Last week, I wrote about the pros and cons of how to onboard credit ratings, part of which often requires some extent of entity matching, especially when leveraging direct agency rating feeds.
With that, I was asked to dig a little deeper into the ins and outs of entity matching from my readers.
If you have different sources of company-related data like entity ratings, the challenge is to identify those entities correctly by matching them. If you take the ratings directly from their source, odds are that no one has done the work for you to align them with other entity ratings sources for a 360-degree view.
So, somebody has to do that work, by employing a variety of entity matching mechanisms. Here’s a breakdown of the ones we recommend our clients use to gain the clearest possible view of a potential investment:
- Common Identifiers: These are the logical first step in cross-matching. Two examples are LEI and FIGI. In addition, Factset ID and Refinitiv PERM ID are often used, though they are proprietary sources. A good data management platform can take advantage of identifiers which are not used across the board but only included with a subset of sources. This still allows to build up a consistent entity master, piece by piece.
- Third-Party Cross-Referencing: These data products link entity schemas to each other and can be used when the non-proprietary identifiers are only available to a limited extent. There are several specialized products in this category, one of which is S&P’s BECRS.
- Instrument-Based Issuer Matching: This is a powerful mechanism because at the instrument level, you usually have a lot of common identifiers. As long as you know a company’s issued instruments, you can match those instruments and then conclude their issuers are one and the same.
- Fuzzy Matching: This uses company names, addresses and other attributes for company identification. In our experience, particularly for complex legal hierarchies, this is not the best means of creating reliable matches, but it does help in creating suspect matches rated by percentage. As a result, you need to feed the results into some kind of manual verification process.
If you choose to do this in-house, you can arrive at some initial considerations using a combination of the four mechanisms above based on what is most effective for your needs. For example, do you need a 100% accurate entity master or are a few spurious matches still acceptable for your purposes?
As long as one can live with a few anomalies, a fully automated matching may be able to do the job. If, however, your use case is very stringent, some level of user intervention to approve or reject the remaining suspect matches is likely required.