In a world in which data is flooding into organizations from all directions, it has long been axiomatic that companies should try to develop a "single version of the truth." The idea is that a customer profile, for example, should look the same to everybody from accounting to sales to customer service.
The impetus to build a single version of the truth comes from many sources. But perhaps the most important is the commonsense notion that companies cannot truly understand their customers, or their performance, or market trends or their operations, unless all the data associated with a given operation is available to all the interested stakeholders. To use even more jargon, data silos must be broken and data integrated across the enterprise.
The classic examples demonstrating the need for a single version of the truth include these: Customers who look great to the sales team, but look terrible to accounts receivable because they are habitually slow payers. Or a supplier is preferred by the logistics group because of its ability to deliver goods just in time, but is an anathema to quality control because of quality issues. Issues like these can be identified before the next sale or purchase is made if everyone in the organization has access to a single version of the truth for a customer or supplier.
The challenge to creating a single version of the truth is two-fold. First, completely integrating all the data about an entity such as a customer, product or even many operational processes is probably beyond the reach of most master data management programs. Sure transactional data can be combined with external data from third parties--think credit reports here--and perhaps even semi-structured data like email. But it isn't easy. And when you add in the local knowledge people have captured in their personal office applications, the task becomes even more daunting.
The second challenge is conceptual. Using the term a "single version of the truth" implies that you can arrive at that destination. But business activities are ongoing. Customers can be great customers until they aren't anymore. Instead of striving to create a single version of the truth, organizations would be better served by building data-robust entities. We can never actually arrive at a single version of the truth. We only can know what we know now and need processes to integrate what we will learn tomorrow.