I'm at the TDWI Boston user group meeting today. Our first presentation is on predictive analytics and is titled "From FICO Score to Risk Score – A Journey Into Predictive Analytics." The speaker is Jens Meyer, First Marblehead.
This is a very interesting presentation explaining the predictive analytics process, what kind of skills are necessary and how people in this space need to continually learn and adopt to changing business, economic and political forces. Using First Marblehead's journey in the student loan business pre- and post- 2008 US financial crisis is a fascinating case study.
Some of the key takeaways:
- Predictive analytics is much more than simple forecasting
- Many statistical models need to be applied to the data with much trial and error along with continual refinement
- You need lots of data from many sources that is often incomplete and not clean
- Building models is a very labor and time intensive process, especially if the models are trying to predict consumer behaviour in a very dynamic market conditions
- The people building the predictive models have a statistical, math or economic background and understand their firm's business and industry models
It is great to hear someone in the field presenting how they have used predictive analytics in their businsss and how critical it is for their company.