Secret of My Success: Minimizing Customer Guesswork
What made you turn to business analytics?
In 2002 Cablecom wanted me to set up a new marketing database, which has about 7 million people in it. We use approximately 120 different criteria, such as demographics, customer behavior, and transactional history, to segment our customers. At first, we were using some basic database software to support our mailing and outbound marketing--very simple stuff. We wanted to go one step farther and start doing some serious data mining.
How did you select a vendor? We were using a Canadian software package, called Angoss, for our data mining. It was limited in its ability to pull data from multiple sources, so we needed something else. We started looking for more advanced analytical software packages and began our own vendor evaluation process. It came down to SAS and SPSS. We chose SPSS for a number of reasons. The total cost ownership was lower, the ability to marry data from different sources was better, and the graphical user interface was easier and more intuitive. So in 2003 we implemented Clementine 7.0. We also implemented SPSS Statistical Services software, since these capabilities weren't as good on Clementine 7.0 as they are today.
What are you using business analytics for? We use Clementine and Statistical Services to determine three main trends about our customer base: propensity to purchase, customer churn, and segmentation of the customer base. We had a lot of success doing this throughout 2005. Then I went to a summit in Europe and found out about SPSS's Dimension product line, which includes a product called mrInterview. It's designed to create and manage customer surveys. I thought it would be great to connect the survey platform to Clementine and Statistical Services. So in 2005 we implemented mrInterview.
How do you survey your customers and use the subsequent data? We survey our customers three ways: over the phone, via email, or in the field from PDAs. We use the survey data to give each customer a satisfaction score. We then compare that against data mining models produced by Clementine to make sure we have a clear picture of our customer base. A good example of this was when we noticed an unusual pattern in the life cycle of our customers. At around the 12-month point we had a peak in customer churn. We did some data mining using Clementine and found that on average, most customers would begin inquiring about canceling their services with us at the nine-month mark.
Now, we send out a customer satisfaction survey at the seven-month mark. If we have 20,000 customers hitting that mark a month, we send out 20,000 surveys. Maybe 5,000 will respond. For those who do, we collect the data, give each customer an average satisfaction score, and come up with a list of satisfied customers and unsatisfied customers.
We can then match the profiles of those who do respond to similar ones of the 15,000 who don't. This allows us to come up with a list of satisfied and unsatisfied customers for the remaining 15,000. It takes into consideration their answers, their profile, prior transactional history, et cetera, and compares that to similar customers so we can gain accurate insight into future actions. The idea is to measure, understand, and predict your customer.
Related Articles
Customer Insights Breed Customer Loyalty
18 Jul 2008
A new Aberdeen report says profitability is dependent on repeat buyers and a focus on lifetime customer value.