10 Critical Myths and Realities of Master Data Management
Prevalent myths surrounding MDM alongside an explanation of the realities.
Andrew White is a Gartner research vice president whose primary research focus is information governance and master data management (MDM). In a recent interview with IT Business Edge’s Loraine Lawson, he explains what Informatica means by “cloud MDM,” as well as the common mistakes companies make when starting MDM. In the second part, he drills down on what other vendors mean by “cloud MDM” and offers advice for how CIOs can cut through the cloud MDM hype.
Lawson: Gartner’s repeatedly said cloud MDM is not mature, and yet vendors are already offering it. In fact, Informatica joined the group this fall. So I’m curious: What do vendors mean when they promote “cloud MDM” and why does Gartner contend cloud MDM is not a mature?
White: The trick, if you will, is in the definition of MDM and even cloud MDM. What Informatica is specifically offering is purpose-built on Force.com only to work with SalesForce.com.
So it’s actually only around prospect data. It is a hosted application, predominantly focused on data integration, data quality, the idea being that business users or the stewards would still actually log onto the application and do all the problem-solving that they have to do with bad data.
From that perspective, it is not all cloud MDM. It’s actually only limited to SalesForce.com data, so it’s very, very narrow. It’s really just data management for SalesForce.com. It’s a bit like an ERP vendor saying, “I want an ERP data management tool for my ERP system.”
It really ought to be called SalesForce.com MDM, really, to communicate to the prospect exactly its limitations, but it is the first of its kind. Not many companies have done it yet.
In fact, the majority of organizations who do MDM internally normally don’t even bother with prospect data in the scope of their governance program, because prospect data is notoriously vague, fluid and you don’t get shot for having dodgy prospect data. But if you have a financial transaction, i.e. they become a customer, then things start to get more important and that’s when information governance really needs to be good. That’s where most of the focus has been for the last six, seven years is trying to govern the core customer data. The idea that we can extend that principle to prospect data is certainly desirable. Visionaries are trying to do that, but the average company, the mass market, they're a long, long way from it.
So, technically, there are some reasonable things here, but there is not a lot of real stewardship support, not the kind of capability you need to stand up to governance process, which, by the way, doesn’t live in a cloud.
The governance process lives on-premise. It doesn’t actually map to where the technology resides. It never did.
Most companies actually don’t even have that technology today; even those implementing MDM aren’t really standing up governance and stewardship in the business very well. That was one of the big topics last week at Symposium. Fifty users — they weren’t customers, they were 50 companies — came to a workshop on first steps with MDM and only two companies said they have really made an effort to make MDM work. The rest said we’re still trying to figure out how the heck to start. And these were big companies that you’d think surely they’ve done it by now.
It’s extremely immature, even now. There’s a difference between technology and discipline. For us, MDM is more discipline than technology, but for some vendors, MDM is just a piece of technology you can license.
Lawson: And what is it for companies? Have they bought the technology and they don’t know the discipline? Or have they just started with nothing?
White: It is actually more of the former — lots of purchases going on. Lots of companies starting MDM, dabbing around at the edges. A number of companies failing for different reasons: They didn’t include the right technology or they had good technology but didn’t focus on the governance, the process and the organization parts. So that’s quite common.
For a given technology, on a Monday you could meet a company who’s done real well with it; on Tuesday you could meet a company that’s done real badly with it. The only discernible difference is not actually in the vendor or the pool, but in the company and its maturity of understanding what governance means, how to set up stewardship, how to make it part of day-to-day behavior by the business, for the business.
This is not an IT problem. IT knows the problem. IT knows how to solve it. The business people have to participate, they have to lead. They have to define and have to steward the data and that’s where the real barriers stand up, unfortunately.
Lawson: When you say sometimes they have the wrong technology, what do you mean?
White: A good example would be a company that has an MDM solution, but they didn’t actually include in their acquisition of the technology the right data quality tools. You can’t do MDM without some data quality capability and there are many of those tools lying around, so you might have the wrong one. We’ve seen that quite a bit, the wrong data quality tool.
Sometimes companies buy an MDM solution, but it turns out their real problem is very heavily work-flow oriented and the solution they bought doesn’t have very good work flow.
Too many companies jump to technologies — if you can just buy it and MDM is like many other initiatives. It’s a lot less about technology and more about how the business behaves.
Lawson: So if you buy the wrong data quality components, how does that happen? I would think those would be pretty standardized. They're not?
White: Oh no, no. Some data quality tools are very good with matching and merging, very common in say customer data. But if your problem is product data and you want to do semantic discovery from a lot of text and engineering strings, you need a different type of skill and not all data quality tools are the same. It’s not very good at say matching and merging and name and address reconciliation and you know, be absolutely useless in semantic text parsing of engineering data for product information. So you have to have the right type of tool for the right type of problem.