The Mainstreaming of Master Data Management
A new survey identifies a strong correlation between doing DW and successfully pursuing MDM.
- By Stephen Swoyer
- May 26, 2010
A new survey from master data management (MDM) consultancy Information Difference makes a compelling case that MDM has arrived.
It's been a long time coming. Although MDM adoption has yet to cross the magic 50 percent threshold, it's close, according to Information Difference. Writes industry veteran Andy Haylor, the consultancy's president and CEO, MDM is being practiced in nearly half (46 percent) of shops that have live data warehouse (DW) implementations.
Not all of these organizations have production MDM deployments, of course. Nonetheless, says Haylor, there does seem to be a relationship between DW and MDM. Think of it in the context of a data management continuum: i.e., shops with DW practices are also more likely to have adopted MDM in either trial or live deployments. DW practitioners are likewise ahead of the curve when it comes to the maturity of their MDM implementations: shops that have both MDM and DW are more likely than MDM-only organizations to be working with live MDM efforts. MDM-only adopters, by and large, are still stuck in the trial stages.
The study's title -- The Link Between Data Warehousing and MDM -- underscores its conclusion: namely, that a mature data management infrastructure, anchored by one or more data warehouse systems, is a good predictor of MDM success. Though the survey's approach raises questions about confirmation bias -- its questionnaire posits a relationship (positive, negative, or as-yet-undetermined) between MDM and data warehousing -- Haylor says there's a strong correlation between doing DW and successfully pursuing MDM.
For example, although a sizeable share (30 percent) of DW shops don't yet have MDM implementations, nearly half (46 percent) do; what's more, just 7 percent of MDM adopters don't already have an extant data warehouse.
"[T]he majority of those organizations with both DW and MDM had live implementations -- compared with just 14 percent for those with MDM alone," writes Haylor, who deems it "encouraging that a relatively high proportion of organizations has chosen to implement MDM alongside their data warehouses."
The Information Difference survey collected responses from 208 people, more than half of whom (57 percent) were North American respondents; just over one-quarter (27 percent) of respondents live in the EU, with the rest (16 percent) represent other regions. Crucially, one-third of survey respondents had business backgrounds; the rest (67 percent) said they had IT responsibilities.
The survey paints a picture of near-ceaseless MDM activity -- among DW-savvy shops, that is. Again, not only are shops that have DWs more likely than shops that don't have DWs to adopt MDM, but nearly half (48 percent) of shops with data warehouses are already supporting live MDM deployments.
Of the remainder, 51 percent say they're either planning or actively developing MDM deployments. In other words, if you've got a data warehouse, there's an even chance you're mulling (or have actually deployed) MDM, too.
Haylor also points out that DW-savvy MDM adopters are thinking big. Nearly three-fifths (57 percent) are aiming for (or have implemented) enterprisewide MDM deployments; another quarter say they're shooting for MDM implementations that span multiple departments or locations.
"[T]his result shows that organizations understand the importance of reaching a broad scope for analytic MDM in order to ensure quality business information," Haylor writes, noting that the number for enterprisewide MDM deployments was much lower (at 36 percent) among shops that are pursuing a master data management strategy in the absence of a DW.
The Information Difference survey also sheds light on the relative size of MDM deployments (the reported range spanned 100,000 to 500 million records; the mean was 109 million records; the median, 4 million) as well as the number of master data domains adopters are working with. In the latter case, not surprisingly, adopters ranked "customer" data as most popular, with "product" data second, "location" data third, and "supplier" and "financial" data ranked at fourth and fifth, respectively.
More surprising domain entries include "intellectual property" data (ranked ninth), "research and development" data (tenth), and "project" data (eleventh).
Most shops are already working with multiple MDM domains: the survey mean was four domains, the median three. Most shops likewise believe that master data should be sourced directly from an MDM system, according to Haylor: four-fifths of respondents endorsed such a topology. What's more, he notes, more than one-third (36 percent) are already doing as much.
It's more proof that MDM is growing up, Haylor argues. "Although 'Customer' and 'Product' received the highest ranking and were most frequently selected, a wide range of data domains was generally selected," he writes. "On average, the number of domains selected was four[,] with a median of three, showing that while 'Customer' and 'Product' still have high priority, most organizations have a need to cover a broader range of data domains in MDM."
Haylor and Information Difference aren't alone in touting a DW-linked approach to MDM. MDM authority Jill Dyché -- a principal with Baseline Consulting (which -- along with several industry media outlets -- co-sponsored Haylor's study) -- has championed a DW-linked but discrete approach to MDM for some time.
When MDM first came to the fore, Dyché notes, some proponents tried to position it as a zero-sum replacement for the data warehouse. This is neither feasible nor desirable, Dyché argues; in the same way, she stresses, it's dangerous to think of MDM as a subset of the data warehouse. MDM, like DW, is a distinct entity; it deserves what she calls a "seat at the table."
"The data warehouse wasn't designed to clean the data, to apply business rules to the data, or to link back to the operational systems," said Dyché, in an interview earlier this year. "All of the stuff that we define in our MDM stack is sort of outside the core data warehouse."
The good news, Dyché said, is that this argument is no longer cast (or perceived) in zero-sum terms: MDM isn't going to replace the data warehouse. Increasingly, she noted, it's being cast as an essential data integration service.
"The good news is that it's really starting to be institutionalized as part of the formal information management architecture. MDM has a permanent seat at the table," she said. "This also blurs the lines again: just as everybody was starting to distinguish the distinct value of MDM -- versus something like an operational data store, versus the data warehouse -- the lines will blur again as everything gets poured into that one data integration bucket."