In 2002, Nextel's churn rate was in the mid- to upper 2% range -- which means more than 2% of customers left each month. That may seem small, but over the year it adds up and eats into profits.

Now, Nextel is an industry leader in retention, with churn at only about 1.4% a month.

How? We interviewed Scott Radcliffe, Nextel's Director of Decision Sciences, Customer Lifecycle Management Group, to find out.

"You have to be able to identify which customers are at risk, why they're at risk, and what you're going to do about it," he says. "Theoretically, what we try to do with analytics is to bring some order to that."

Radcliffe joined Nextel to build a customer lifecycle management team to reduce churn among Nextel customers, and to make customer lifecycle management an integrated business process -- fully aligned with marketing, product development, customer service and other business practices.

His team focus on more than just reducing churn for churn's sake. "Are we keeping customers, and are we keeping profitable customers?"

By gathering customer data across all touchpoints Radcliffe and his team learned to treat customers based on their churn risk and value to Nextel -- not based on the size of the account alone.

Here are specifics on the three steps they used in the process:

Step #1. Define drivers of churn

To improve retention, you must first know which customers are at risk of bailing; otherwise, you're spending retention dollars on customers who plan to stay with you anyway.

Radcliffe's team has identified a dozen drivers of churn. Using a churn model, they look at these drivers in combination with each other to determine if a customer is at risk of leaving.

Drivers include:

--Changes in usage pattern over the last 90 to 180 days. "Some people, if their usage goes up, it's risky, because the next factor is, how does your usage match your rate plan? If your usage has gone up and you're blowing out your rate plan, you're likely to churn," says Radcliffe. On the other hand, if usage goes down and the customer is using far less time than his contract allows, he may be thinking, "I don't need this."

--Business customer vs. consumer. "We're primarily B-to-B but we have a substantial B-to-C base as well. Consumer churn is higher than business churn."

--Network performance. By looking at blocked calls or dropped calls, Radcliffe's team can understand at least partly how the customer views the network. Poor network performance increases churn possibility.

Step #2. Build customer care center slowly

"You can't take a care center and say, we're going to make this a combination care center and cross-sell center," says Radcliffe. "You have to develop a team of people."

Nextel began by drafting people from the general care side; some had been in customer care but want to be sales people, others had a strong proclivity for sales. They made those people a separate group.

There are now teams in Nextel's cost centers who do nothing but service those customers who have exhibited behavior characteristics that indicate they may be ready to leave, Radcliffe says. Those customers are flagged within the database, and when a call comes in from one of them, they are immediately transferred to the retention program team.

(Note: this is not an outbound campaign. The retention program is based solely on customers calling in to Nextel.)

Depending on the value of the customer, the sales person is empowered to offer different benefits: a higher-priced handset, for example, "as opposed to the undifferentiated strategy, where I pick a middle handset because I think that's all I can afford, but then I have a lot of high-end customers who go away because they don't think I'm valuing the relationship," Radcliffe explains.

Some lessons learned:

--Sales rep performance cannot be based on number of contracts signed a day, but on survival rate of those contracts over time.

--Though employees are trained to make some decisions, "a factor in this program is you essentially have to do as much decisioning behind the scenes as possible," says Radcliffe. "The customer care rep has to be directed to offer X or Y, but they're not sitting there making the decisions."

--Some campaigns can actually harm customer retention and increase churn. Occasionally, there have been times when churn increased in certain segments, because what the customer care rep was offering actually reminded customers of an unpleasant experience they may have had with Nextel. "So we needed to bring together information about [a customer's] past calls to care and what the results of those calls were."

--A culture of employee-staffed call centers is important. "They have much more of a vision of trying to get [customers] to complete a transaction that makes them more 'sticky,'" rather than simply trying to get people to sign a contract, says Radcliffe.

Step #3. Educate marketers about the uses of analytics

When marketing managers come to Radcliffe for analytics, he requires them to go through a process to help them understand exactly what information they need from him. "I don't just take an order for a predictive model or report," he says.

The first thing he asks is for the marketing or sales channel manager to identify the business opportunity. "You'd think any business would make people do this, but that's not the case."

Then he asks for a face-to-face meeting in which he walks through the business process for which the manager wants analytics. "Marketing or answering phones or servicing customers is a process, so what we talk about is, where are the potential optimization points?" he says. "Is it response rate? Amount of people under contract? Profitable mix of handsets to sell or give to a portion of the population?"

He adds, "We want to optimize how customers respond to our action -- or, in some cases, how we respond."

Take, for example, the service and repair call center. "They've always got plenty to do, so who am I going to try to get to first -- a profitable customer who's been using the same handset for two years, or someone on the least profitable price plan who tries to get a new handset every 90 days?"

When Radcliffe walks through the processes with business managers, "We'll sit down and whiteboard that thing," he says. Together, they find the optimization point and put together a business case for the analytic report and how it will add value to the campaign.

Another example: Nextel's web services. There's a certain type or profile of a customer who uses that capability.

Nextel obviously can't call all customers who don't currently make use of the capability to offer it to them. "So I'll optimize all the contacts by building a profile of customers who are strong users of Nextel online, and go look for the people who look like [the profile] but who don't yet have the capability," he says.

A final word: When investing in analytics, be sure to do it right, says Radcliffe. "You want a broad view of the customer, and you want the ability to pull it all together, so that takes the right kind of analytical system."

Useful links related to this article

SAS, the analytical system Radcliffe uses http://www.sas.com/index.html

Nextel http://www.nextel.com