Large organizations typically have numerous operational systems that each store and produce competing versions of master data such as "customer" and "product".
Master data hubs is one approach to breaking down these application silos and producing a unified version of "golden copy" master data e.g. a common source of customer account information or product codes.
The master data management (MDM) market has seen a battle of ideas over the last couple of years between "cross domain" approaches and specialist hubs for separate classes of master data, particularly "customer" and "product".
This war is now warming up. Vendors from a heritage of dealing with customer data are typically strong on matching potentially duplicate customer information, often in real-time, and often have high performance hubs capable of dealing with large volumes of customer data.
Vendors from a product heritage typically have elaborate hierarchy management capabilities and are capable of dealing with complex data structures with many levels of depth.
The reverse is rarely true. Until recently these two worlds have barely met, but customers are increasingly demanding an integrated approach to their different types of master data, and the vendors are scrambling to respond.
The specialist hub approach has drawbacks from an architectural standpoint, especially in situations where companies are multi-national. Firstly, there are many types of master data beyond customer and product (such as asset, location, supplier, contract,...) , something beginning to be acknowledged by the industry.
Customers buy products, so it is a problem if the master data system specifically for customer data is different from the one dealing with products.
The whole idea of MDM is to improve the current messy situation where application silos compete over the ownership of master data.
Yet a proliferation of separate specialist hubs may result in a new generation of silos. For multi-national companies the problem is compounded by the fact that each country or cluster of countries will have its own set of operational systems.
Not only does master data need to be managed across systems within a country, but subsets of this data, such as international product codes, need to be co-ordinated across the enterprise as well.
In many cases it is impractical for enterprises to deploy a single hub or hubs at the enterprise level.
A single cross-domain hub at the enterprise level will have serious operational demands placed on it. The more systems it is connected to, the scarier it will be if the hub runs out of steam or breaks - are you going to stop opening new customer accounts while your master data hub comes back on-line?
Eventually what will be needed are technologies that can avoid this bottleneck by achieving synchronisation across multiple instances of a managed federation of hubs (rather as enterprise email directories are replicated in a managed way across multiple systems), and yet few vendors have even begun to address this issue.