≡ Menu

Live from DAMA – Case Study: Implementing a Securities Master Using Flexible Data Models at Lord Abbett

Share

Almost done now and next up is Len from Universal Data Models talking about a Case Study: Implementing a Securities Master Using Flexible Data Models at Lord Abbett. Lord Abbett is a securities company and a very old, established one at that, who regard securities trading as a craft deserving of excellence. Manage $112B in assets and a full range of funds, plans and accounts. Started with 3 asset classes – equities, bonds for instance. Securities master data is required by multiple processes – it’s the core data across the front, middle and back offices. Accurate reporting and understanding require good master data management for securities but different groups have different data they keep for securities.

Could have started with a clean state and started gathering information requirements, building a model and evolving it. It is, however, a problem well known to be difficult. There is no one right answer, lots of disagreement etc. Lord Abbett decided to combine build and buy using knowledge from outside, standardized and mature models/patterns and inside expertise. Level of abstraction is considered one of the greatest sources of variation in data models and this was the main issue at Lord Abbett.

They used some Universal Data Models – templates for common data scenarios such as common and industry data models. Classification, features and roles are the three UDM patterns being discussed.

  1. Security Classification is a big issue as different groups within and without Lord Abbett classify securities differently and the classification is constantly changing. Can use a large number of type classes to handle each kind of classification but can result in many classes and constant data model change. Instead use a meta model including classes like Entity, Entity Category Classification, Entity Category, Entity Category Type. This allows anything to be classified in any number of ways as each new kind of classification is an instance of these classes. This allowed constant evolution without having to change the data model.
  2. Features created a similar problem with over 500 distinct kinds of features identified. New asset classes tend to create new features. Also there are business rules for different features such as interest rate not applicable to a certain kind of asset. Built a meta structure to allow Securities to have Security Features each of which had potentially many Security Feature Parameters.
  3. Roles was the third pattern. A person or organization can play many roles and the roles evolve and change as new areas are covered. Used the Part, Party Role, Party Role Type model – Party in Role. Party Role subtypes exist.

Key to success is not re-inventing the wheel and using established patterns while still leveraging internal knowledge. Working incrementally and communicate the value of abstract models in supporting change. Interesting presentation, I shall have to think about how rules and analytics match to this kind of structure.

Share

Comments on this entry are closed.

  • Karen Lopez March 24, 2008, 10:38 am

    The most interesting part about Len’s presentation I blogged about was the fact that there are many more of these universal patterns (not just universal data models, but universal patterns) that we should be identifying and using.