Master data management -- MDM. It's new, it's hot, and nearly every corporate technology department is talking about it -- and for good reason. The problems associated with data quality have been building within enterprises for years. As businesses strive to become more nimble and customer-centric, the endless parade of enterprise applications they've deployed have become a series of silos--and silos are precisely what MDM, at its best, does away with, revealing scads of actionable insight hiding within.
But with this emerging popularity has come the common integration and deployment woes associated with any enterprise IT system. MDM, which cuts a broad swath across nearly all enterprise applications and users, only magnifies and multiplies these problems. As a result, enterprises have been stymied in their attempts to deploy MDM, says Jill Dyche, a partner and cofounder of Baseline Consulting. "There's been a slower adoption curve than we expected," Dyche says, "especially given the increased maturity of the software. But for those companies that have overcome these deployment woes, the ROI can exceed their wildest expectations."
A staple of any CRM initiative, securing senior-level buy-in is even more critical for an MDM project, especially considering the level of confusion and lack of knowledge surrounding this technology. C-level executives have long been believers in the benefits of data warehousing, which has become the de facto standard among enterprise IT departments for providing the proverbial "single version of truth." But management's belief in an existing solution makes taking it to the next level a tough sell. What carries the day is being able to quantify MDM's added business benefits. "IT wants to push this 'MDM hub' concept, and the natural response from business leaders is, 'Don't we have that already with data warehousing?' " Dyche says. "Many companies are still approaching MDM from the infrastructure/IT standpoint, and that's just not as sexy or alluring to the business leader who doesn't fundamentally understand how it will improve his or her business."
The answer is in quality and breadth, far beyond the sheer volume of data in a data store or warehouse. With one of those, data integrity and validity are optional: The information a data warehouse collects could be good or bad, depending upon the surrounding environment, and lacks the data-standardization, -cleansing, and -reconciliation capabilities inherent to an MDM hub. Any pitches for MDM should play up the business benefits that come with the added functionality. Dyche offers healthcare as a perfect example: "Using MDM we can create a master patient index so our affiliates and partners can all look at the same customer data, or so we can accurately report spending guidelines to the government because we can individualize physician information," she says.
So now, CRM magazine offers the best tips and pieces of advice that industry pundits have to offer in terms of implementing an MDM solution. Many of these tips may have a familiar ring for those who already have CRM deployments under their belt, says James Kobielus, principal analyst of data management at Current Analysis. "MDM is just like CRM: It takes an enterprisewide shift in the way you conduct business to be successful, but done properly, the payoff is huge." (See Kobielus' four-part series for CRM magazine on data and MDM here.)
1) Small Strokes: Keep Things Focused
Simply put, MDM projects are big, thanks largely to the scale and number of enterprise applications and data sources they touch upon. The resulting upheaval can be quite disruptive, and will impact users at nearly every level of the organization. That said, rather than trying to boil the ocean with an enterprisewide implementation, many industry pundits now recommend taking a point approach by starting with a particular type of data domain, such as tackling a customer data integration (CDI) implementation or by resolving disparities in product information with a product information management (PIM) solution. Any entry point will be driven primarily by the type of industry your company operates in and the data domains that are most business-critical. These initial gains will form the basis of your ROI, and can then be used as a beachhead into further MDM expansion. "Many times, your initial starting point will expand naturally as you look to include other data sets, such as product information or account and billing data, or other lines of business," Kobielus says.
That starting point -- and the eventual expansion -- should also drive your vendor selection, Kobielus notes. "If all you want is CDI, guys such as DataFlux should be on your shortlist, but if you're looking for a comprehensive MDM vendor, you'll want to turn to IBM, Oracle, or perhaps Purisma," he says, referring to the pure-play data vendor recently snapped up by Dun & Bradstreet. In other words, when it comes to selecting a vendor, make sure to balance any short-term gains against your long-term expectations.
2) Don't Mix Your Paints: Differentiate Between Governance and Management
All too often, companies fail to differentiate between data governance and data management, leading to a fundamental breakdown in communication between technology folks and executives. Data governance refers to the corporate rules surrounding data, which in the context of MDM would include defining the attributes and characteristics of different data sets, determining who owns the data, and how employees should access and use it. "It's the policy-making around decision rights and who owns the data," Dyche says. "For example, it's not a good idea to have a programmer during the middle of an [extract, transform, and load] job determining if revenue data should be categorized as either booked or billed data. It's not that person's decision. That's what data governance needs to address."
