After reading about Angoss – who we used in another lifetime – in their press release http://t.co/8eCheSND, we realized there are some [...]]]> We have been in the modeling, mining, predictive analytics business for years.  In fact, we helped pioneer some of the more advanced uses of SQL Server Analysis Services for Data Mining.

After reading about Angoss – who we used in another lifetime – in their press release http://t.co/8eCheSND, we realized there are some general misconceptions about “predictive analytics”.  First of all – it’s really hard work!  Getting the right data at the right time with enough cases to evaluate has always been the problem.  So where do you start?  Although it might seem rather obvious, the best place is at the beginning – and don’t be afraid to start small.

We recommend to our customers the following:

• If you have historical data, great, but make sure it’s complete.  A lot of customers want to answer questions (or make decisions about the future) with incomplete data.  This is a waste of time and energy
• Start today.  Amass as much detail as you can today so that it can be used later.
• Focus on the most critical decisions.  We often preach about decision support rather than predictive modeling because often times cost savings are easier to identify and will only help performance-oriented goals in the end.

Examples?  Check our projects category for those!

Scope:  300MM input names with a target population of 1MM/month

Goals and objectives:  Reduce the cost of ordering more than 2,000 lists per year to support customer campaigns