Data Quality: Looking Beyond Vendor Promises
by David M. Raab, Raab Associates.
With billions of data points generated daily by potential buyers, it’s easy to believe that technology can identify a prospect’s interest and intent as part of a marketing campaign. While most data vendors avoid promising 100 percent accuracy or universal coverage, marketers may assume that their data is as close to perfection as possible.
However, as anyone who’s ever been given an inexplicably roundabout route by their car’s GPS system already knows, technology is fallible. So, it’s important to test any new data sources before making a large investment in using them. Here are five factors worth considering:
Send the vendor a sample of your existing customer or prospect list and see how many names they can match. Results can be eye-opening, as vendors who claim to track every company in your market may match just half the names you give them. Of course, you’ll want to be sure the names they don’t match actually refer to real companies – it could be that your own list contains invalid information.Examine the matches they do find to be sure you’re confident that they are valid. Missing companies doesn’t mean a data source is bad, but it does mean you’ll want to supplement it with other sources to get more complete coverage.
One way that data companies increase their match rates is to ensure they have different versions of the same company’s data on file. This includes alternate contact names and multiple addresses. There is nothing wrong with this as long as the alternate versions are correctly linked to the same master entity. (If they’re not linked, then you’d have several incomplete partial profiles rather than one comprehensive shared profile).
Testing for duplicates poses challenges, since your own data may contain unrecognized duplicates. One approach is to purposely submit alternative versions of the same entity and see whether the vendor unites them. Another is to have the vendor show you all entities within a narrow group, such as a small geographic area, and have someone familiar with the companies look for duplicates.
Errors creep into data assembled from multiple sources over time. After you had the vendor show you matches against your customer list, see how well the provided information matches your own list. Significant differences should prompt follow-up by researching a few companies the vendor provided. By doing this, you’ll be more confident that the vendor is providing a prospect list that is verifiably new.
Vendors get data from different sources. This entails everything from country of origin, to industry criteria and software used to obtain data.This matters most when you’re expanding use of a data source into a market segment, since you can’t safely assume that quality in the new segment will be the same as quality in segments you’ve tested before.
There’s nothing like a test to prove whether a data source is reliable or not. But you should also find out the vendor data source and verification process. For business data, the gold standard is still telephone validation on a recurring schedule. Additional validation resources are company Web sites and social media activity.*
*A quick note on this: Vendors without reliable validation processes aren’t necessarily going to give you bad data, but you’ll certainly want to look at their products more closely.