Reltio Connect

 View Only
  • 1.  Data Modeling and Governance Scenario

    Reltio Partner
    Posted 09-15-2021 18:30

    Looking for some feedback regarding best practices around data modeling and governance.

     Hypothetical situation

    • Implementing account 360 as a starting point.
    • Data will be coming in from different states
    • Matching will be configured to only match when the states are the same.
    • Each state has Data Governance team that sets
      • Matching rules for their state
      • Survivorship rules
    • Cleansing / validation may need to be handled at the state level as well
    • Roles will be created based on states so that privileges will be set at the state level (e.g. someone with access to CA should only see CA data)

     

    Considering the above, is it generally best practice to:

    1. Use entities Organization and Contact with a state field or
    2. Create state specific Organization and Contact entities that extend Organization and Contact

     

    For #1,

    • I would be concerned with matching as the number of match rules could easily exceed 100+, much more than the recommended max of 10 per entity.   Would this matter since they would all have an exact state match as a component?
    • The maintenance and security configuration may be more difficult since everything would be done on two entities, but filtered by state. Would this be the case?
    • Are there any other performance considerations?

     
    For #2,

    • This would multiply the number of entities by the number of states configured.  Would this be a concern?  I didn't see anything related max entities in a model.

     

    Any feedback would be greatly appreciated.


    Thank you,



    ------------------------------
    Ed
    ------------------------------


  • 2.  RE: Data Modeling and Governance Scenario

    Founding Member
    Posted 09-16-2021 09:47
    Is the data model identical for all states, or does each state have a unique data model?

    Why would 100 match rules be needed? As you have already concluded, this will not work in Reltio.

    My personal thumb rules on merge/match:
    • If the match is "strong" or is on a unique identifier, should automerge or better yet, use a surrogate crosswalk
    • If the match is "weak", consider the work involved to manually review and action all of the suspect matches identified
    • Less is more - keep merge/match "lean & mean"


    ------------------------------
    Walt Feldman
    Solutions Architect
    Tenerity
    ------------------------------



  • 3.  RE: Data Modeling and Governance Scenario

    Posted 09-16-2021 11:26
    Is data structure very different from state to state? Perhaps the 1st step is to standardize the data injection.

    ------------------------------
    Vivian Wu
    AbbVie
    ------------------------------



  • 4.  RE: Data Modeling and Governance Scenario

    Reltio Partner
    Posted 09-16-2021 11:47
    The data elements/model required would be consistent across the states.    I'm not 100% sure if there would need to be 100 match rules because there would be overlap, especially for those that are strict rules.    

    These would be implemented by state or groups of states, so we just want to understand the best approach so it can be easily managed without impacting performance. 

    We would also like to be sure permissions can be based on an entity attribute.  If we are able to keep one 'Organization' entity, can we use the 'State' attribute to grant permissions to state specific roles.? I've reviewed the documentation but haven't been able to find how to set permissions by entity or attributes.

    Thanks!


    ------------------------------
    Ed Mosier
    Slalom
    MO
    ------------------------------



  • 5.  RE: Data Modeling and Governance Scenario

    Reltio Employee
    Posted 09-20-2021 13:41
    Hi Ed,

    You won't need a match rule for each state.  Simply stating that the State 'Equals' will ensure that the states are the same between matched records.

    ------------------------------
    Gino Fortunato
    Senior Solution Engineer
    Reltio
    ------------------------------



  • 6.  RE: Data Modeling and Governance Scenario

    Reltio Partner
    Posted 09-27-2021 15:05
    The distinction here is that each state may have different requirements for their match rules, not just that that they can't be matched together.    I realize that we would want to re-use as much as possible but there may be the need for flexibility.

    ------------------------------
    Ed Mosier
    Slalom
    MO
    ------------------------------



  • 7.  RE: Data Modeling and Governance Scenario

    Posted 09-24-2021 11:15
    This is such a interesting use-case. 

    Wouldn't the desire to define state specific match rules/validation/cleansing and more importantly survivorship necessitate an entity per state model?

    I can't imagine it would be pretty to try and maintain up to 50 survivorship variants * n attributes. Even if you could, the reality of 50 different governance teams trying to make releases would be... stressful.

    ------------------------------
    Ted Sager
    Qlik
    ------------------------------



  • 8.  RE: Data Modeling and Governance Scenario

    Reltio Partner
    Posted 09-27-2021 15:09

    Ted, 

    That was my original thought too, but wasn't sure if that's what the consensus would be.  This is a hypothetical scenario but I feel that it illustrates some design considerations when there are large implementations with complex governance structures in place, each with different business rules.  



    ------------------------------
    Ed Mosier
    Slalom
    MO
    ------------------------------



  • 9.  RE: Data Modeling and Governance Scenario

    Reltio Employee
    Posted 09-28-2021 10:13
    Hi Ed and Ted

    We went through something similar to this a number of years ago when one of our life sciences customers was migrating away from just over 60 separate legacy MDM instances around the world to Reltio.  The original assumption was that we would preserve match rules in each of the regions because a lot of time was spent designing the rules and each region had its own "unique" needs.  After gathering all the distinct match rules (Over 130 if memory serves), it became quickly apparent to the stakeholders that A) there are only so many permutations that actually make sense before you go into crazy land when it comes to matching and B) a simpler model is way easier to maintain and can deliver good results quickly.  So the net was an initial set of common rules that worked well across regions and then if a particular region felt they had some very unique needs, the data governance team would review and decide if they wanted to add some specific rules (with filter conditions). 

    Hope this helps.  Guy​

    ------------------------------
    Guy Vorster
    ------------------------------