build predictive churn models analyzing potential customer value versus the degree of interactions and frequency with the firm to target at-risk customers. Companies can retain 20 to 25 percent of their revenues by protecting just 5 percent of their most profitable customers. Operators should work to control their service-cancellation processes as well. For telecoms, specialized call center teams should be employed to address exit calls and influence the customer’s perception of the service value to retain the revenue base. In mature telecom markets, on average, it takes three years to pay back the cost of replacing each lost customer with a new one. Contact Özkisacik at kozkisacik@1to1.com With a customer base that’s largely going the prepaid route, wasn’t it difficult to determine what the churn rate actually was? The company defined a “churned” customer as one who had been inactive or had low activity for at least 90 days. The problem with that is, if you’re not proactively acting on them before the 90 days are up, they’re long gone. For the churn prediction model, we decided that a customer would be “inactive” if he did not make at least three transactions—voice call, service usage, SMS, etc.—or a recharge during the target period. What was the next step? We conducted workshops with the company’s relevant business and analytical units on churn management strategies and churn modeling techniques and processes. Then we started on existing churn management activities—defining churn, setting data requirements and business priorities, and building churn prevention programs. This led to the identification of focus points and the required data set for analyzing each of their 5 million customers’ potential churn behavior. Based on these definitions, the company prepared a churn-modeling data environment and PRG started to develop a predictive churn model. We ended up developing a “traffic light” approach to segment the customers: A red flag indicates that the customer fluctuates significantly from his usual behavior, an orange flag means that he fluctuates at a lower level, and a green flag says that he behaves as expected. Then, as a complement to the first model, we developed models for estimating the churn probability of each customer, which will be used continuously and form the basis of all retention management activities 1to1 magazine 47 http://www.1to1media.com/view.aspx?ItemID=29293
For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.