R.R. Donnelley was going through a growth spurt. It was 2005 and the company's revenue had jumped from $2.4 billion in 2003 to $8.4 billion, much of it coming from numerous acquisitions the printing giant had made.
But all that growth was creating information management challenges because customer data from the acquired organizations was stored in multiple source systems. And Donnelley, like many companies its size, had to rely on a manually intensive process to get a comprehensive picture of its customers, such as which Donnelley unit they might be doing business with and what additional products they might want to buy.
So the company set out to find a better way to manage its data.
At the end of 2005, Donnelley embarked on an MDM program, an effort to create a single set of identifiers throughout the enterprise for its customer and vendor data. Having a master of each record would enable the company's applications to tap the most accurate, complete and up-to-date information.
Donnelley is one of a growing number of companies seeking the benefits of MDM, which can increase corporate effectiveness, reduce costs and meet compliance requirements. IT research company Gartner expects that when MDM software revenues are tallied for 2008 they'll hit $1.3 billion, a 24 percent increase over the previous year. While implementation cost estimates range from a couple of hundred thousand dollars to $2 million for big companies, organizations can see an almost immediate return. Some realize a payback in as little as three months, according to veteran MDM consultant William McKnight, a partner with Lucidity Consulting Group, although he cautions those kind of results are only possible if the deployment is done properly.
Indeed, the benefits of MDM aren't always easy to come by. Most efforts experience organizational, political or governance bumps, according to Andrew White, a research vice president at Gartner who focuses on MDM.
Companies need to agree on how data will be mastered, how data discrepancies will be resolved and who'll make those decisions and be in charge of the process. There are also the budget, staffing and integration challenges that come with any project that spans an enterprise.
And, at Donnelley, the company was dealing with specific challenges that would be expected in an organization that recently made a number of acquisitions. These included issues such as data stewardship - identifying who in the business should "own" the customer information - and inconsistent data definitions that resulted from customer data being in multiple systems.
From a management standpoint, "if you asked one group in the business what 'customer' means, and you asked a different group what 'customer' means, you'd get two completely different answers," says Kim Fahey, a senior director of IT at Donnelley. One group may think of a customer as a specific billing location, another might consider the customer as the legal parent entity.
In the third quarter of 2006, Donnelley brought up its Customer Master Data Store, which now contains more than 250,000 records and integrates numerous source systems - mostly those of its major acquisitions - allowing the company to quickly assemble data from acquired companies, identify top customers and spot sales opportunities. It's become a vital sales and marketing tool.
The data is used every day, according to Fahey, who says the company is now ready to master a second data set - its vendor information.
The IT team had to overcome a number of obstacles, but the effort to date has earned the praise of industry experts. "It's an early success story that others can hope to emulate," says consultant McKnight.
Three Steps To MDM
R.R. Donnelley today is an $11-billion-plus commercial printing and services company offering direct mail, financial printing, forms and labels, and other services. Since the Chicago-based company's $3 billion purchase of commercial printer Moore Wallace four years ago, it has picked up a number of printing-related companies in a series of acquisitions, including its $1.3 billion purchase of printing and supply chain management company Banta, a deal that closed in January 2007.
It was this history of acquisitions that made MDM so essential to the company's operations. Its accumulated size made it impractical to put all its information into one system, necessitating the need to work off multiple data stores. And prior to 2005, the company's existing processes for consolidating data were predominantly manual and time-consuming.
For Donnelley, Fahey says, "MDM was not an option, it was something [the company] had to do."
The company built its current MDM platform in basically three stages: a discovery, strategy and architecture stage; a building stage; and a renovation stage.
When it began with the discovery stage, it recognized that whatever MDM solution it chose, the system would need to be flexible enough to accommodate further corporate acquisitions with their corresponding data sets.
The company also understood that the deployment would need to be minimally invasive to the source systems, so that there would be no disruptions to the business. The solution was a registry model where customer information comes in from source systems, is registered in the hub, cross referenced with other records and placed in a master file - leaving the source system untouched for reporting or analysis.
MDM consultant McKnight says a registry model is the quickest way to get MDM up and going. Under this scenario, companies don't build a separate master data store; instead they identify where the master data resides in the existing system and point to it. The downside of the model is that it creates an extra "jump" every time someone needs the master data, which may degrade system performance. "It's a tradeoff," he says.
Once discovery was complete, Donnelley started setting its strategy.
The company assembled an executive steering committee for budget approvals and to make sure the MDM effort was properly aligned with the business. In addition to IT, the steering committee included the client financial services (CFS), global strategic sourcing, marketing, finance and sales. It was a broad group of different perspectives and priorities, but it did reach a series of key decisions.
One decision was that "customer" would be the first data entity to master.
Why customer? Donnelley really wanted to focus on customers and the ability to identify opportunities to provide them with services across all of its product lines, Fahey says, and the existing process for analysis was time-consuming. The company felt it could be streamlined.
The steering committee also assigned the CFS group, which is responsible for creating new customers files, checking credit scores and defining credit limits, to oversee the effort because of its deep knowledge of the customer base and systems.
CFS went right to work establishing a data governance council to define and enforce MDM policies, set objectives and monitor data compliance. It also recruited data stewards, employees who would review questionable data and reconcile any discrepancies.
With the people and oversight in place, Donnelley started framing out how the system would work. It came up with this five-step process:
- Step 1: Create new customer. New customer records are created in one of the company's source systems.
