In our search for techniques for Rapid BI, one idea is to Fail Fast. What I mean is to take some of the risky bits of a project and do some fast development and test out approaches. We learn much more from such a trial that attempting to design the risk out of the project.
The difficulty with this approach I think is that when we come up with a design and approach we find it difficult to abandon the approach and go back to the drawing board. Generally our approach is to attempt to make the existing approach work by debugging. Generally I find I get so committed to one approach I cannot see beyond the detailed problems. I have always admired the people who can go back to the problem and come up with a new approach. This mental flexibility is a great gift and one we should all cultivate.
I am recalling a great example of a project that would have been quite different if it had failed fast. A system was developed using sophisticated modeling to assign gates at a large airport. The system would not work and after much work somebody realized that if they assigned planes instead of gates the problem would be resolved. The new primary key solved the design problem. Maybe if they had not been so attached to one solution by failing fast, the problem would have been resolved much faster.
An example of failing fast was a prototype we developed for a financial services client who wanted to offer all their services in a branch through whoever was facing the client. This project would have quite revolutionary for them at the time and a very expensive project. We suggest we could prototype all the data very inexpensively for one branch and try it out. The technology part worked just fine but the people in the branch rebelled. Each person responsiblefor a product did not want the other people in the branch seeing the data and discussing it with the client. The financial services firm had major organizational issues to deal with before introducing the system. We saved that client millions of dollars.
I realize that Fail Fast is used in many contexts. I would like to hear how others have benefited from failing fast.