Print this article | Return to Article | Return to CFO.com
The dirty little secret of customer relationship management projects: bad data.
Tim Reason, CFO Magazine
December 1, 2001
It seemed like a small, simple problem. In January, four customers of New York-based AXA Financial Inc. reported difficulty using a new self- service feature of EQAccess, the Web site that has provided individual customers with access to their life insurance and annuity accounts since 1998.
Simple maybe — incomplete address data triggered error messages. But not small. Three or four addresses could be easily fixed, explains AXA technology consultant Eric Sebo, but a quick check showed that 934 other customers would have the same problem if they tried to complete transactions.
So AXA's IT department swung into action. In two weeks, it altered the application and, working with the customer service department, tracked down and fixed the source of the bad addresses. "We didn't lose those four customers," says Sebo. More important, he says, the problem was fixed before the other customers knew about it.
This close call illustrates the dirty little secret of today's customer relationship management projects: bad data. Long the bane of customer service departments, bad data can cripple expensive CRM projects.
"Once you have the infrastructure in place and confidence in your data, you can do almost anything for customers," says Sebo. Unfortunately, all too often confidence in data is either low or dangerously misplaced. And, as AXA's experience shows, bad data can be especially devastating when it turns up on the Web.
Theoretically, the Web is the best friend a CRM manager ever had. Internally, it is the ideal tool to tie together the disparate, product-based systems that contain customer data. Externally, it can provide all customers with a similar and consistent experience, no matter what products they buy.
But as Chandos Quill, vice president of strategic marketing for Orange, California-based credit-rating and data-services provider Experian, notes, the Web generates its own data problems. One of the presumed benefits of customer self-service on the Web is that customers assume the task of keying in their own information. And while it is generally true that customers try to be accurate if they're attempting to complete a transaction, mistakes happen.
"Customers can open up all different types of accounts over our Web site," notes Angela Maynard, senior vice president and chief privacy executive of Cleveland-based KeyBank. "I can do all the training in the world for our call center to make sure customer reps get the right data, but I can't do that with customers on the Web," she says.
And data errors and deliberate omissions soar when customers who are simply seeking information are asked to register at a site — far more so, says Quill, than when similar inquiries come into a call center. Perhaps the number one sign of customers' unwillingness to part with personal data, she says, is the fact that "the most common name listed in corporate databases is Mickey Mouse."
The magnitude of the CRM data problem is starting to leak out. A Meta Group report last May noted, "CRM implementations continue to be limited by data integration and quality issues, which are more often stumbled upon, rather than anticipated." And in August, a Gartner report cited seven reasons CRM implementations fail. Topping the list: "Data is ignored." As both Meta and Gartner suggest, companies often charge ahead with expensive CRM projects without first assessing — and budgeting for — the work needed to whip existing data into shape.
A System of Systems
One reason data is such a stumbling block is that CRM isn't so much a new addition to a company's technological tool kit as an amalgamation of existing legacy systems, data warehouses, applications, call centers, and Web sites. "There's no such thing as a single CRM 'system,' " explains Bryan Foss, a customer loyalty solutions executive for IBM Global Financial Services. "The sheer number of systems [making up a CRM system] for a typical bank or insurer is 100 to 150," he says. Even if the integration of all these different data sources goes smoothly — a grand assumption — there's no guarantee they all contain good information. That can hamstring a CRM from the start.
For example, notes Experian's Quill, "One of our telecom clients found that when it combined its call center database and its marketing database, it had 300 million customers — more than the total number of people in the U.S." In addition to obvious evils, such as expensive duplicate mailings, the main problem with such data dreck is that it makes a mockery of the analytical capabilities and decision-making tools that are touted as CRM's greatest benefit. Clearly, no marketing or strategic business decisions can be based on information that is so flawed.
"We have customers who have thrown away million-dollar systems because they were too dangerous to use," agrees Lacy Edwards, CEO of San Francisco-based Evoke Software Corp., which provides data-profiling software. Consultants who do the integration work are so aware of the potential pitfalls of bad data that they are careful to insulate themselves from the consequences. "Most integrators have built into their contracts that their work is only as good as the underlying data," says Edwards.
"CFOs need to take a leading role in the decision [to begin a CRM project] and make sure the data risk is understood," notes IBM's Foss. "These decisions are very substantial; they cost millions of dollars and could be the turning point between a company's success in the market and complete failure." Too often, he says, CFOs aren't aware of the data cleanup problem until a request for additional CRM funding lands on their desk.
