Silos are here to stay.
I know when I’ve been beaten, and I’m raising the white flag on this one: Silos are just the way things are, and all the signs suggest we might as well accept that.
Data has just gotten too darn big to store in one place — and that’s true even before Big Data. Why do you think Hadoop is such a big whoop? It’s because it takes all those tiny storage nodes and allows them to function like one big computational machine.
But the biggest reason to accept the silo-ification of the enterprise is that we’ve tried the other approach — building one huge “system of record” — and it simply hasn’t worked.
I recently talked with The Hackett Group’s Chief Research Officer Michel Janssen and Global IT Advisory Program Leader John Reeves about the “borderless business,” and why integration work is growing rather than shrinking.
I asked what they thought of this idea that silos would never go away, and as a result, the idea that we just need to embrace integration as the status quo for enterprise technology.
“There have been companies in the past that have tried to build literally a physical center of the universe for the information that is growing exponentially,” Reeves said. “I don’t think you’ll ever be able to realistically get there.”
In fact, he later identified it as a “worst practice” that companies have tried and realized won’t work.
So if we can’t have one system of record, where does that leave us?
It leaves us with a significant amount of governance work.
“When I talk to folks about a single source of truth or consolidating information, what you're really doing is getting to a set of standards that decides what data looks like with the structure of the data,” he said. “And then at the end of the day, who really owns it? Which system really owns the master copy of that? It doesn’t mean that you have one copy, it just means that you may have 15 different copies, but at the end of the day you know who has the truth, that single record.”
That’s why master data management becomes more important — not because it’s going to act as a single repository, but because it can be used to manage a single record across different data sources, he said.
“The benefit in master data management has absolutely nothing to do with the tools that you use,” Reeves said. “There is some implications to that in terms of the implementation, but all the benefit comes from the data governance, which is a very difficult problem to solve. It has nothing to do with technology. But if you can crack that, that’s where you're going to get the benefit.”
But, and there’s always a “but,” there are two types of MDM solutions: the central repository MDM solutions and the registry or index-style MDM hubs.
This is the kind of situation where an MDM registry would shine, since it doesn’t require you to move anything, but rather maintains an index of the data while leaving the data in place.
But — didn’t I just say there’s always a “but?” — as Aaron Zornes of the MDM Institute pointed out earlier this year, many vendors have sort of “forgotten” about that option.
It may be time for them to remember, because as Reeves points out, migrating everything to one information platform has proven too hard, not just in terms of technology, but politically as well.