When Good CRM Data Goes Bad

Updated: December 27, 2007

Issue

There are many ways good CRM data can go bad. The most common good-data-gone-bad scenarios tend to revolve around incorrect data, missing data and data that's unclearly defined, as well as the lack of a data-quality initiative. Luckily, there are ways to avoid these pitfalls.

Analysis

Good data gone awry is often a closely guarded secret. And more likely than not, these kinds of problems are granular in nature. "If you get one customer's name wrong, it only affects one customer at a time," explained Gareth Herschel, a research director at Gartner, an IT research and advisory company. "It's insidious, though, because most companies don't know how bad their data situation is, and it's often very difficult to discover something is wrong. If the data is too old, any analysis may be incorrect — but they won't know that it's because the data is old and are likely to fault the analysis as being flawed."

Incorrect data involves misspelled names, mistaken addresses, wrong dates of birth and more. There are also many cases in which data becomes incorrect because it's outdated. "This information is basically too old and no longer a valid representation of the customer, although it may have been true," Herschel said. "For example, your customer used to call a call center for help but now uses the Web. It's information that was accurate at a particular moment in time, but it's no longer an accurate reflection of the relationship you have with that customer today."

Missing data is another potential negative influence on a customer relationship. "Say you've built up an accurate reflection of the customer's requirements using your call center. But that information isn't integrated into the Web channel. The fact that the customer is going to the Web for support is missing from your profile on them — it's something about them you don't understand," said Herschel. Another example: You might have data about a customer's purchases, but you don't have that information tied into what they're returning. It looks like they're buying a lot from your company, when they're actually returning 90 percent of their purchases.

Data often comes in the wrong format or is unclearly defined. You might have some information about customer profitability, but others looking at it might not understand the definition or what calculations of profitability entail. If you don't understand how the information is defined, using it is a risk.

In addition, businesses can't ignore data quality. All organizations should think about a data-quality initiative.

Poor data quality often leads to botched data execution. "Many years ago, for example, there was a story about a bank in the U.K. that was doing some segmentation of their highest-value customers to test a marketing campaign to ensure they were using the software correctly and that the processes were properly aligned. They unwisely named their test group ‘Rich Bastard,'" said Herschel. "And then they accidentally executed the campaign and sent out a mailing to their best customers with the salutation: ‘Dear Rich Bastard.' Obviously, it was a colossal mistake. In this case, it was a data execution gone horribly wrong. If you've got the right processes in place, you should be able to catch issues like this before they turn into costly, embarrassing mistakes."

Recommendations

Herschel offered some quick tips to help businesses steer clear of bad-data embarrassment:

1. Examine the way your company is using data to determine what level of accuracy is required. Realistically, most organizations are never going to have perfect data. Adopt a strategy to ensure that the data is of sufficient quality. Quality isn't absolute — it's in the eye of the beholder. Figure out how good your data needs to be, then find someone to validate it.

2. If the data is old, it's important to know how old it is and how quickly the relationship changed. Look at different segments and the cycles of the relationship within each of the segments. For example, if you're running an airline, you would break out a leisure traveler who flies once per year and a consultant who flies twice per week to different locations and into separate categories to monitor the relationship. Looking at data going back two to four years for the leisure traveler and data from just the past year for the consultant will provide a good reflection of their business and the relationship.

3. If information is missing, take a good look at four different types of data — descriptive, transactional, social network and attitudinal — and ask whether your relationship with your customers is likely to benefit significantly if you start capturing more information. If you have descriptive and transactional data, chances are you have plenty of information to do a good job of relating to the customer. But you might get extra value by trying to tap into the customer's attitude by text or speech mining, or even by tapping into the customer's social network so that your company can do a better job of marketing to them as a group. On the formatting side, this means ensuring good companywide training to make sure that everyone understands how this data can be used. Describe what the data includes and excludes to avoid making false assumptions about the quality of data you're dealing with.

For more information on data collection, check out this brief by Focus Expert Pam Baker.