- Step 2: Notify MDM. A notification is sent that a new customer record has been created or updated. The information flows into the MDM platform through an enterprise application interface layer.
- Step 3: Match and approve. Records are run through a data cleansing process that eliminates or corrects out-of-date, incomplete or incorrectly formatted data, then they're matched within the hub. If a new record for a Sears location is loaded, for example, it's bumped up against all the records in the hub. If the system discovers there's a match with a master by looking at the name, address and other elements, it will tie the information together. If a few major information points don't seem to be lining up correctly, the system moves the record into a queue for the data stewards to review and resolve. And if the information doesn't match up at all with files in the hub, the MDM system automatically creates a new master.
- Step 4: Enrich the data. Donnelley augments its data with hierarchy information (how a company's divisions are related to its parent, for example), industry classification and, in some cases, credit profile information. Much of this information comes from going outside and accessing a Dun & Bradstreet database that contains more than 130 million business records.
- Step 5: Update the reporting environment. The information is married with the company's enterprise sales data mart, which sits in Donnelley's data warehouse, and then is appropriately made available for research and analysis throughout the organization.
Once the architecture was defined, the company looked at various solutions and met with different providers to figure out whether to go with a commercial package or to put an MDM platform together itself.
The company, which felt it needed to ensure the platform would be responsive to its needs, made the decision to build its own based on a commercially available MDM framework (the name of which it didn't want published).
In the third quarter of 2006, Donnelley rolled out its Customer Master Data Store.
Fork in the Road
As Donnelley started to use its new MDM hub, it saw a need to improve data loading. The Customer Master Data Store was able to automatically take care of about 85 percent of the records that were sent in either by assigning records to masters or automatically creating a new master. That meant 15 percent of the information coming in had to go to data stewards for checking, resulting in 15,000 to 20,000 unresolved records after the initial load.
The problem, Fahey explains, was that the matching engine wasn't sophisticated enough to make more connections.
The company hired temporary help. They got through the records, but the temps didn't understand the business process well enough and wound up making decisions on records that didn't accurately reflect the relationship of customers to each other.
In addition to the data matching challenges, it was difficult to prioritize records, according to Fahey. Records of customers of varying sizes were simply queued up behind one another.
Over the next 18 months, the company rolled out a series of upgrades, but working out enhancements took time.
Meanwhile, Donnelley continued to make acquisitions, adding more customer names that needed to be mastered. "We hit a fork in the road," says Fahey. "Do we continue with our custom package or do we start looking at a package that has all the features and functionality [we] wanted, but was taking [us] a long time to build internally."
At that point, Donnelley began looking again at commercial MDM packages. One of the products Donnelley had checked out during its initial product evaluation was Purisma's Data Hub. Interestingly, D&B bought Purisma in November 2007. Since Donnelley forged a strong relationship with Dun & Bradstreet and because of D&B's alignment with Donnelley's technical requirements, Fahey says going with Purisma seemed like a natural thing to do and last year it brought the MDM software in-house.
Purisma's Data Hub is designed to reconcile fragmented data from multiple sources and create a base of complete and up-to-date information about an enterprise's customer, prospects and market opportunities, says Catherine Pedersen, product marketing director at D&B Purisma.
A key feature of Purisma is the tight integration with the D&B database that D&B built into the system. Another is its Continuous Learning Engine, which, through use, picks up the nuances of how customer information is handled. This facilitates future matching.
Donnelley was also impressed with the software's configurable user interface, which allows data stewards to customize their environment. The user interface "takes some getting used to," explains Donnelley's MDM architect Bob Short. But, he says, once data managers familiarize themselves with the interface, it does a good job.
Fahey cites additional benefits: a flexible hierarchy management, which enables multiple hierarchies to be created that in turn let sales managers see customer information from different angles; better workflow queuing, which allows for the prioritization of customer information by certain criteria; and the smarter matching engine.
While Purisma had the features and functionality Donnelley was looking for, implementation was not without some work.
The company had to load up the Purisma system, which meant records needed to be checked again before they became masters in the news system. Although she had to tell management that staff would be reworking data that had been mastered once before, Fahey had support. "It wasn't a difficult conversation to have," she says.
The Bottom Line
As far as Donnelley is concerned, the effort was worth it. With Purisma, the percentage of information needing manual review went from 15 percent to 5.5 percent.
"You're never going to have perfect data," says MDM architect Short. "But if you always have a process and can demonstrate that you have a process for constantly improving the data and helping your business users see how they can affect the data, it's a big win."
Fahey says it's difficult to quantify the return on its MDM efforts in hard dollars, but the company is seeing a lot of benefits. Now, when Donnelley acquires a company, it can quickly see a list of overlapping customers.
And the company, which offers a broad mix of products and services, can now tap into its sales data mart to find out what kind of business, and how much business, it has with a particular customer. Similarly, it can take a list of prospects and bump it up against the customer data hub to see what business the company is doing with a particular customer to determine additional sales possibilities.
These days, Donnelley is scoping out next steps for its customer master program, and it has plans to begin mastering vendor data next year.
"Our work has really just begun," says Fahey. "As users are starting to understand what MDM can do for them, they keep coming to us with more and more demands of how they can leverage our data or how we can move to different entities within the organization. So it's going to continue to grow."
Register or login for access to this item and much more
All Information Management content is archived after seven days.
Community members receive:
- All recent and archived articles
- Conference offers and updates
- A full menu of enewsletter options
- Web seminars, white papers, ebooks
Already have an account? Log In
Don't have an account? Register for Free Unlimited Access