I got a couple of e-mails from a friend over the past few days, and he asked some great questions about matching (or entity resolution, as Infoglide Software prefers to call it).
His first question was about AbiliTec IDs for individuals. He wondered if this was an Axciom product or an industry standard identifier. He was looking for a way to uniquely identify customers from the web, retail point-of-sale (POS), and other marketing channels for a client who doesn’t have any useful way of identifying the same customer across channels.
I’m familiar with AbiliTec IDs (Acxiom calls them AbiliTec links). It’s a persistent identification scheme for consumers and addresses, issued and controlled by Acxiom, similar to how D&B issues and controls the D-U-N-S Number for businesses. There’s a good brochure available on Acxiom’s UK website.
To answer his specific questions, it’s definitely an Acxiom product, not an industry or open standard identifier. So one can only get AbiliTec links by working with Acxiom in some way, shape or form.
As far as uniquely identifying individuals across channels like the web, retail POS, and other marketing outlets, that’s pretty much exactly what AbiliTec is used for – think of it as a high speed matching engine that will return AbiliTec links for every consumer record you feed it. The downside is you’ve got to pay Acxiom for the privilege. But like D&B, they’re pretty good at helping you create the business case to justify the expense.
Then my friend came back with another question: are there any competing products for the AbiliTec link?
My answer was: none that I know of that are as widely accepted. The three consumer credit bureaus (Equifax, Experian and TransUnion) all sell their own persistent consumer IDs too, but in my opinion, Acxiom’s is very good at that particular task.
Products like DataFlux, Trillium, Informatica, IBM’s QualityStage and SAP BusinessObjects can also generate a key with the same value given varying input, so that different representations of the same person, such as
- John O’Connor, 30 Palomino Lane, Westwood MA 02090
- J. J. Oconnor, 303 Palomino Lane, Westwood MI 02093
- John Joseph Connor, 30 Polo Pony Court, Westawooda RI 02021
will all generate the same key value. So you can use that generated key to identify the person across three different channels.
Of course, it all depends how well you write and tune your business rules, too. If you write them too tight, then the three records above won’t generate the same key after all. Write them too loose, and a bunch of other records will also generate that same key value and you’ll end up with a bunch of false positives (non-Johns).
The beauty of Acxiom’s approach (and D&B’s for that matter when you’re interested in businesses) is that they’ve both got anywhere from hundreds of millions to billions of records of reference data to work with – both to refine their match engine and business rules, and to match against.
So when you’re comparing records, you’re not just comparing these three representations against one another, you’re comparing them to all of the historical addresses this person has lived at over the last 25 years, and all of the other versions of their name of file from marriages, divorces and name changes. Don’t underestimate the power of the database!
You can retweet this by cutting and pasting < New on the Hub Designs Blog: “Matching (That Is, Entity Resolution) Revisited” at http://wp.me/p5Tdn-yS >
Image courtesy of Infoglide Corporation