Top-Down vs. Bottom-Up In Data Warehousing

Data warehouse systems have gained popularity as companies from the most varied industries realize how useful these systems can be. A large number of these organizations, however, lack the experience and skills required to meet the challenges involved in data warehousing projects. In particular, a lack of a methodological approach prevents data warehousing projects from being carried out successfully. Generally, methodological approaches are created by closely studying similar experiences and minimizing the risks for failure by basing new approaches on a constructive analysis of the mistakes made previously.

This chapter considers a few methodological approaches mentioned in the literature that describe how best to manage data warehouse lifecycles. This chapter also defines our methodological approach to a data mart project.

Top-Down vs. Bottom-Up

When you consider methodological approaches, their top-down structures or bottom-up structures play a basic role in creating a data warehouse. Both structures deeply affect the datawarehouse lifecycle.

If you use a top-down approach, you will have to analyze global business needs, plan how to develop a data warehouse, design it, and implement it as a whole. This procedure is promising: it will achieve excellent results because it is based on a global picture of the goal to achieve, and in principle it ensures consistent, well integrated data warehouses. However, a long story of failure with top-down approaches teaches that:

• high-cost estimates with long-term implementations discourage company managers from embarking on these kind of projects;
• analyzing and bringing together all relevant sources is a very difficult task, also because it is not very likely that they are all available and stable at the same time;
• it is extremely difficult to forecast the specific needs of every department involved in a project, which can result in the analysis process coming to a standstill;
• since no prototype is going to be delivered in the short term, users cannot check for this project to be useful, so they lose trust and interest in it.

In a bottom-up approach, data warehouses are incrementally built and several data marts are iteratively created. Each data mart is based on a set of facts that are linked to a specific company department and that can be interesting for a user subgroup (for example, data marts for inventories, marketing, and so on). If this approach is coupled with quick prototyping, the time and cost needed for implementation can be reduced so remarkably that company managers will notice how useful the project being carried out is. In this way, that project will still be of great interest.

The bottom-up approach turns out to be more cautious than the top-down one and it is almost universally accepted. Naturally the bottom-up approach is not risk-free, because it gets a partial picture of the whole field of application. We need to pay attention to the first data mart to be used as prototype to get the best results: this should play a very strategic role in a company. In fact, its role is so crucial that this data mart should be a reference point for the whole data warehouse. In this way, the following data marts can be easily added to the original one. Moreover, it is highly advisable that the selected data mart exploit consistent data already made available.