The problems data integrators face are legion, but disparate data sources are especially odious to them. Organizations often support between five and eight different database technologies, and 50 different sources of data from the operational side, according to Kevin Strange, vice president and research director at Gartner.

One reason data tends to be Balkanized is for proprietary reasons, as each department jealously defends its own little fiefdom. But this can result in very real costs to the company, such as a California insurer that offered both vehicle and home insurance.

According to Strange, the company had taken out a series of space ads, asking customers that held homeowners policies to call and get a special rate on auto insurance. Of course, if that data had been available in a centralized location, the insurer could have made a series of outbound calls rather than relying on mass marketing.

“Data belongs to the enterprise, not to a specific project or department,” Strange said, adding that it is critical for companies to implement this thinking through the entire enterprise.

Other times data can be flat-out incorrect. Strange gave the example of one firm that discovered during a data audit that social security numbers, which it used to track customers, were often incorrect. It turned out that it was paying a foreign firm to enter its data, and the bonus structure was based on volume of records processed. As a result, keypunch operators were throwing in any numbers that happened to come into their heads, if not using the same number over and over.

Assuming that data can be cleaned and verified, training users across an enterprise presents its own set of difficulties. A common trap firms fall into is bringing in an outside consultant who cheerfully teaches hundreds of employees how to use a piece of reporting software. But the employees are never taught how to analyze raw data. This leaves them able to use the information the tool serves up, but not how to take the data in new, and potentially more profitable, directions through analysis.

This type of training isn’t cheap, and the hidden costs can come as a surprise to organizations that want a plug-and-play system. Strange offered a graphic view of potential surprises in CRM expenditures, likening professional service and software licensing fees to the tip of an iceberg bobbing above the water level.

To further the metaphor a ship, labeled the “S.S. CRM Project,” blithely steamed toward the beneath-the-waves reality of data acquisition costs, customer data integration headaches and other data quality issues, which can make up 80% of an implementation’s cost, and in Strange’s graphic a similar portion of the iceberg. And like the business portion of the iceberg waiting, beneath sea level, to interact with the ship’s hull, these costs are not often clearly visible.

Sometimes determining which data is actionable can only happen after a system has been in the field for a while. Strange told the story of a health insurance call center that was taking an unused field on the customer record and, for certain callers, entering the code for hemorrhoids.

It turned out that these were particularly difficult customers, and service reps were trying to cue each other in when a particular client was a “pain in the butt.” It wasn’t a formal code, Strange said, but internally it was a useful one.

Strange made his comments during a session at the Gartner CRM Summit in Baltimore.

The problems data integrators face are legion, but disparate data sources are especially odious to them. Organizations often support between five and eight different database technologies, and 50 different sources of data from the operational side, according to Kevin Strange, vice president and research director at Gartner.

One reason data tends to be Balkanized is for proprietary reasons, as each department jealously defends its own little fiefdom. But this can result in very real costs to the company, such as a California insurer that offered both vehicle and home insurance.

According to Strange, the company had taken out a series of space ads, asking customers that held homeowners policies to call and get a special rate on auto insurance. Of course, if that data had been available in a centralized location, the insurer could have made a series of outbound calls rather than relying on mass marketing.

"Data belongs to the enterprise, not to a specific project or department," Strange said, adding that it is critical for companies to implement this thinking through the entire enterprise.

Other times data can be flat-out incorrect. Strange gave the example of one firm that discovered during a data audit that social security numbers, which it used to track customers, were often incorrect. It turned out that it was paying a foreign firm to enter its data, and the bonus structure was based on volume of records processed. As a result, keypunch operators were throwing in any numbers that happened to come into their heads, if not using the same number over and over.

Assuming that data can be cleaned and verified, training users across an enterprise presents its own set of difficulties. A common trap firms fall into is bringing in an outside consultant who cheerfully teaches hundreds of employees how to use a piece of reporting software. But the employees are never taught how to analyze raw data. This leaves them able to use the information the tool serves up, but not how to take the data in new, and potentially more profitable, directions through analysis.

This type of training isn't cheap, and the hidden costs can come as a surprise to organizations that want a plug-and-play system. Strange offered a graphic view of potential surprises in CRM expenditures, likening professional service and software licensing fees to the tip of an iceberg bobbing above the water level.

To further the metaphor a ship, labeled the "S.S. CRM Project," blithely steamed toward the beneath-the-waves reality of data acquisition costs, customer data integration headaches and other data quality issues, which can make up 80% of an implementation's cost, and in Strange's graphic a similar portion of the iceberg. And like the business portion of the iceberg waiting, beneath sea level, to interact with the ship's hull, these costs are not often clearly visible.

Sometimes determining which data is actionable can only happen after a system has been in the field for a while. Strange told the story of a health insurance call center that was taking an unused field on the customer record and, for certain callers, entering the code for hemorrhoids.

It turned out that these were particularly difficult customers, and service reps were trying to cue each other in when a particular client was a "pain in the butt." It wasn't a formal code, Strange said, but internally it was a useful one.

Strange made his comments during a session at the Gartner CRM Summit in Baltimore.