Critical. Authoritative. Strategic.


CBR is proudly produced & published
by Technews
Issue Date: March 2007

Data master class

March 2007

Master data management, or MDM, is touted as the silver bullet for ridding companies of inconsistent and untrustworthy business data. But does its real value lie in being an enabler of SOA? Madan Sheina investigates.

A 'single version of the truth' or 'a 360° view of the customer' might be familiar, even hackneyed, phrases in corporate IT circles, yet both remain elusive goals for many enterprises today.
Getting to that 'truth' requires a basic constant - the consistent use of key data shared across the organisation. But this simple goal is difficult to accomplish; companies have tens, if not hundreds, of applications and databases, each of which has a different definition of customers, products, suppliers and more.
It also costs companies millions of dollars through delayed ERP rollouts and lost production, inefficient procurement and supply chain planning and poor customer service - not to mention the impact on the brand.
Consider the case of a major oil company that drilled a borehole into the side of an existing well because two systems had swapped x and y coordinates. It cost $5m to rectify. Or two subsidiaries discovering they were bidding against one another for the same contract because they each had different coding schemes for the same customer.
Companies have not been ignorant of the issue. Many have implemented ERP and CRM systems and enterprise data warehouses to deliver a single consolidated repository. But ironing out inconsistencies in these siloed systems is proving to be a complex and expensive endeavour. Clearly the task of managing data is far too strategic to be handed off to a single ERP application or data warehouse. What is still lacking is an enterprise-wide approach that encompasses all corporate data, both operational and analytic.
Master data management (MDM) is a new technology and discipline that promises to do just that, by helping both IT and business to rationalise overlapping and conflicting data in disparate systems to drive data consistency and quality across functional business units.
Technologically an MDM system references a single, authoritative source, or 'master', that feeds into front-end applications as required. It also identifies and manages various interrelationships of the data and their derivations, creating business rules so that master data can be applied across the enterprise. MDM also defines organisational processes for how master data definitions are mediated and shared across different constituencies, including business divisions and partners.
MDM is not new by any means. So why do proponents think it will fare better than past approaches? There are several reasons. First, MDM works by establishing master data rather than by focusing on operational (transactional) data. Second, MDM is an active approach to managing the entire information life-cycle, allowing companies to define new data, monitor exceptions as the data changes and rationalise and synchronise data as it is updated.
Last, and perhaps most promising of all, is that MDM takes a services-based approach that is in line with modern service-oriented architecture (SOA) initiatives. In other words, it treats data as another enterprise service in an SOA. Some refer to it as the 'fourth layer' of an IT architecture that sits between data and the business logic and presentation layers.
Dealing with silos
Master data management has gained prominence in recent years because of a highly fragmented application landscape caused by lines of business applications optimised for specific functions or departments. Having these multiple ERP and CRM instances and home-grown IT systems creates data silos that make it difficult for a business to get a single authoritative source against which all other data is compared, or, in MDM parlance, a 'golden record'.
"The problem stems from the fact that companies do not have all their business data in one place," notes Cliff Longman, chief technology officer at data warehousing and MDM firm, Kalido.
“The problem stems from the fact that companies do not have all their business data in one place. It is created and stored in multiple applications and databases, each of which engenders its own account of information.” Cliff Longman, Kalido CTO
“The problem stems from the fact that companies do not have all their business data in one place. It is created and stored in multiple applications and databases, each of which engenders its own account of information.” Cliff Longman, Kalido CTO
"It is created and stored in multiple applications and databases, each of which engenders its own account of information that is usually tied to transactions or, in the case of data warehouses, analytics. When you are dealing with 17 instances of SAP it can be a headache to find out that product data in one SAP system refers to another set of product data in another SAP system."
It is precisely these multiple versions of data that create multiple versions of the truth for business, which can be a headache in businesses such as retail banking, insurance, healthcare and high-SKU retailers managing tens of thousands of SKUs.
Arvind Parthasarati, senior director of solutions at data integration firm Informatica, notes that the problem becomes more pronounced when siloed data is particularly volatile and subject to frequent change, communicated over phone, fax or email: "It is a big problem in dynamic environments where the churn on master data is much higher," says Parthasarati. "If you sre changing your distributors every week, or experiencing churn every week, you can no longer rely on manual stuff."
Longman ultimately blames a lack of strategic thinking about data as one of the root causes of the problem. "Companies did not treat data as an asset," he says. "There was no holistic discipline or training for managing it like financial assets." For instance, financial accounts software supports a consolidation process for a unified chart of accounts, mapped to multiple financial systems for integrated reporting and performance management. So why are other types of business data not treated with the same level of respect?
The drive to get business advantage from the integration of multiple business processes and the disparate applications that serve them has also pushed master data issues into the spotlight. Anurag Wadehra, CEO of customer data integration specialist Siperian, notes there is a pressing requirement for business processes to span former silos of applications. "When applications did not talk to each other, MDM was not an issue. But now companies expect them to, and are looking to break down these application silos with new architectures such as SOA," he says.
“You can enable all the SOA you want, but if it is not done with MDM then it does not add much value.” Anurag Wadhera, Siperian CEO
