Share what you know with millions of people

Analytics projects lends themselves very well to a phased implementation. Unfortunately, a rather common failing in these types of projects is to try to do too much all at once. To be successful, you have to be able to implement quickly and frequently using agile development techniques.

You start by identifying the core business processes that you want to model, and then prioritize them based on value to the business. You start at the process level because that's the smallest sized subject-area that yields actual value within a resonable amount of time. Anything smaller, and you will have an incomplete picture of the process being modeled; anything bigger and you'll likely spend so much time doing requirements analysis that the requirements start changing on you before you can get anything implemented (assuming that you ever do).

Once you've identified a business process to be modeled, start your requirements gathering with the people who actively work within the business process you have chosen. How do these people want to use their data? What questions are they trying to answer with their data? This is referred to as the "top-down" approach. The "bottom-up" approach focuses on the data itself rather than gathering good business requirements up front. Both approaches can work, but the top-down approach has a greater chance for success because it is driven by high-value business requirements rather than the if-you-build-it-they-will-come mentality associated with the bottom-up approach.

This is a high-level discussion to be sure, but as you do more research, you'll see that these steps also happen to be the first steps in any agile development methodology. Using a top-down, agile development process, you'll find that your success rate is high, with lower overall risk, within a fairly short time-frame. What more could you hope for?

Good luck to you.

David