Data stewards from the business, as well as gatekeepers from IT, compose a CRM team charged with driving new data management solutions. But the business users are always in front.

Describe, Define, Govern

Similarly, when Scott Sullivan joined Pitt Ohio Express, a $238 million mid-market transportation company, as its VP of IT and services, one of the first things he did was sit down with his business users and help them define what exactly the term customer meant to them. Sullivan helped the business narrow its list of customers from 450,000 to 10,000 active consumers of its services. Sullivan also pulled the plug on an ERP system roll-out because he thought it wasn't going to satisfy the company's needs and was going to take longer than had been originally projected. (The project was green-lighted before Sullivan joined Pitt Ohio in 2001.) Since then, Sullivan has integrated an assortment of existing applications to form a customer management system for the sales and marketing group and the operations department. (For more on the integration challenges confronting mid-market CIOs, see "Stuck in the Middle", page 48.)

Sullivan also spent time ensuring that Pitt Ohio Express's customer data was clean. Dirty data is hardly a new problem, but the fact that CIOs are still complaining about it, analysts are still noting its prevalence, and vendors are still selling solutions to address it indicates that it hasn't gone away. Dirty data problems are amplified by the number of systems and users that touch customer data, especially if there are no established governance processes or technology safeguards. For example, Sullivan points to the disconnect in address requirements between the sales and marketing department and the operations division. The sales and marketing group needs exact addresses, whereas drivers can get by with more inexact data. "If the address is 'the back gate at the Kmart plaza'," he says, "that's OK for the driver, but not so great for sales and marketing." And if no one takes ownership of making sure the data is consistent, "there can be up to 10 to 15 different versions of your customers [within your company]", says Tom Reilly, IBM's VP of master data solutions.

Once your management team has formulated a data management strategy - say it wants to improve the ways in which the company targets and contacts prospects - it's time to consider the technology options available to integrate all the customer data so that sales and marketing will be going after the most appropriate customers. You can go the enterprise vendor route, or have your CRM systems hosted by an on-demand vendor. Or you can integrate existing customer-data systems by building a service-oriented architecture or structure using the Web to knit together all the customer information contained within a company's business applications. UnumProvident's Dolmovich decided to go the Web services route. He chose IBM's WebSphere Customer Centre product to pull together the pockets of customer data on account activity, payments and premiums.

Dolmovich says the first data loaded into the CDI hub in late 2005 came from business customers (companies or employers that bought or sponsored UnumProvident's disability products) and brokers (the independent businesspeople who sell them). With the new system, Dolmovich says, "We are now able to assimilate and display a broker's entire block of business and create some statistics and a profile of our relationship with that broker". UnumProvident is now working to create individual profiles of employer customers so that every time a new customer account is created or accessed - perhaps to change an address or add new customer information - all employees of the insurance company, regardless of what system they are using, will see that change at the same time.

The New, New Hype

Whenever a new CRM solution emerges, it's inevitably followed by hype, complexity and confusion. CDI is no different, says Colin White, founder of consultancy BI Research. One challenge for companies embarking on a master data management strategy is getting all parties to agree on common definitions and labels for categorizing customer data. Starwood Hotels & Resorts Worldwide is currently in the process of sunsetting its legacy mainframe system in order to move to a SOA environment. The aim, says Song Park, Starwood's director of pricing and availability technologies, is to allow for more real-time and online reservation capabilities and transactions for its 900 hotels in 80 countries. But a major pain point for the groups working on the SOA migration has been hammering out the data labels and definitions for the Web services that will be consistent across the SOA implementation. How, for example, one group defines a specific hotel's property identification label can vary from PID, to pID, to property ID, to name just a few of the possibilities, Park says. "How do you synchronize [those labels]? Who owns that data? Who's mapping those things?" Park asks.

Park says he's pushing for a data dictionary of pre-established services so that the developers working on the project can employ a common set of labels. "And the developers need to talk to each other," Park adds.