The majority of Data Warehousing initiatives fail on the first attempt.
In almost every case, that failure can be traced to a tendency to put
showcasing the technology ahead of
solving business problems by supporting decisions.
Business Intelligence initiatives are fundamentally different from traditional IT development projects that go through design, development, testing, and deployment in sequence. In fact, attempts to build enterprise data warehouses in this fashion have nearly always been abandoned or cut short well before their completion. The reason is that these applications are meant to support the process of business decision making across the entirety of the enterprise. This means that any significant business event can change the design assumptions, as well as the desired solution. Therefore, any enterprise-level design risks becoming obsolete, at least in part, in a matter of months, if not weeks. There are countless examples of enterprise data models, representing many person-years of effort, rendered instantly obsolete by a merger, acquisition, outsourcing decision, or change in business strategy.
One response to this phenomenon has been to abandon the notion of enterprise BI in favor of packaged point-solutions that promise rapid implementation and address a specific set of decisions within one subject area like finance, marketing, or human resources. These occasionally take the form of complete analytic applications like CRM, financial planning, or balanced scorecards, each of which require its own dedicated data repository, or ‘data mart’. Such solutions may end up providing value to their own sponsors, in aggregate, however, they tend to create a redundant web of technologies, data flows, and conflicting versions of reality that destroy credibility. Such arrangements quickly become far too expensive and burdensome to maintain.
Other unique aspects of decision support applications are that they place a premium on timeliness, flexibility, and ease of use. Moreover, many user requirements are not even anticipated, much less defined, until the tools are actually put in the hands of the users. This makes it necessary to adopt an iterative approach to development that emphasizes the use of prototypes and proofs of concept.
To those firms who choose to get out in front of these issues, we advocate an iterative methodology that views enterprise BI as a set of discrete releases by subject area with limited scope and timeframe. Six months is generally a good limit to impose on any single project. These would include the implementation of an analytic application like CRM or an ERP application delivered with integrated analytics. The key is to define what these individual applications must share in the way of data flows and definitions, technology standards, administration facilities, and other resources to work together efficiently, while delivering information that is as consistent as it is credible.
guides its clients as they navigate through the complex and rapidly consolidating Business Intelligence tools marketplace. Our extensive experience with the leading products puts us in a position to separate the hype from the value.
knows the heart of any successful data warehouse is a sound data model. We also know that each model must be customized to the leverage the strengths and weaknesses of the database and reporting products being used for that application. Our specialization in relational database technology enables us to develop data models that are not only flexible and efficient, but specific to the client’s business Intelligence platform.