A common data model lets organizations develop a real sense of how to extract insights from the data, and to do so quickly and cheaply.
Ask any data manager what their biggest pet peeve is, and there is a good chance the response will be data complexity. This should come as no surprise. Quite simply, most organizations are creating or collecting more data than they can keep up with or easily derive genuine business value from.
A number of factors contribute to data complexity. The most obvious are the sheer volume of data and the fact that it is being created by different applications, which often don’t talk to each other. This makes it time and resource-consuming, not to mention extremely expensive, to gain business insights from data on an organization’s customers, products, partners, and more.
All of this has motivated many organizations to seek open data model solutions to unify their software and applications. For many organizations, the solution to managing and analyzing this data and achieving business insights in real-time is to adopt a common data model approach. In simple terms, a common data model unifies hundreds of data sources into a standard view with no procedural ETL required. In that way, when business teams are sharing their experiences, they will all be talking the same language. That enables the organization to develop a real sense of how to extract insights from the data, and to do so quickly and cheaply.
The Great Buy vs. Build Debate
Once an organization decides to migrate to a common data model, it must then decide whether to build it on their own or purchase a commercial solution. Many organizations will try to design and build a common data model on their own, which entails building the data model on top of existing data analytics tools. It also generally requires a significant amount of effort and cost to build, maintain, and map all of the data to it, etc.
Organizations that opt to purchase a commercial solution tend to either augment the common data model they have already built or purchase the entire data model out-of-the-box. A commercial solution can also provide deeper industry-wide depth to the data model than the organization could have achieved on their own. On the flip side, when an organization builds the model on their own, they may only develop what is currently needed for data management and have a harder time scaling the data management and analytics practices later.
Determining Data Model Ownership and Avoiding Pitfalls
Normally, when you buy the data model, you’re actually buying a way for your business team to own it. When you build it, the data team owns it on behalf of business users. Building a common data model is a significant expense, both from the perspective of just how big and deep your data expertise has to be. So if you have decided to build it, assume you are taking on a very significant project that will require significant resources. My advice: start small initially.
Also, with building a common data model yourself, by the very nature of doing that, you are becoming a service organization. The internal data team builds the data model for the business users, and immediately becomes a service organization that has to answer all subsequent business requests regarding data.
There are also quite a number of drawbacks that organizations should be aware of before deciding to build a common data model. Among them is a clear lack of agility as the business asks for more from the model. This may be unacceptable to many organizations, especially ones that need real-time business insights from their data.
Consider some of the challenges: Any desired change in data management and analytics processes can take a long time because you have to re-do the code and systems to manage whatever the business is trying to do. The cost of changes or new additions in features or functionality can be very high.
So if you build something and a new source comes in, or a new use case emerges that requires an enhancement to the data model, it can quickly become a significant engineering project. Add to that the fact that data teams are typically understaffed and under-funded from a data perspective. So by taking on the project yourself, you risk always lagging behind what is really needed.
Challenges with Creating an Industry Standard Data Model
One of the biggest challenges with a do-it-yourself approach is with creating an industry-specific comprehensive data model that is available immediately to quickly deploy solutions that business users need. With most organizations, there are already significant industry initiatives on the business side to really understand data and assess what data is available. At the same time, every organization is always unique in some sense. Typically, 80 percent of your business is running on applications that are standard, and 20 percent are applications unique to you.
The advantage of adopting an industry-standard whenever possible is that it lets the organization focus on its key differentiators, so they can add to the data model the things that it really is an expert on and unique in. In addition, reusing a single common data model can prevent the need for multiple and app-specific ones.
When adopting a third-party solution, an important consideration is how easy it is to adapt the common data model to a user’s unique business requirements. Organizations must assume that customization is going to be a key piece of it, both because companies are a unique mix of business entities in how they run their business, and also because the exact business rules depend on where the business is.
You always have to assume that the data model will be customizable, otherwise, it’s not worth your time. If it’s completely fixed, it’s not going to serve your needs. And the business has to own that customization.
Best Practices for Realizing Top Benefits
Finally, organizations should follow best practices for getting the top benefits from a common data model. One is the direct involvement of the business domain experts in managing and updating the data model – really being particular about giving ownership to the users who understand the business and can maintain and define the model, rather than outsourcing it.
The second important thing, even though the data model covers most use cases, is for organizations to start piecemeal. One thing the data model can allow you to do is to start taking business use cases and give value and impact at low cost immediately. You would have to do a massive data program to bring all the data together. But you don’t have to do that. You can start with one business entity, then add a second business entity, and start to really leverage the value of the common data model in a piece-by-piece manner.