Hitmetrix - User behavior analytics & recording

DMers Must Decide What Is Correct When Cleaning Data

SAN FRANCISCO — Data quality vendors produce different results when cleaning business-to-business data, so the marketer ultimately must decide what information is correct.

That was the focus of a presentation yesterday by Ruth Stevens, president of eMarketing Strategy, and Bernice Grossman, president of DMRS Group Inc., at the Summer National Center for Database Marketing Conference here.

In May, the panelists performed a test with data quality vendors to determine how well the companies could clean “dirty data.” Participating vendors were Acxiom, DataFlux, Donnelley Marketing and Harte-Hanks.

Grossman compiled a live sample file of 10,000 names with 12 fields such as last name, first name, phone number, e-mail address, business title and company address, including city, state and five-digit ZIP code. Vendors were to perform as many data hygiene tasks as their software process allowed, show counts for each field corrected and answer key software questions. The questions included:

* What is your company’s definition of data hygiene?

* What are the five most important ways your work differs from that of your competitors?

* How do you characterize bad BTB data?

The vendors were asked to do the job in 30 days. They did not charge for their services.

Data supplied by the U.S. Postal Service, such as address data, was standardized across all vendors. But for other types of data, such as phone numbers and business titles, there were discrepancies. Some vendors had phone numbers and titles that differed from the live data supplied by Grossman while others didn’t.

Which ones were correct?

“The takeaway here is that services provided by the postal service, such as NCOA updating, are highly reliable,” Stevens said. “However, if your interpretation of clean data involves updating specific phone numbers, then that’s another issue.”

These corrections, however, can be made for an extra cost.

Grossman said it is important to be specific about what you want done to your data.

“It really becomes incumbent upon you to learn how to ask for exactly what you want, as opposed to hoping you are going to get it,” Grossman said.

Related Posts