"With the economy the way it is, the quick buck for our customers is MDM on the BI side. Over the past two years the appetite for broader strategic projects was just not there from a CIO perspective," said Aaron Mahamainathan, senior director for solution marketing at SAP AG, the German business software giant.
Implementing an MDM system for BI is not just less difficult than tackling operational systems; its impact is significant in its own right, according to Mahamainathan. Because CEOs increasingly see BI projects as a strategic tool, the data quality in a BI initiative is critical: It makes the difference between a profit or a loss, a successful ad campaign or a bust.
"What we've seen is CIOs saying, 'Let's fix the data going into BI, maybe starting in one area, for example, customer master data or supplier; and in the end we'll know we have a better-quality spend-analysis or a better dashboard, and so on,'" Mahamainathan said.
The term "master data management" typically refers to the processes, tools and technology needed to form and maintain accurate and consistent lists of an organization's critical business information. The business benefits of having a so-called "single version of the truth" are obvious. The screwups resulting from a lack of enterprise MDM are legion: the software company that tries to sell its newly acquired product line to a customer that owns it already; the bank that tries to sell a mortgage to a customer who already has one.
The holy grail of an enterprise MDM system is an application-agnostic central hub that feeds master data across the enterprise. ERP and other systems still manage the type of master data they need to run their business processes, but the key element -- supply data, product data, and so forth -- comes in an integrated fashion from the central hub.
Even MDM enthusiasts, however, acknowledge that getting to the so-called single version of the truth -- in a centralized hub, no less -- is tough. "It's a journey," said Colin White, president of consultancy BI Research in Ashland, Ore., who has covered MDM extensively.
Indeed, applied to the entire enterprise, MDM can be a multiyear process, White said, that starts with the identification of all the sources of a company's master data. Master data records must be matched to eliminate duplicates, not an easy task. Along the way, the initiative can require changing business processes to accommodate the master data that will reside in a central hub, as well as putting in the rules to make sure the master data stays "clean."
"This is not a trivial pursuit, to manage master data on the transactional side. It tends to require a lot of governance, and a there is a lot of change management involved in it," SAP's Mahamainathan said.
Because a BI system is a consumer of data that is created elsewhere -- for example, in an ERP or a customer relationship management (CRM) system -- applying MDM to BI data is less complex than fixing the master data in transactional systems, according to these experts. Once the data is cleaned, maintenance is minimal. A BI system still stores the master data it uses, but it should not manage, govern or own it. The central MDM hub does that.
BI on the fly
Not everyone is convinced applying MDM to BI is worth it -- or even feasible. Joseph Marcella, CIO and director of information technologies for the city of Las Vegas, has opted to mine the databases of his highly distributed environment "on the fly." Las Vegas' city government has 18 business units, each of which is vertical in nature, comprising multiple governmental divisions. Building, maintaining and managing a separate database for master data is challenging. To take just one example, the fire department defines the term "public safety" very differently (getting to a fire in under six minutes) from the way other areas of the city government do.
"Everything we do has a public safety component, and every one of the databases has a public safety component," said Marcella, who uses Oracle Corp.'s Business Intelligence Suite-Enterprise Edition. "How in heaven's name am I ever going to be able to always gather all of that in a timely fashion with a level of integrity in a separate database? Very difficult to do."
To mine anything for any one particular department, Marcella said, he would want to make sure he was looking at data "with a level of integrity that represents what has really happened at a business level from both a data and time perspective. Even if the data is wrong, it would be consistently wrong in the same way," and measureable.
"If you are capable of mining [data for] your organization as your organization is running, according to the business it is in, then you are able to give management the kind of information they need to run the place," Marcella asserted.
Let us know what you think about the story; email Linda Tucci, Senior News Writer.