28 January 2006

When all you have is a cube...

Multi-dimensional analysis tools, or cubes can be very useful tools for data analysis and the marketing behind them would suggest that they are the silver bullet of business intelligence tools. While I agree that cubes are great for a small segment of information consumers a lot of the time and are possibly good for a larger group of consumers some of the time but they are way too complicated for most users, most of the time.

One of the oldest tricks in the book is to make products that in order for a person to make use of the product, they have to invest in it. WordPerfect used that years ago as did Lotus and others. Who remembers how to do the contortion required to do a table of contents entry in WordPerfect? How about changing fonts. Do you remember how many levels deep into the Lotus 123 menu you had to go to change a chart series?

I don\'t, but I do remember it was painful and the only way to get good at it was to train your nervous system to memorize the motions (which was a considerable effort and investment in time) before you could even begin to be productive.

Once you had made that investment, how eager were you to switch to Word? Many people who held out for a long time because they had made such an investment into WP that they couldn't consider switching. Add the investment in macros and printer configurations and it was very compelling to just stick with the tried and true.

Cubes are set up in a similar way. They are sufficiently difficult to learn and become proficient at (similar to user learning the keystrokes) and they require enormous investment in ETL and building to automate the production (similar to macros and configs) that once you have made the investment it is difficult, and very expensive, to get out.

Cube software makers, and the consultants that promote the software, would have you believe that all of your business intelligence and information products are best served up in cube format, with their proprietary web front end of course. They are promoted as a single solution to a host of problems. All you need is cubes. If you were to follow that advice you would look at everything as a cube problem when in fact a simple report or chart would not only suffice but is simpler, faster and requires far less sophisticated technology to make it happen. Cubes are a useful tool in the BI arsenal at best and an expensive source of automated confusion at worst.

Cubes and Multi-dimensional analysis tools are a great addition to a larger strategy of information tools including data warehouses, data marts, reports, charts, portals, forums, etc.. Promoting them as the silver bullet for business intelligence does a disservice both to organizations and the cube products. Positioning them as power tools for power users which make a wide variety of analysis data available and useable is much smarter.

No comments: