By Elliot King, Ph.D.
The buy-versus-build question has haunted the development of data warehouses and analytic applications since Bill Inmon started discussing the underlying concepts in the 1970s, and Barry Devlin and Paul Murphy published “An Architecture For A Business and Information System” in the IBM System Journal in 1988. Since then, many vendors have offered packaged applications that facilitate the development of data warehouses and analytic applications. Nevertheless, many companies still opt to build their own custom data warehouses rather than buying pre-packaged solutions. The question is, why?
According to recent studies, that question can have several answers. For years, vendors of pre-packaged data warehouses and analytic applications simply did not deliver on their promises. One approach vendors took was to build analytic applications based on mapping pre-built metadata into operational systems. That method failed because those applications drew on too narrow a data set to be useful, and they often impeded the performance of the operational system. Alternatively, some companies with pre-built applications did try to create data marts and data warehouses prior to performing analytics. In those cases, the ETL process often was the weak link that led to failure.
The second reason companies have opted to build rather than buy is cultural. Most companies think their business situation is more unique and cannot be addressed by “pre-packaged” applications. However, this belief is generally unfounded. For any given subject area, organizations can find packaged applications that cover 60, 70, 80 percent or more of the data elements they require “out of the box.” The questions that organizations want to address (Who are our most profitable customers? What is the history of our AP aging over the past few years? Which vendors have the best track record for on-time delivery?) happen to be common for most businesses.
In my opinion, the traditional objections to investing in pre-built data warehouses and analytic applications no longer apply. Pre-built technology has improved considerably and companies are slowly coming to the realization that most of their data is just not that special. Get over it – your enterprise just isn’t that unique!