Data from Market Research to improve Predictive Analytics?

[Tomhca]

Does anyone have any good experience using secondary data or data from primary ad-hoc market research help improve prediction models of your real-time data?

This is something we have currently been working a lot on. Curious to hear your opinions/experiences?

Tom H. C. Anderson, Managing Partner
[b]Anderson Analytics, LLC[/b]
[url]http://www.andersonanalytics.com[/url]

[gsafarz]

Although not specific to real time models I have had some success in using primary market research data for predictive models. Since you cannot deploy a model using primary research data, unless you surveyed the entire population, the key is finding or creating a surrogate for anything you find is important. For example, you may ask in a survey if a customer views their relationship with your company as their primary relationship for a particular service. If you find this can be very useful in your prediction, you can either find a direct surrogate or create a predictive model as a surrogate. The power of this really depends on how good of a surrogate prediction you can come up with. In addition, this may give you insight into what type of data you need to collect and/or purchase in the future.

Hope this helps.

[fedec65]

I've build a platform that combine and ORACLE db with CLEMENTINE 10.0 and MrInterview (the online survey software of SPSS).

I use in real time data that come from our productive systems, from external sources (list brokers...), and from the online or CATI surveys.
That approach allows me to use the survey results to check the dm models and vice versa.