Lesson #3: Data Warehouse as The Analytic Foundation
In many of the web companies that I have talked to, I'd astounded by the lack of appreciate for good data warehousing techniques. Many of these folks believe that the infinite amounts of processing power obviates the need for good data warehousing design and fundamentals. Let's review why this is a dangerous assumption, and let's learn how to leverage the 20 years of data warehousing training, techniques, methodology, and tools that we have developed and learned. Let's treat data warehousing with the respect that it deserves!We have learned over the past twenty years that a well-designed data warehouse is still the best vehicle for aligning disparate data sources. We’ve got years of experience in building agile and extensible data warehouses. There are scores of books (from leading thinkers such as Ralph Kimball, Neil Raden, Claudia Imhoff, and Howard Dresner) and several outstanding education opportunities (at places like The Data Warehouse Institute and Kimball University). And one of the HUGE benefits of a well-designed data warehouse (read that to mean one that uses conformed dimensions and the data warehouse business architecture) is extensibility.
Data Warehouse Benefit: Extensibility
Extensibility is critical in an environment in which most web companies compete. There are new marketing tactics (rich media, smart ads, video, mobile, IPTV) being developed every day. And the analysis of the effectiveness of these new marketing tactics requires an extensible data warehouse foundation. For example, the correction attribution of multiple marketing tactics on a customer event requires the integration of data from multiple operational or serving systems.
Unfortunately, many in the web world don't think data warehouse design is all that important. Why does one need to spend time on data warehouse design when you can store ALL the data at the lowest level of granularity (the benefits of infinite amounts of storage) and have unlimited computational capacity to apply against that data (courtesy of cloud computing)?
Since we are running out of space in this issue of Decision Support 2.0, let me share in the next issue a couple of examples that can not be easily solved by infinite amounts of storage and unlimited computational power.
Until next time, Decision Support 2.0
