Article Preview
Top1. Introduction
When it comes to the failure rates of IT projects, estimates differ widely, depending on the understanding of what constitutes a failure. There are number of measurements that can be used to assess the success or failure of a large IT project. The success measurements can be divided into two categories (Nelson, 2005). One category encompasses process-based measures: project on schedule, project on budget, and project meeting requirements. The other category includes outcome-based measures: is the project result actually used, does the project result provide value for the organization, and does the project result enable learning that helps prepare the organization for the future. In addition to the fact that some IT projects are complete failures and some are definite successes, other concepts, such as “failed success” and “successful failure” have been recognized (Nelson, 2005). A failed success is defined as successful from the process perspective, but a failure from the outcome perspective. On the other hand, a successful failure fails on one or more process-based measures, but ultimately delivers solutions that succeed from the outcomes perspective.
Data warehousing has become a standard practice for many companies worldwide (Jukic, 2006). Within the past decade data warehousing projects have been receiving a growing amount of attention and resources in the majority of large and mid-size organizations. A recent study reports typical cost for creating a one terabyte data warehouse of several million USD with a typical implementation time of several years (Gray, 2006). There are no definitive numbers on the failure rate of data warehousing projects, but estimates vary from as little as 20% to as high as 90% (Watson et al., 1999; Frolick & Lindsey, 2003; Watson, 2005; Hwang & Xu, 2007). As is the case with estimating the failure rates of all IT projects, one of the reasons for the wide discrepancy in estimated failure rates of data warehousing projects is the absence of an agreement of what constitutes a failure. For example, there is no unambiguous answer to the question: does abandoning the initial design, scope, strategy, infrastructure or technology of a data warehouse design constitutes a failure? In some cases the answer is a definitive yes. On the other hand, in cases where such abandonment is accompanied by learning lessons that allow for the adoption of successful alternatives which eventually result in a properly designed and used data warehouse, it is appropriate to exempt such cases from a label of outright failure, given that the outcome is in line with the concept of a successful failure.
Although it is hard to establish a precise overall failure rate of data warehousing projects, the fact remains that some data warehousing projects (just like any other IT projects) fail. The literature indicates that there are many, often intertwined, factors that can cause data warehouse project failure, such as budget overruns, unacceptable performance, poor quality data, weak sponsorship, and lack of long-term planning, etc (Stackowiak, 1997; Adelman & Moss, 2000; Goldman, 2001; Frolick & Lindsey, 2003; Hayen, Rutashobya, & Vetter, 2007).
As do most information system development processes, data warehousing projects follow some form of a System Development Life Cycle (SDLC). SDLC is the overall process of developing information systems through a multi-step process including steps such as planning, analysis, design and implementation (Dennis, Wixom, & Roth, 2006). One popular data warehouse-focused variation of the SDLC is the Data Warehousing Lifecycle (Kimball, Ross, Thornthwaite, Mundy, & Becker, 2007) illustrated in Figure 1. Certain steps (such as product selection, project initiation, etc.) are omitted for brevity. The depicted steps are common to any data warehousing project:
Figure 1. Abbreviated data warehouse system development lifecycle