Reltio Master Data Management Match Rules

View Only

Reltio Master Data Management Match Rules - FAQ's Part 2

By Joel Snipes posted 05-30-2021 11:55

Recommend

In the first part of our blog on Match Rules, we covered an overview of Match Rules and how they work. Now let us dive into the options available to you in implementing a match rule.

Match by Operational Value (OV) only

In Reltio we consolidate multiple records that will call crosswalks into a golden record that we call the operational value (OV). The process of consolidating the records is called survivorship. When developing a match rule you have to decide whether you want to consider all values that have contributed to that record (all of the crosswalks) or just the OV. Because the quality of your crosswalks can vary, in most cases I see match rules using OV only left on, but there are use cases where you might consider otherwise.

And/Or Statements

At the top level there are and/or statements, just like in any programming language. Each row in the match builder contains an “And/Or” and a Comparator (Exact, Fuzzy, Equals, etc) and an attribute. If you are looking to exclude matches from your result set “ And not” and “Or not” are also available, but not necessarily advised. Negatives in rules impact performance and should be avoided if possible.

Exact

This is the plain jane of match comparators. It compares two fields to see if they contain the same value.

Exact or Null

This works very similarly to exact Consider if we wanted to match a name with a suffix – so senior (John Paul, Sr) to a non-suffix name (John Paul) – if we used exact, they would not match because in Reltio null != Sr.

Exact or Null allows you to match a field, if they contain the same value, or one contains a null.

Exact or All Null

If you have a match rule on Exact(FirstName, LastName and Suffix) you will have cast a very small net. This is because Exact will not match null to null. If you have a field like Suffix, that is rarely populated, but could improve your match quality when it is, then Exact or Null is a fit.

Equals

Equals works as a filter, coming from the relational world it would compare to a where clause. This is useful for reducing the set of data you are looking to consider for matching. For example, you may be matching individuals, but would like to exclude your employees. In your match rule you could have Equals EmployeeIndicator false.

Fuzzy

Fuzzy matching allows you to match attributes that don’t have the exact same value, but represent the same thing. For example: Robert Jordan and Bob Jordon are likely the same person if they share a birthday and address. The only difference is a nickname. Fuzzy matching helps us find these using comparators like Soundex, and Name dictionaries.

RuleScore: Stand Alone and Incremental

Not all match rules are created equal, and if an Entity in your tenant has multiple potential matches, you need a way to rank them. That is where rule scores come into play.

The stand alone rule score is a number you assign based on your feeling for the accuracy of the match rule. If you feel the rule is strong you might set it to 80, if it is poor you might set it to 20. With this you can have your better rules be considered by your data stewards first.

You have two match rules, Rule 1 and Rule2, which both trigger a potential match. Rule 1 has a standalone score of 50, Rule 2 has a standalone score of 55. Which gets listed higher? Trick question! You don’t have enough information. Incremental rule score works with your other rules to create an aggregate score. If two rules get triggered together, their total rule score is their standalone plus the incremental score of the other rules that were triggered.

Q: How do I see the Match Rule for an existing profile?

A: For a suspect rule, go to profile, then the Same As/Matching view on the left (the icon of two people standing next to each other) and expand the carrot in the right column. For an auto match it will appear in the history or activity log. brought in.

Comparators and Match Token tools

Reltio has a variety of tools available regarding characters and words. These allow for the more advanced comparison of attributes than Exact, Equals, etc..

There are also a Noise Word Dictionaries that can be applied that wipes out things like Inc., Corp., LLC. These words would prevent an exact match, and could even stop a fuzzy match, so before comparing the strings, this dictionary will remove those noise words.

There is a wide variety of tools available here too:

Basic String Comparator – the simplest form of matching, and what Reltio uses for exact, exact or null, and fuzzy all use this comparator class. Foreign Unicode characters are accommodated with this comparator.
Damerau-Levenshtein Distance –calculates how many characters are out of place, deleted, or transposed.
Metaphone/Double Metaphone/Soundex Comparator – allows for phonetically pronunciation comparison rather than spelling, i.e, bookkeeper vs. bookkeeper, Sean vs. Shawn.
Organization Name Comparator – includes a Noise Word dictionary that excludes words (Inc., LLC) that would stop an exact match, and could even stop a fuzzy match. You can also take the original default dictionary and add in additional words, then just upload it to an S3 bucket or a Google bucket and point at that.
Address Line Comparator –used with exact or fuzzy, though it has limited use in fuzzy. Generally, locate will cleanse the same address to the same exact value, resulting in a lot of tight matches. A Noise Dictionary is also available to cleanse words like road, avenue, etc.
Range Comparator – numerically similar US zip codes tend to be geographically adjacent to each other. It’s useful for identifying family relationships as families tend to live close to each other or a store location close to customers.
Proximate Geo Comparator – takes two cleanseed addresses, and uses the address’ latitude and longitude down to a few decimal places to determine their linear distance. You can set a threshold distance, say of 100 feet. If the two coordinates are within that threshold it is a match.
Custom Comparator/Build Your Own – offers the ability to customize your own comparator by combining existing comparators.

The existing token classes correspond virtually one-to-one with the comparator classes. You must apply them both to your Match Rule, which can be done in the Match Rule builder via the Row settings on the right.

Learn More with the Reltio Community

The Reltio Community is a great place to learn more about how to use the Reltio products and connect with Master Data Management peers. Rely on the expertise of Reltio partners, customers, and technical experts.

Reltio Connect