Data management, on the other hand, is the tactical execution of such policies, and primarily makes sure that particular IT personnel handle specific issues to guarantee and maintain data quality, privacy, and access issues. "Data governance is an overall business management issue," Dyche says. "Data management is an IT issue. It's ensuring that the proper people are tasked with data-specific jobs and not roving linebackers who address an issue when they have a free minute."
Dyche says it remains one of the biggest obtacles to current MDM adoption. "We'll have clients ask us to spend a day explaining data governance to them, and we'll show up expecting to formalize the decision-making processes around data," she says. "Instead, they want to know how their metadata repository should be accessed and updated by the IT department. There's a great deal of confusion in the market, and it's critical [that] companies understand and define the differences between the two."
3) Clean Brushes: Data Quality
Another key to any MDM implementation, and the primary determinant of how much an MDM rollout will cost, is the underlying data quality and integration tool sets. The installation of enterprise information integration (EII) and operational data stores means the company has laid a firm foundation for MDM, says Ray Wang, principal analyst of enterprise applications at Forrester Research. "The beauty of the MDM hub approach is that many organizations already have the pieces in place," he says. "They just need to find a way to pull it together." Depending on the level of preparedness, a company could spend up to $5 million for licenses and implementation services for an MDM implementation, and spend anywhere from a year to 36 months rolling out the software, Wang says, though that figure can drop to as little as six months for a well-prepped company.
"MDM is essentially data warehousing on steroids," Kobielus says, and companies that have taken the time to embed data quality practices and data governance workflows have a head start. "It's just like the old saying," Kobielus says. "'Dirty data in means dirty data out.' Without data quality, it's not MDM, it's Steven Colbert: 'There's a truthiness there, but I wouldn't trust that truthiness.' " Before pursuing MDM a company must establish data governance processes and technology safeguards across the enterprise, because "departments that have been operating independently of each other must operate in unison for the greater good of the company's data quality practice," says John Radcliffe, a research vice president at Gartner.
4) Select the Right Easel: Clarify Your MDM Architecture
MDM solutions come in various shapes and sizes. Accordingly, businesses should consider the types of architectures that their organizations could opt for when deploying MDM software, Radcliffe says. Companies can pursue two avenues with MDM: a transaction style, or centralized architecture, which leverages a unified hub that acts as the "golden record" of data definitions and attributes, pushing data out to enterprise applications and end users as requested. The other is a federated (or registry) style, which is a distributed model that allows data to reside in its respective repositories, pulling data through a centralized hub that transparently points enterprise applications in the right direction.
Each path has its advantages and disadvantages, many of which are contingent on the kind of business environment you operate in. For centralized, IT-driven businesses that have limited numbers of data domains and lines of businesses, a transaction style might fit best. For multinational organizations operating multiple databases and systems, a registry style might be the best choice. The key, Kobielus says, is making sure the architecture you select best matches your data governance and workflows. "Just like CRM, it's critical to make sure that the technology aligns with the business processes," he says.
5) The Canvas Beneath: Services-Oriented Architecture
The enabler of MDM -- and the reason it wasn't developed earlier -- is SOA, which is what gives MDM the ability to synchronize data across systems as a set of Web services. Accordingly, most industry pundits categorize MDM as being an "enabler" of SOA, but that's just half the story: SOA is also a precursor of MDM -- a symbiotic relationship that many enterprises aren't prepared for. "A lot of companies aren't ready yet with their SOA environment, so in that sense, MDM is ahead of the curve," Dyche says. "They need MDM badly, but they're relegated to using it in batch modes until they can get their service architecture solidified."
Dyche recommends making sure that your technology department has a firm grasp on its architecture and roadmap before moving forward with MDM, lest the project be hindered by a flawed or inadequate SOA. "The company should have at least laid the basis for their services structure. You don't need SOA to implement MDM, but you need it to really make MDM flourish, which in turn will lay the foundation for SOA."
6) Your Art Patron: Find Yourself a Solid Service Provider
Make sure to partner with a vendor or service provider with a significant track record in master data management, because chances are you don't have the proficiencies within your in-house IT staff. This is due to both the relative immaturity of the MDM market and the domain expertise required to implement an MDM solution. "Just as important as your vendor is your service provider," Kobielus says. "MDM brings together a lot of things, so it still confuses a lot of people."
Kobielus also recommends making sure that whichever third-party integrator you decide to team with is knowledgable about your industry. "Most MDM solutions have been heavily customized to your specific horizontal requirements, but not necessarily in your vertical. That's important to note, because nearly all MDM solutions require some degree of customization -- customization that will be primarily driven by the data domains inherent to your industry."
You can reach the editors of CRM magazine at editor@destinationCRM.com.