In profiling our data, we have learned that our best suspect match is the phone number, but we filter out all of the "garbage" phone numbers (i.e. 0000000000, 1111111111, etc.). We never actually use the suspect match to perform an actual "merge". Instead, we use suspects to help our customers identify cases where they have created multiple accounts over the course of several years. We use the suspect matches to help customers "get to one" account.
------------------------------
Walt Feldman
Digital Data Lead
Subway
------------------------------
Original Message:
Sent: 05-03-2021 10:02
From: Angela Wawrzaszek
Subject: Match/Merge rule best practices
Our current suspect rules involve Name and Address Only. 1. Removing noise words 2. Using Phoenetics. These 2 rules do introduce alot of not a match. Can you share M&M you have used that are lean and mean?
------------------------------
Angela Wawrzaszek
Manager, MDM governance and Application Enablement
Original Message:
Sent: 05-03-2021 09:56
From: Mike Frasca
Subject: Match/Merge rule best practices
I would echo #3 from @Walt Feldman above. I often see customers start with many many suspect rules, but only have staffing for 1 person to review the potential matches. This just leads to thousands of unresolved potential matches and extra overhead on the platform. For any manual process, there is really no benefit to creating more in the queue than can be reasonably resolved by the stewardship team.
------------------------------
Mike Frasca
Original Message:
Sent: 05-02-2021 09:17
From: Walt Feldman
Subject: Match/Merge rule best practices
In my 4+ years of of working with Reltio, these are my 3 takeaways on match/merge:
- You really need to know your data extremely well before defining match/merge. If you don't, be prepared to refactor your rules.
- If you have one or more attributes that are so unique that you would use them for auto merge, consider the use of surrogate crosswalks. This has made Reltio much more manageable and performant for us.
- Keep your suspect match rules "lean and mean". Consider what you will actually do if/when a suspect match is identified. Define your suspect match rule in a way that only the best possible suspect matches are identified, without making the rule overly complex.
------------------------------
Walt Feldman
Digital Data Lead
Subway
Original Message:
Sent: 04-30-2021 11:26
From: Angela Wawrzaszek
Subject: Match/Merge rule best practices
Thank you we have utilized the documentation. I am looking for examples of successful M&M models by clients in use. There are multiple ways to create M&M, hopeful to learn from others what has worked from them vs. 'trying' different solutions.
------------------------------
Angela Wawrzaszek
Manager, MDM governance and Application Enablement
Original Message:
Sent: 04-29-2021 14:24
From: Venki Subramanian
Subject: Match/Merge rule best practices
Hi Angela,
We have extensive documentation on Match, Merge and Survivorship in our documentation portal.
You will find that here - https://docs.reltio.com/matchmerge/matchmergeoverview.html
There are a few illustrative examples of match rules that you will find here.
https://docs.reltio.com/matchmerge/exampleoverview.html
We have introduced a Match Rule Analyzer that helps you analyze the match rules and get recommendations for improvements of optimizations.
https://docs.reltio.com/tenantmanagement/matchanalysis.html
Hope these resources are helpful. Let us know if you are looking for something more specific.
Thanks,
Venki
------------------------------
Venki Subramanian
Original Message:
Sent: 04-29-2021 08:55
From: Angela Wawrzaszek
Subject: Match/Merge rule best practices
Hi everyone - Can you share Match/Merge best practices? What features have you enabled to improve the quality of the suggested matches?
------------------------------
Angela Wawrzaszek
Manager, MDM governance and Application Enablement
------------------------------