Straight Talk on Matching

Download the Whitepaper

Should match perfection be your data destination?

The pursuit of perfection often impedes improvement.

— George Will

Business identity resolution in data management is the process of matching a collection of data points (a ‘record’) against a trusted set of external data. The goal is to provide a single view of the business that record is referring to.

Identity resolution allows for stronger, more personalized, business interactions, especially when it comes to one-to-one customer communication, When we consider the volume of data flowing into an organization, compounded by ever-increasing digital interactions, maintaining a clear view of our business partners – whether customer, prospect, supplier, or partner – across our organization is a business imperative.

Sales and marketing operations are one of many use cases where matching plays a key role. Matching incoming data to a reference database supports data consistency and avoids pitfalls like duplicate records that often result from subjective human input. The downstream impacts of missing or improper resolution can be significant. As such, users of identity resolution services naturally look first to maximize their match rates.

Match perfection may be counterproductive

All too often, business leaders are under the false assumption that matching is an all-or-nothing proposition. Should 100% match on every input record be our goal? No. It’s not reasonable nor as valuable as one might think.

Better match rates aren’t the same as better match quality, and, depending on our specific use case, striving for 100% match rates can be counterproductive. Rather than aiming for perfection, organizations should focus on improving match quality and understanding the drivers that keep them from that 100% mark.

Companies invest considerable time and resources in the pursuit of flawless data, but the fact is that large quantities of data will rarely achieve perfection when it comes to identity resolution. There are three primary reasons for this:

  • The quality of the data, whether input data or reference data, just isn’t where it needs to be for a perfect 100% scenario
  • Process factors including constraints of the matching algorithm and how it’s configured
  • There’s just not a ‘real’ entity out there that we can match to 10 best practices for getting the most out of identity resolution

In our latest whitepaper, Straight Talk on Matching; Why 100% Resolution is Unrealistic – and May Be Counterproductive, we describe in detail the challenges associated with matching, and the difficulty – and often the impossibility – of reaching 100% match.

Here is a preview of 10 recommendations that may provide a more productive lens for measuring success. Here we recast the challenge in terms of effort versus value.

  1. Don’t expect 100% match. Unless your datasets are small enough to curate manually, some records will inevitably fail to match, and some will return imperfect matches. The whitepaper goes into greater detail to help you set expectations within your organization.
  2. Focus on value. Direct your optimization efforts into identify resolution requests that offer the greatest business impact. Provide as much data as possible in these requests.
  3. Know your limits. Some match processes have broad capabilities, and others are intended to have a narrow focus. Understand your reference universe before you attempt to match data against it.
  4. Understand match process. Understand the specific process’s minimum match criteria, prepare request data to meet them, and consider delaying requests for records that do not fulfil them until more data points have been gathered.
  5. Configure the match system. Take advantage of any tuning configurations that allow you to align matching behavior with specific expectations and needs.
  6. Fix the problems you know. Discuss and resolve systemic data quality issues. For example, if a source regularly supplies bad city names, work with them to fix the problem before allowing the data to move forward.
  7. Learn the lingo. Understand the metadata returned with match output and what it implies.
  8. Have a “Plan B”. Establish a secondary protocol for addressing identity requests that do not meet match acceptance criteria.
  9. Watch for trends. Track success rates over time to quantify improvements and to monitor for performance issues.
  10. Ask for help. Dun & Bradstreet has a team of identity resolution experts who specialize in optimizing match performance for specific use cases and business goals.

Straight Talk on Matching; Why 100% Resolution is Unrealistic – and May Be Counterproductive explores the typical scenarios that lead to failed identify match requests, and some of the ways to address these situations. It’s critical reading before you embark on your next identity resolution exercise.

Download the Whitepaper


George C. L’Heureux, Jr, and Cecilia Petit, Principal Consultants for Data Strategy at Dun & Bradstreet, co-authored the whitepaper and contributed to this article.