Sunday, June 24, 2007

Prototyping with a spreadsheet

Question: Why is prototyping a good idea?
Answer: It allows the users to see something in a short amount of time. The model built can often be used in the final implementation so the work is not duplicated or wasted.
Question: What will the user see if the data is empty? Answer: They will see no results. The user may also receive ODBC errors if a source is defined in the repository, but not physically available.

Prototyping accelerates the entire building process to gain user buy-in and executives support for the project.
Anyone can build something quickly. Often the initial prototype becomes the final model with little effort. It is not a throw away effort. This approach allows you to build a prototype in the real application rather than building it in excel or some other technology that will not be
used in production. I keep versions of Access databases that use in prototyping specific subject areas of an Oracle Analytics Application. I just change the database and connection pools as well as to perform some minor changes in the webcatalog so I can show dashboard of the subject area that my client is interested in as an initial implementation of much larger in scope of an Oracle BI solution.

Oracle BI EE/Siebel Analytics strength is the ability to access non-relational structures and create logical dimensional structures. This is unique functionality that is seamlessly integrated within the platform. Other products require you to use different user interfaces for different types of analysis and back-end systems.

The difference is in not creating any physical joins. If you think about it, it makes sense since the data is flattened into a single wide row of information, so no join is required.

If the physical source changes, the logical model is still correct and doesn't need adjustment.

Create the logical Joins in the Business Model layer.

Notice the dimensions are combined here for ease of use.

Additional notes when prototyping with a spreadsheet:

Spreadsheet does not support count distinct. Be sure this feature is
turned off in database features table.

Disable connection pooling. That will allow you to open up the spreadsheet and make changes to the data without shutting down the Analytics Server.

No comments: