To Buy or Build a Datamart
The question in my last column was, "Is there a data mart in your future?" For most organizations, I concluded the answer is probably yes. There is an additional question, however, that needs to be addressed: "Do I buy or build a data mart?"
Many companies believe they have unique information needs and, therefore, should build their own decision support systems. Of course this kind of thinking has been challenged by the advent of packaged enterprise resource planning (ERP) applications. Many organizations, especially large, global companies with distributed operations, are choosing to implement packaged applications for key business functions such as accounting, manufacturing, shipping, sales and marketing and human resources.
There is a strong business case to be made for implementing packaged analytical applications, as well. Several categories of analytical applications have emerged over the past few years, and there are several reasons to believe this trend will continue into the future. Analytical applications deliver a lot of functionality out of the box; the cost of development is amortized by the vendor over a large customer base; and query time is significantly faster than it would be if you built a comparable application using a suite of tools.
Packaged analytical applications have been centered on several key business functions, such as managing and analyzing budgets, forecasts, costs, and financial statements. Lately, however, applications are focusing on the hot area of customer relationship management and marketing automation. Companies as diverse as Broadbase Information Systems, Influence Software Inc., Oracle Corp., Information Advantage Inc., Exchange Applications Inc., Paragren Technologies Inc., and E.piphany have solutions focused on improving the effectiveness of monitoring customer activities, establishing profitability profiles, enhancing customer loyalty and reducing resources consumed by unprofitable customers. Then there is the big kahuna of analytical applications: the ERP solution. SAP’s Business Intelligence Warehouse, for example, is a decision support system tightly coupled with an R/3 product.
You should look for three key features from a packaged analytical application. The first is a predefined, domain-specific data model. Domain expertise is essential to developing a database design optimized for analyzing customer data, manufacturing data, or human resources data, for example. The second key feature is a complete set of predefined queries and reports that are aware of the data model and enable end users to begin obtaining useful information from the data model once it has been populated. The third feature is a preconfigured extraction and transformation model for obtaining, cleansing, formatting and configuring data to load it into the analytical database. The extraction logic should be able to access data in the databases of the popular application vendors in your market sector.
Finally, assess the abilities of the package for customization and tailoring, which may be necessary to suit your company’s specific requirements. There may be a concern over how to integrate the decision support solution with the rest of your company’s information architecture. In my last column I discussed the three types of data mart architectures: independent, dependent and federated. A data mart from a software vendor will be an independent data mart, since the vendor will not have the opportunity to define either a dependent architecture, which requires a centralized data warehouse, or a federated architecture, which uses a common metadata repository and common database keys.
You may, instead, elect to build an analytical data mart rather than purchase one if no satisfactory solution is available on the market. For example, while there are many analytical applications for customer relationship management and financial management, there are relatively few offerings available for manufacturing process quality control. A manufacturing company, therefore, would probably need to build a customized application.
The good news is that the data mart vendors have made significant progress in delivering integrated tool sets that reduce the work required to do the building yourself. They provide capabilities in the three areas of data sourcing, designing the core database, and desktop queries. Many vendors deliver tool sets in all three of these critical areas with varying levels of integration and interoperability. --Robert Craig is vice president of marketing at Web Engine Inc. (Burlington, Mass.), and a former director at the Hurwitz Group Inc. Contact him at rcraig@webengine-db.com or via the Web at www.webengine-db.com.