Is it better to have a centralized or decentralized master data structure?

Is it better to have a centralized or decentralized master data structure?

I just have a quick question. Which do you think is better — to have the master data (sales, plant, material, finance) centralized or decentralized? I am just thinking it is best that it is centralized to keep master data aligned to each other, but I'd love an expert's opinion.
A quick question, huh? I suppose you go to the Appalachian Trail for a "quick walk," and grab a "quick bite" at Le Cirque? But I'll give it a shot. The management of the data needs to be centralized, but where the data is persisted can be decentralized. I know that was quick, but like your question it's deceptively complex.

You see, as much as we like to focus on the platforms, data management — the tactics around defining and maintaining business rules, definitions and access to data — is the real hard part. Best-practice companies have learned that centralizing data management as an enterprise service is not only the most efficient way to support MDM, but a whole range of enterprise initiatives.

As for master data, it really does depend on whether you're using registry-style MDM to quickly match and link data across source systems, or a persistent approach that stores the data in a centralized hub. Each has its advantages and disadvantages. To understand which is best for your company — registry, persistent, or a hybrid of each — see my prior Q/A on MDM functional requirements .

 

This was first published in August 2010