No Room for Error
KeyBank's Maynard is all too familiar with data inconsistencies. The 2001 Gramm-Leech-Bliley Act requires all financial institutions to give customers an opportunity to refuse permission for banks to sell or share their personal data with unaffiliated third parties. At Key, says Maynard, the decision was made to send a single mailing to all customers no matter how many accounts they held. "We didn't want customers to hear from us several times," she says, "because it would appear that we didn't understand our relationship with them."
That meant combining data from almost 30 data warehouses and individual systems. Maynard's team had to comb through a total of about 200 systems to make sure the information in those data warehouses was up-to-date and complete. Ultimately, Key mailed 11 million notices, but the effort to come up with a single list highlighted so many inconsistencies that executives decided to accelerate corporatewide data-governance and data-quality initiatives.
Nowhere is the importance of CRM data quality more critical than with the Web applications that serve customers. "Real-time delivery [of information] makes data integration and quality really key on the Web," says Experian's Quill. Because customers view and use the data in real time, there is no opportunity for customer service reps or direct-mail managers to correct or interpret common glitches. As was the case with AXA's EQAccess, many Web applications simply won't work if the underlying data isn't right. That can frustrate and discourage Web customers. Worse, it can cause call-center volumes to swell with callers looking not for products or services, but for technical support.
Making sure data is good and applications are easy to use is increasingly important. "We find that customers who go to the Web tend to be our best customers," says AXA's Susan Brown-Reitz, vice president of service delivery. "Their account balances tend to be 5 to 10 times higher than average."
Getting It Right
When done correctly, CRM can be an enormous boon to a company's finances. At Cincinnati Bell, the wholly owned subsidiary of diversified telecommunications provider Broadwing Communications, any service order entry for any product immediately populates all of the company's data warehouses. That's even true of the company's Web site. "Customers can go out to the Web and order a service, and that can get provisioned to the switch without any human intervention," says CFO Kevin Mooney.
Three years ago, notes Mooney, the goal of the Residential Service Center at Cincinnati Bell was to answer customer calls with the shortest possible hold time. Today, armed with companywide data, the task of the rechristened Residential Sales Center is to sell products from Broadwing's array of businesses, including local service, long distance, wireless, Internet access, directory advertising, ADSL, and home security services. "About 9 percent of our new wireless customers are people who called us for something else," says Mooney.
That's not just new revenue for the wireless business, he notes; it's a reduction to SG&A. Acquiring a wireless customer is a pricey proposition that averages about $350 per customer. Half of that is the subsidy for the mobile phone itself, but the other half is a commission to outside sales agents — which Broadwing doesn't have to pay if the customer calls them. "If I can halve the acquisition costs, I [will] have a customer who is much more profitable much more quickly," explains Mooney.
CRM, Mooney adds, is an evolutionary process. "Most people overscope the effort to begin with," he says. "They want to create a grand architecture and design that is going to change the world. That is very challenging, and I have never seen one succeed." In fact, when Broadwing acquired IXC Communications to create its national fiber-optic business in 1999, that company had just such a project under way. "We balled that up and scrapped it on day one," says Mooney.
Tim Reason is a contributing editor at eCFO.
Customer Dissatisfaction: You Can't Always Get What You Want
Are satisfied customers worth the trouble? "There was no correlation between how satisfied customers said they were and their economic threshold for switching providers," says Paul Bascobert, senior vice president at Chicago-based Braun Consulting, who studied the cable TV industry to try to put a dollar value on customer satisfaction. Highly satisfied customers would still switch for a 2 percent discount, while even big savings wouldn't budge some disgruntled customers.
Not only is happiness no indicator of fidelity, he notes, "there are some customers you just don't want to be happy — they are credit risks, they tie up your call centers, they abuse your services." Indeed, in a time when companies are reducing inventory, trimming expenses, and cutting head count, maybe it is time to think about getting rid of certain customers. Telecommunications companies were among the first to learn this, says Mike Schroeck of PricewaterhouseCoopers's iAnalytics practice.
"The numbers were shocking — some telcos had a 70 percent annual churn rate," says Schroeck. The companies studied customers with a propensity to switch providers, he says, "and they realized that maybe if some of them left, it wouldn't be the worst thing in the world."
Rooting out the 5 to 20 percent of customers who are "negative contributors," says Bascobert, is difficult because "the best, cleanest data is from the billing system." But individual customer revenue data ignores what it costs to acquire and care for each one. "What it comes down to is, what is the income statement of that customer?" he says. Schroeck agrees: "The goal of the integrated CRM is to understand who the most profitable customers are."
The moral? Maybe companies shouldn't spend too much time on customers who can't get satisfaction. But still they try. And they try. And they try. —T.R.