Seriously, Are Data Silos Really "Bad For Business"?

November 20, 2018 Dennis D. McDonald

Are silos bad?

When I read the Forbes article Why Data Silos Are Bad For Business I immediately put my snark hat on and thought to myself, "Tell me something we didn't know 20 years ago!"

Still, reading the article had me nodding in agreement: "Yep -- that's true. A sIngle view of the customer is good" and "Yep -- that's true. A failure to collaborate can lead to costly mistakes," and "Yep, that's true, too; without a single source of "truth" it's hard for someone to build and make decisions around a single strategic view."

This is relevant when modern ideas of "digital transformation" call for all parts of an organization to collaborate around the management of processes, technology, and data in ways that align with and support a strategic vision of where the organization needs to go.

Lofty goals

Yet this is a lofty goal that many organizations aspire to but may find difficult to achieve. Such “failures” may not occur for want of trying nor are they necessarily due to a lack of top management support. Legacy systems and processes can be difficult to change, especially in larger or more complex organizations that already have to deal with constant changes in markets and how business is conducted.

Blurred boundaries

Another factor making digital transformation complex is the modern data ecosystem which no longer respects traditional organizational or departmental borders and may already be undergoing a move to cloud services.

We no longer have a "single source of truth" about customers and their transactions given the buying and selling of personal data from multiple sources that can occur online. To think that even a serious data governance effort run out of the IT department that concentrates on master data management can address all the natural variability in data is naive. This is doubly true when naturally occurring variations occur not only in data but also in data literacy. Hiring data scientists to make sense of huge amounts and kinds of data won't help if the consumers of their services don't understand how to interpret data and the context in which predictions are made.

All of this suggests to me that it may not be such a bad idea to let well managed data silos exist as long as how they interact is understood and managed. Effective data governance doesn't automatically have to mean that obtaining a "single view of the customer" should be an automatic goal for all organizations if the resulting administrative overhead or process disruption overwhelms the benefits.

How much change in data governance is necessary?

I feel somewhat guilty in making that assertion since many of my original ideas about data governance were forged in traditional data transformation and database management environments. Nevertheless, I am increasingly aware that effective data governance need not automatically require a monolithic change in how systems and processes are managed -- as long as those systems and processes are effectively supporting the organization's goals and objectives.