It's no accident that JetBlue continuously tops J.D. Power's airline satisfaction survey and is one of only two North American airlines to receive four-stars from Skytrax's World Airline Star Ranking. Such distinctions are largely due to substantial analysis of the airline's customer and operational data.

"Our sample set is so large that we're able to do significant analytics around it and find out what's important to a larger set of our customers," says Bonny Simi, director of customer experience and analysis at JetBlue.

In doing so, JetBlue relies on Net Promoter Score (NPS) as its gauge of customer satisfaction. The airline's belief is, the more promoters it has, the fewer customer will defect. "It's about measuring and monitoring what they like and doing a lot less of what they don't like," Simi says. "It's about driving more promoters and fewer detractors."

To keep a vigilant eye on its NPS, the airline analyzes volumes of data monthly. It gathers 50,000 email survey responses, results of text analysis, social media data, and Web mentions, and then analyzes those findings to uncover customer pain points. Next, the airline combines that data with operational and commercial data. On the operational side, the airline determines elements like whether the airline is keeping its customer promises and how specific teams are delivering service. On the commercial side, it analyzes such data as whether the TVs work and airplane turn times.

Next, Simi's team drills down into three areas: C-level, departmental, and front line. At the C-level, Simi's team analyzes high-level macro issues; at the departmental level, they analyze elements like comparative airport data; and for the front line, they may look at flight attendant data to determine which employees require additional coaching. Once all the data is analyzed, JetBlue provides the necessary training or makes the appropriate changes to move the needle on its NPS.

Additionally, JetBlue applies predictive analytics. Annually, Simi's team combines the customer, operational, and demographic data with historical data, like on-time performance, and sets monthly NPS targets. Then if the team sees that, for example, it's likely that NPS won't hit a sufficient target six months down the road, they mitigate that through treatment strategies. In doing so, JetBlue has been successful in reversing the decline of NPS in specific cities and regions.

The entire West region, for example, was experiencing declining NPS, so the airline changed signage in airports, fixed public address systems, and even convinced the Transportation Security Administration to add more lines. In January the West region jumped from the lowest-scoring NPS region to the highest-scoring. "You don't get that by having a great recipe, you get that by changing the recipe over time," Simi says. "You can wow somebody once, twice, or maybe three times, but after three times, they're not wowed anymore so we're constantly looking to get better."

What you find depends on where you look

JetBlue is taking a holistic view, applying data from several areas to get a more accurate view of potential churn. The airline is enhancing that data by adding voice of the customer (VOC) insight. According to Chris Cottle, executive vice president of marketing and products at Allegiance, the ability to collect unsolicited feedback and combine those elements with business data is critical. "If you can combine voice of the customer data with operational, CRM, and financial data, then you've got a really unique area that we refer to as VOC intelligence," Cottle says. "At the heart of it is a value proposition of having greater insights."

Dhiraj Rajaram, CEO of Mu Sigma, agrees that companies need to invest in the right data, and that traditional sources might not suffice. Businesses must understand customers' needs, and conduct real-time monitoring in the contact center and in social media. "Understanding the customer is the key," Rajaram says. "From knowing who your customer is to figuring out what his or her current sentiment is becomes they key in building good customer relationships."

Another data source for churn propensity, one that is often overlooked, is new-customer behavior. Connie Hill, president of Veracentra, says that while understanding every customer at every touchpoint is crucial for churn prevention, stopping customer defection happens back when a company first onboards a customer. "You can conduct analytics to understand if customers are at risk of defection, but typically what you see are the [churn] drivers being a lack of engagement or purchasing," Hill says. "To prevent churn, start at the beginning of the cycle and take steps appropriately to get value from customers."

Changes in engagement or purchasing, as well as finding patterns that lead to those changes, is also valuable insight later in the customer lifecycle. "Companies try to identify drivers of attrition by analyzing the transactional behavior and demographic information of customers who have defected in the past," Mu Sigma's Rajaram says. "Reduction in current spending and reduction in frequency of store visits can all serve as valuable clues as to what customers are likely to do in the future."

Dan Thorpe, senior director of advanced analytics in insights and innovation at Sam's Club and former senior vice president of statistics and modeling at Wachovia, says that a newer metric that banks in particular are putting into place is a "new/lost ratio," the ratio of new customers to lost customers (the definition of lost customers varies by industry. In a membership organization, one way to define a lost customer is one who does not renew membership. In financial services, it may be a customer who has dropped three of his seven services.). But according to Thorpe, the ratio alone is not enough, as it does not immediately reflect churn. "We also need to understand and take into account the tenure of a customer," Thorpe says.  "The struggle is to understand what is a lost customer and how can you predict it if you're losing a customer."

Anyone can build a downgrade model, or a departure model, or a model that will predict churn," he adds. "The trick is predicting it early enough where you can take action."

The power of predictive analytics
Thorpe says that most industry metrics are backward looking. Banks, for instance, can calculate a loss ratio to determine churn, but what matters most is achieving foresight. "We want to predict where [churn] is going in the future so we can take future action now. That's where the financial services industry has to go…The big move will be to forecast your churn statistics, understand what's influencing that, and take action now for future value, not just immediate value."

Predictive analytics is essential in every industry, and can help companies assess not only churn, but also changes in customer value. Veracentra's Hill says that predictive analytics help move the needle to understand if the customer is becoming less valuable over time. "It's like a shiny new penny, she says. "Everyone wants to leverage their data, but if you can use predictive analytics to [enhance] strategies, then you can leverage that in a more powerful way."

Hill adds that effective churn practices depend on the creation of a baseline—understanding current churn and retention levels. Then companies must work to analyze customer behaviors, design a customer contact strategy that helps to improve retention, measure the results of that strategy, and create a roadmap for continued success. "Basically what you want to do is try to keep it simple," Hill says. "Once you understand what data can do for your programs, create a roadmap for continuous improvement."