“You can enable all the SOA you want, but if it is not done with MDM then it does not add much value.” Anurag Wadhera, Siperian CEO
Paraic Sweeney, vice-president of product information management solutions at IBM, concurs. "Traditionally the inventory of apps that most enterprises have built up have focused on specific processes and verticals," he notes. "But fast-forward to what companies have been trying to tackle for the past two or three years and it is all about cross-functional process optimisation. Things such as sharing customers across multiple brands after an acquisition, or supplier optimisation through electronic information exchange."
Changing business dynamics, such as the growth of outsourcing, are also driving this integrated process view. "Business can now easily shift manufacturing from Japan to China, but their IT systems struggle with that change," explains Parthasarathi.
Sweeney admits it is a difficult task, as most enterprise data elements are moving and changing. "It is not just about connecting your web channel to your call centre. Integrating the flow of data is not enough. The key is getting an integrated view of that as it is being updated or changed."
SOA enabler
Clearly, having a shared and flexible IT architecture that works with consistent, accurate and up-to-date master data is vital. For that reason many point to MDM as a seminal development for IT setting the goal of moving towards SOA.
"You can enable all the SOA you want. But if it is not done with MDM then it does not add much value," Wadhera asserts. "SOA is about connectivity and integration. It is one thing to create common interfaces, but if you start pumping inconsistent data definitions between them then there is a problem." He claims the 'dirty secret' of many failed data warehousing and CRM projects can be traced back to a lack of common data definitions.
Parthasarathi agrees there is a symbiotic relationship between MDM and SOA. "The MDM pill is harder to swallow without SOA," he says. "And without MDM it is hard to do SOA. Getting a single view of the customer is the number one thing that everyone wants through an SOA. It often becomes the first bell-wether of your SOA."
The MDM approach seems to sit nicely with how SOAs are supposed to work – ie, by decoupling master data from business applications. The real promise of SOA is that you do not need to know in advance which applications will be talking to each other. But with that paradigm it is critical to first agree on common data definitions.
Longman argues that one cannot have SOA without having an enterprise data service. "MDM inherently creates that data service component within an enterprise SOA, actively synchronising clean and consistent data to applications as an on-demand service," he says.
Sweeney argues that generating an aligned perspective in SOA requires the master data to be decoupled from business applications, leaving the latter to worry only about business logic. "The whole idea of SOA is identify a core set of services that can be used to construct changed and new business processes that are embodied in existing applications, rather than building them from scratch," he says. "The applications themselves should not have to worry about the semantics of the information to access it."
Sweeney adds: "Decoupling data from applications frees up information from departmental silos so it can be shared with other systems. It also shields data from changes in applications and business processes."
Sunil Gupta, director of solutions marketing at business applications firm SAP, certainly sees MDM as a fundamental piece of SAP's data foundation, in particular its SOA-enabled composite XApps strategy. "NetWeaver MDM lets us effectively ensure a consistent source of truth in all our composite applications by embedding it into our enterprise services architecture," she argues.
Need for governance
Putting in place a robust, SOA-based data architecture is, however, only half the story for MDM. It is no coincidence that successful MDM deployments are always backed up by strong data governance polices and processes in the organisation. Hence MDM is as much about operational process alignment as it is about data consistency.
Most MDM vendors acknowledge their software can only work if there is a strong culture of information sharing. The key, according to Longman, is to establish a data governance programme that addresses MDM from an enterprise-wide perspective in order to identify and define data elements, hierarchies and taxonomies, specify policies and rules for how master data is maintained, and explicitly assign data stewardship roles and responsibilities to mediate over definitions and resolve potential conflicts over data ownership.
But old habits of data ownership die hard, and require considerable negotiating prowess with various stakeholders and constituencies in the organisation. Negotiating a consensus on the definition and dissemination of master data across the line of businesses is not a task for IT, Longman notes: "It is one that needs to be controlled by subject matter experts tuned into the business."
Longman believes it is imperative to assign people data stewardship roles to take responsibility for maintaining master data and provide them with workflow tools to manage the process of moving, validating or enhancing data.
"It is not about data ownership anymore," he says. "Companies must change their mindset to maintain processes that create and manage the lifecycle of master data, rather than the data itself, which often extends outside of a single department."
Sweeney claims more of these coordination and negotiation functions are starting to appear in companies. "We are starting to see new job titles such as data governance director and master data architect to ensure coordination and alignment across various facets of the business," he says.
However, he warns that MDM is not a one-off effort. "Data governance needs to be continual and sustainable," he says. "It is an ongoing journey and you have to start small, have a big vision and expect it to take longer than you thought."
CBR opinion
MDM, as a services-based approach for delivering a clean and consistent set of business data to business applications, has emerged at the right time to meet the demands of SOA initiatives. Companies that are serious about SOA will have to deal with master data architecture – otherwise their efforts will be largely wasted. But MDM software can only go so far. Its implementation needs to be accompanied by a strong data governance programme. Companies should not underestimate the challenge of negotiating a consensus of data definitions across different organisational hierarchies and divisions.

Others who read this also read these articles

Others who read this also read these regulars

Search Site

Search Directory

  • Search for:


Previous Issues