Data detectives

I knew that dame was trouble the minute she called. I'd seen her type a thousand times, calling these same lines, looking for the same deals. Sure, she looked great on paper - good family, good neighborhood. But she was a jumper, ready to cut out on our services at any time to take up with some other carrier down the block that walked a little smoother walk, talked a little smoother talk. And who could blame her - staring at all those glitzy offerings every night? Yeah, lucky for me, I knew the dame was loaded.

Competitive markets have changed the way carriers look at the world. They're worried about losing customers at every turn. And banking on that film noir notion that a mix of smarts and toughness always wins, they're turning to data warehouses, data mining and good old-fashioned investigation, hoping to scrape together enough useful information about their customers to keep them around and sell them more along the way.

It's questionable whether all of the hype about data warehousing and mining is true (see related story on page 26). Currently, statisticians are riding high in the tech-savvy telecom market, building huge data warehouses and buying huge analysis engines because CEOs believe in the technology (Figure 1). Carriers such as U S West, Bell Atlantic and Alltel claim that they are seeing results from their labors, but the techniques still are in their infancy. The technology has just started to catch up to the minds of the mathematicians behind the algorithms.

But if everything goes their way, service providers will be able to give customers what they want, before they ask for it.

The start of a beautiful friendship

U S West's move into data warehousing and data mining is a story like most others. The company was sitting on its monopolistic mound, king of its world, until competition came rolling into town.

"U S West traditionally has been a monopoly. As a result, we were very product-centric," says Jovan Barac, acting executive director of database marketing at U S West. "We didn't try to get customers because they didn't have a choice [in service providers], so we focused on the product."

Previously, all information was controlled by the IT department. It took too long to get the information for it to be useful. "When we built a model, we actually had to write the code to implement it, so it took two to three months to score," Barac says. By setting up a data warehouse and working on data mining, the company is hoping eventually to focus all ofits energy on the customer, making products for them rather than trying to make them like the products.

To pull this off, the company has built a data warehouse that uses the SAS Institute's decision support system to score its data and uses Exchange Applications' Valex campaign management software to interpret it. The hardest part of building any statistical system is to make the raw data usable, says Mark Brown, program manager at the SAS Institute. Figure 2 shows the company's data modeling and scoring scheme.

U S West is still new to the data warehousing game. Halfway through the three-year project, the RBOC has just moved passed the initial challenges and is starting to realize the benefits of such a system. But a quick high from a data warehouse is common. "There is an initial wow," Brown says. "But the profitability is to do this over a long period of time."

Bell Atlantic's early days were similar to U S West's. The company realized it had little islands of data from various systems strewn all over the place, which made the information useless. The RBOC needed to combine that data in order to reach the customer - especially because it's in one of the most hotly contested competitive local exchange carrier areas.

"Before it would have been, `We can make it, go sell it.' Now it is from the customer's standpoint," says Sharon Mosse, vice president of market research and intelligence for Bell Atlantic, adding that her company got into the data warehousing game a few years ago. "Having all those different resources and databases was silly, so we combined them. We wanted to make sure that we were looking at one customer."

The company wanted to get all of that good information, which was collecting dust in the information systems department, and put it in the hands of the marketers, who could put it to use to stave off competitors. "We did an inventory [of our systems] and worked very closely with the information systems department. We became partners in the development and the marketing," Mosse says.

Total investment in the system was fairly inexpensive. "We basically needed to connect the dots and get the software," Mosse says, although she would not discuss specifics of the system.

Now Bell Atlantic, like most others, is looking to move forward with its system so it can create relationships with its customers, based on call records. "What we're seeing is a switch from [carriers thinking], `How do I get all this data from this data jailhouse into one place?' to `How can I use this data to help my business?'" says John Santaferro, global marketing manager for Hewlett-Packard's business intelligence group. "Their service has become a commodity, and they have to build a relationship with their customers." Figure 3 illustrates HP's call behavior analysis solution.

