Data quality issues plague CRM

CRM’s potential to provide better customer service and in turn boost revenue has caught the attention of many companies. But a number of CRM deployments are thwarted by faulty, inconsistent data sets that prevent enterprises from having a clear, unified profile of each customer.

“It’s massive,” says Derek Strauss, CEO of Bethesda, Md.-based BI (business intelligence) applications hoster Assurenet Ltd., of the problems of data quality in CRM applications. “You have to have accurate information, and most of the front-end systems which deal with customers do not have accurate information about the customers. There’s disjointed [data]; there’s a lot of blanks in some of the critical fields.”

Even seemingly minute mistakes such as being one digit off on a customer’s street address can plague data sets, says Joe Butt, senior analyst at Forrester Research Inc. in Cambridge, Mass. This type of problem can lead, for example, to a company sending out multiple mailings to the same customer, because the address is listed in two different ways.

Analyst firm Gartner Inc. believes that the problem is widespread. Gartner in July wrote that more than 75 per cent of enterprises engaged in CRM initiatives cannot combine a comprehensive view of a customer with actionable, personalized advice to customer service and sales agents.

Data-quality issues are particularly vital in health care, according to Jim Bodenbender, president of Chicago-based Madison Information technologies, which provides data cleaning services for health-care facilities.

The average hospital has 25 different record systems, Bodenbender says. “A lot of those are very departmental in nature, and what happens is the data for a different patient may get propagated within those different systems,” with an error rate of 10 per cent, Bodenbender says.

“What it means is they have duplicate medical records for the same person,” leaving caregivers without a single view of a patient, Bodenbender says.

Data-cleanup services from Madison cost an average of US$500,000, with some costs rising above US$1 million.

Integrating customer data sets is challenging, says IBM Corp. official Bryan Foss, a Customer Loyalty Solutions executive in London who has been working on CRM deployments both internally and at customer sites.

“CRM has a broad scope,” Foss says. Existing systems, such as manufacturing and ordering, must be integrated to enable a single view of a customer base, he says.

“In a typical bank or insurance company, for example, there could be 50 to 150 different systems which contain customer data, and if you want to gain a single view of that customer and the value of that customer and their needs, then you need to combine that data,” Foss says. This requires pulling together different data stores with data of varying ages on different databases, Foss says. These data stores use multiple programming languages and data formats, he says.

“Some of the data is [of] poor quality. Maybe it hasn’t been used for a long time,” Foss says. “To bring [together] all this data is a big task,” Foss adds. “It’s an enormous task. It can’t be done overnight.”

Foss is using Evoke Software’s tool of the same name to automate the process of data analysis. The product finds different relationships in customer data from multiple sources and helps with restructuring and cleaning that data, he says.

IBM is using the Evoke product to build data warehouses and to profile data for a Siebel CRM application.

Evoke offers an alternative to hiring systems integrators to manually examine data files, says Rick Cortese, president and CEO of the San Francisco-based company. “[Manual intervention] is tedious, time-consuming, and expensive and by definition not very accurate,” Cortese says.

The cost of Evoke Software runs from approximately US$500,000 to several million dollars, depending on configuration and services.

Assurenet has deployed another product called MetaRecon, from Research Triangle Park, N.C.-based Metagenix. The product analyses data in source systems, including looking at data values. “It tells you early in your project where your problems lie,” Strauss says.

MetaRecon provides data profiling and analysis, and also can perform code generation, including producing online analytical processing specifications, according to Metagenix President Greg Leman. MetaRecon finds inconsistencies and anomalies in data and determines relationships. Additionally, the product can reverse-engineer business rules, enabling users to compare these rules to what is supported by the data. MetaRecon costs US$25,000 for an entry-level version, and US$250,000 to US$300,000 for a perpetual, enterprise licence.

Lanham, Md.-based Group 1 Software offers both packaged software, called Enterprise Data Quality, and an ASP-based offering, called HotData, to help clean up records. Using Group 1’s technology, a company could, for example, take data feeds from SAP, PeopleSoft, and Siebel applications and combine them all in a single database. Group 1’s software can then analyse and fix anomalies in the data, such as fixing names that are spelled with a slight variation in two sets of records.

Group 1 charges 4 cents to 16 cents per transaction or a perpetual licence fee ranging from US$200,000 to US$1 million.

Web-based data entry complicates data-quality problems, says Tim Wagner, CTO of Group 1. “A competent data-entry clerk has an average error rate of 2 [per cent] to 4 per cent. What we’re seeing on the Web is a 10 [per cent] to 15 per cent error rate because John Q. Public is actually entering the data.”