Zurich-based Cablecom is the largest cable network operator in Switzerland, providing customers with cable TV, broadband Internet access and telephone service. This essentially retail outreach masks a powerful b-to-b underpinning: 70% of the company's TV service contracts are with apartment building landlords, who then bill the final users.
But despite its market leadership, the company faced challenges. Prior to 2005 its marketing outreach was focused on acquiring customers, with little attention paid to retaining them. When the market reached a saturation point, Cablecom began to focus on ways to reduce churn and undertook an initiative to expand its customer analytics staff, invest in supporting software and establish new procedures to keep customers from leaving.
The connection between customer-service analysis and marketing is clear to Federico Cesconi, Cablecom's director-business intelligence.
“It's very difficult to win back customers after they've left you,” Cesconi said. “And in Europe, the win-back rate is only about 10% to 15%. So, our intent was to assess the satisfaction of every single customer.”
Industry statistics indicate that unhappy customers begin to complain at around nine months into their contract and abandon contracts at about 12 to 14 months. Considering this, Cesconi decided to analyze customer attitudes at the seven-month mark. The challenge was how to assess real sentiment from up to 40,000 feedback responses per month.
There was no way to analyze that many open-ended, often rambling responses except through automation, Cesconi said. In early 2007, he turned to Chicago-based SPSS Inc.'s Clementine text-mining solution to analyze customer views collected via the Web, automated phone surveys, tech help desks, call centers, sales and other points of end-user contact.
“That was our original, small idea,” Cesconi said. “To put the customer point of view in all our processes. Of course we worked with specific unsatisfied customers, but in addition we categorized the drivers of satisfaction and dissatisfaction to understand where the gaps were in our operations.”
Cablecom put text mining to work analyzing “unstructured” data, such as free-form responses to open-ended questions or call centers. It also used the SPSS tool to analyze audio content coming in from the company's interactive voice response systems after converting the MP3 audio files into text.
The analysis revealed that customers felt too pressured by high-powered sales approaches. In response, Cablecom revised its sales and marketing. The feedback also gave the company critical information on specific product features, aiding design and manufacturing, as well as customer relations and marketing, Cesconi said.
To predict his ROI, Cesconi conducted an experiment using a control group and a test group of end users. The test group's feedback was analyzed via text mining while the control group's was not. The control group's churn was 14%; the test group—with customer-service follow-up—saw a churn of only 2%.
Following the test, text mining was applied broadly. Cesconi said that after six months of use, Cablecom's new approach was able to pinpoint unhappy customers and increase satisfaction in 53% of all cases.”
“And that really is improving retention because there is a strong correlation between satisfaction and churn,” he said. M