Just the facts

On the older side of database marketing is Alltel. The progressive company built its first data warehouse in 1992 and began segmenting its customers. Initially, the company started its data warehouse for the same reason that the others did: to bring together disparate information to make it useful. "We saw that there were lots of islands of information, and we wanted to get everything together in one place," says Colleen Sullivan, director of information technology.

Alltel wanted to figure out what types of customers it had, and the majority of information came from the billing system. In the early days of the data warehouse, the information at Alltel still was controlled by the IT department, which allowed for a more complex interface. But now Alltel has moved away from that system, allowing more marketers to do research themselves, Sullivan says.

Alltel has between 60 and 100 employee users. Although the data warehouse is being used more, the number of users has not grown very much over time. Sullivan divides Alltel's users into three different segments: casual users, who conduct two parameter searches; sophisticated users, who perform simple SQL searches; and power users, who do advanced SQL searches.

Often, companies don't realize all that they can do with a data warehouse until after it is installed. "You don't see all the hidden magic until you get it up and running," Sullivan says.

The most important aspect in data warehousing and data mining is for people to put their faith in the information, she adds. "Your warehouse is only valuable if people trust it. It is important to foster a trusting relationship between users and [the IT department]." To accomplish this, Sullivan continually tries to educate the users and enhance the system.

As carriers continue to develop their data warehouses and make more sense of the data, they realize that they don't need every bit of information in the system to make it worthwhile. There is a limit, and data about customers has to be manageable to be useful.

"Your initial hunch is to put as much information in as you can and see how it works," Barac says. But during the past year, U S West learned to be more careful. It is not a question of whether the technology can handle it, but whether putting all of that information together actually is useful, he adds.

"You have to be scientific about your data approach. You can drown your users with too much data," Barac says.

Most service providers have reduced the amount of data in their warehouses by restricting the amount of historical data in the system. Information older than 24 months usually is not included. And from that controlled information, only samples are taken to reduce the time involved in analysis. This usually doesn't effect the usefulness of the information because the sampling information is accurate, says Teresa Wingfield, research director at The Aberdeen Group.

Tying it together

The basic drive in all information systems is to move faster, and data warehouse developers have been trying to decrease the amount of time between the conception of an idea and its fruition.

Closed-loop marketing is helping them move in that direction (Figure 4). The idea is simple: During any customer contact - direct mailings, customer care calls, telemarketing or a customer's random visit to the carrier's Web site - a carrier should be able to target products to the customer on the fly, using all the information from the system.

Carriers strive to create a system that connects all inbound and outbound traffic - one that is always accessible. "You have to think about where your opportunities are to deploy your marketing," says Robert Rich, director of product marketing at Exchange Applications.

Using closed-loop marketing software allows companies to continually track their marketing campaigns while the campaign is in progress, rather than waiting for a few months to scrape together information from the data warehouses. With an almost instantaneous turnaround of information, companies can react to the results of their campaigns, Rich says.

There are two drivers for developing a constantly updatable system, says David Dague, director of closed-loop systems at American Management Systems. "Information technology has evolved to a point where the [carrier] can actually use [the data], and the competitive model is making them focus on their greatest asset - the customer."

Besides directly marketing to customers with closed-loop systems, carriers can reduce the amount of churn that occurs, which is the "cancer" of the industry, says Clifford Wagner, senior vice president for the Americas at MSI, a developer of wireless software. Using software such as MSI's Ceer or Exchange Applications' Valex, carriers can look through the eyes of the customer rather than looking at the system from a technical standpoint. With this knowledge, the idea is that the interaction between customers and carriers will improve continually. "Information is utterly worthless unless it is processed in a way that makes the customer experience superb," Wagner says.

Carriers are excited by the possibilities of closed loop systems and data marketing in general. But the most important thing about data systems is that no matter how much information is given to customer care representatives or marketers, it is only useful if they can make sense of it and act on it. "I don't think anything is science [in data analysis]. Data requires a lot of judgement, and that is a good thing," Mosse adds.