Jill Dyche’s favourite New Yorker cartoon shows an old woman talking on the phone to someone who is obviously a telemarketer of some kind. She is saying, “If you don’t stop touching base with me, I’m going to call the police!”
Dyche, a partner with Sherman Oaks, Calif.-based Baseline Consulting, used the cartoon to illustrate a common problem: different parts of the business keep harassing the customer because the systems in each division have different information about that customer.
That’s why Dyche is helping enterprises create master data management (MDM) strategies.
“We’re talking about who someone is, rather than the fact that they used an ATM to take out money,” she explained to local IT executives during a recent seminar in Toronto. “It’s not transactional. It’s subject area-specific. It’s the groupings, categories and hierarchies that exist across a company.”
Vendors have been promising a “single version of the truth” for years, but Dyche predicts more companies will be setting up an MDM “hub,” or repository of master data that becomes the basis for a “golden record” feeding into other systems.
Here are some answers to some common MDM questions to help IT managers get rolling with their own projects.
? How do you define this stuff?
Dyche simply calls it “data about stuff,” but San Mateo, Calif.-based Ventana Research has a more elaborate explanation. “Master data includes information about customers, products, suppliers, regions, hierarchies, business rules and other detailed aspects of a company’s business,” an addendum to a report on MDM it released last month reads.
“An MDM system delivers the capability to define master data (including definitions, references and metadata) uniformly across the organization and synchronize its use. It allows organizations to retain and derive maximum value from their existing IT investments.”
? What if I’ve already deployed CRM tools or set up a data warehouse?
Not good enough, the experts say. In many cases, you’re more likely to take on MDM because those things failed.
“ERP systems didn’t solve the problem, CRM didn’t solve the problem, data warehousing — go down the list,” says Marty Mosley, CTO with Initiate Systems in Chicago.
Anurag Wadehra, vice-president of marketing and product management with Siperian in San Mateo, Calif., agrees.
“Data warehousing is good for reporting and analysis after the fact. It is not where data conflicts are resolved,” he says. “Managing the master data — the identifier and relationships and the resolution of conflicting identifiers and making a source of truth across all systems — is a completely different task.”
? Should I pick a point product or a module to an existing platform like SAP or Oracle?
The point solution vendors, not surprisingly, don’t advise the latter. “A company having a homogeneous system, a single platform that manages all their assets, is very, very rare,” Mosley says. “SAP works great if you’re an SAP shop, but especially if you’re using software as a service, you’re dealing with a lot of data residing outside the firewall. How does a closed big box architecture like SAP or IBM play in that world?”
David Stodder, vice-president and research director of information management at Ventana Research, says MDM is more about figuring out what problem you want to solve first.
“I think it’s important for people to look at the whole market of possible tools that includes data mapping, profiling and data relationship discovery,” he says. “We’re discovering that there are MDM projects that fall apart because they can’t even find the data.”
? Isn’t a hub just another form of data warehouse?
Um, yeah. Kind of. “It’s like a data warehouse where you’re trying to load what you know about the data systems that have data about customers and pull that together so you can establish a single view. I think that’s a natural way to go about doing it,” Stodder says.
“What’s beyond that is somewhat more distributed kind of approach to the problem that doesn’t create this new hub that has to be managed all by itself, with another set of staff, hardware and software and so forth.”
Wadehra, whose company sells a hub product, says it should, a) identify customer, product or other master data across multiple data sources b) resolve any discrepancies between identifying information among various systems c) relate pieces of data, and d) route it to other systems.
Mosley says a lot of the MDM thinking is being driven out of the DW/BI world. “People end up thinking we’ll just build another DW and we’ll call it MDM. ‘So we’ve got a hub, let’s stuff everything in it.’ You have to really focus on one kind of data and forget the rest,” he says.
? How long will it take before we see results? You’re probably in this for the long haul, but most people suggest you could expect to hit some milestones within a 90-day period.
“Start with a high-stakes process that addresses some underlying data issues,” Wadehra suggests. “It could be your CRM process, your cash to order process, whatever. You’ve got to pick up one domain and deliver a system that goes live in that particular area and resolves the conflicts.” Initiate promises “time to value” within three to six months, which Mosley says is a lot better than the tens of millions for data and process modelling with other firms.
“What you’re doing is going in and instead of trying to do data modelling, you’re touching the data that seems to matter most — profiling, running through data quality checks, matching, getting stats on duplication rates. Then you’re populating that in a service that is available in your enterprise.”
Stodder notes that the whole psychology of software development has shifted towards programming where you can have some sort of deliverable results incrementally.
“The ultimate journey may take a couple of years,” he says. “MDM ties into the whole idea of data governance, which is a continous process. This is really something you have to think of as a destination.”
