Episode 8: Master data management with secure and scalable architecture.
In this episode we are discussing Master Data Management (MDM).
The most often cited reason for implementing MDM is to reduce bad data in the organization. Cleaning data is a difficult, repetitive and a cumbersome process. Your data also lives in different silos such as a CRM, Accounting, and HR. There could be hundreds of data sources. Who holds the truth? These projects often fail for a number of reasons. Complexity, politics, lack of domain expertise.
So, Today we get a chance to speak with Cort Fritz, a Master data specialist with a number of years in this field. He is always fun to talk to. We get into what is master data, why it’s important, what is a good starting point, what to do with old systems, APIs, and if there ever needs to be an intelligence aspect to MDM.
There is a huge IT spend on MDM projects yet so many fail. You’ll want to listen to this episode for expert insight into making these projects successful.
Master Data Management and the quest to reduce dirty data in the organization and eliminate silos.
What is Master Data and Why is it Important? (2:03)
2:03- Defining Master Data
Data produced by all the parts of an enterprise throughout a normal business cycle.
Master data = the eternal entities in an organization. Includes core and transactional data.
But it's hard to define in a complete, satisfying way.
Anytime you dealing with master data, you're dealing with CRM.
Complexity can be overwhelming. Especially understanding how systems transact with each other.
It's important to get a 360 degree view. Need a hub. Cost of Metcalfe’s Law.
Domain-Driven Design. Concepts and resources. Shared language within a domain allows an organization to move very quickly.
Differences Between Master Data and a Dashboard View or Visual (13:02)
13:02- Distinguishing uniqueness of master data.
Correlation and rationalization/harmonization requires master data.
How Does an Enterprise Start Master Data Management? (16:20)
16:20- Starting points for master data management
Understand your domain and/or perform a data inventory. Where things are and how they're being used.
Need to map the functions and what they do (what is a buyer? who is a customer?)
Start small with mapping (5-6 domains)
Master data management projects are heavily back-loaded with benefits not being realized for, sometimes, a year or two or more. Need buy-in and upfront understanding.
Master data projects can be a real fight inside some organizations.
Need to allow people to do their jobs while generating the tools and insights the business needs.
Legacy Systems with Difficult Data (28:21)
28:21- How do you deal with the old stuff?
Dump them and kill them. (Usually doesn't work very well - it's really hard to shut off old systems completely)
Phase out old systems in an organized way
Replace the old system or augment it. But be careful because the replacement needs to be better.
Architect the new system to be protected against the dirt of the old systems.
Earned Complexity as a Guide for Leveraging Hard Technology in Master Data Projects (35:58)
35:58- Earned complexity applied to master data management
Old school is powerful.
Master data is a primitive industry right now.
There isn't space yet for fancy tools like ML and AI.
Build and trust your models and patterns. Build them incrementally and scale slowly.
Understanding what AI and ML is and isn't should be a required business skill in modern business.
Cort Fritz has been leading technology teams since 1995. He is the founder of rheli.co - a software and hardware product development outsourcing provider.
His teams have built global-scale, consumer-grade systems for Xbox, Disney, Yahoo, MTV, Showtime, and many other global brands.
His current interests include Internet-of-Things, the elixir programming language, music, and modern lit.
He is most proud of being a dad to his two daughters, Liberty & Reiko and lives in Los Angeles, CA.