And just to circle your comment back, I think I have seen more “wasted analytics” using data mining than any other approach, because folks simply don’t know what it is good for and what it is not good for…

A clear goal for each and every analysis must be articulated before any data is prepped. Actionable outcomes (what do we implement as a result of the analysis) must be laid out, budgetted and agreed-upon by the pertinent decision-makers. Then, after the work is done, results that show how the program improved (hopefully) as a result of the analysis must be shared–communicated within the organization.

If all of the above are done the right way, you’re laying the groundwork for future success. Job security is another side benefit :)

Data-oriented technology (data mining, web analytics, CRM, etc.) is often sold as a “cure” when it simply provides a path. The need for marketers (or consultants) to understand both what the technology can do and how to use the output is very real.

This problem is particularly acute when the Marketer has a background in general advertising; they have trouble moving from conceptualizing everything in terms of “who” (demographics) as opposed to “what” (behavior). The creating of marketing solutions is very different for each.

And too often the analyst acts as an “enabler” in this regard…just because something can be analyzed or modeled doesn’t mean the output will be helpful or actionable. Each side needs to push back on the other.

Problem there is, you can’t push back if you lack the background to do so.

Hmmm…maybe I should write a book on solving this problem.

Oh yea, I did!

You have rightly pointed what data mining can do and what it cant do. But I find something missing in your post and that is, what exactly marketers neeed to do and what an analyst need to.

I think, the need of the hour is a very strong consulting who is aware of both the analytics and the marketing. Analytic vendors (in house analytics team) need to sale the strategy based on analytic results (analysis + models) and not the raw results to the marketers. Then half the problem is solved.

I suggest good Customer Profiling after segmentation and Decision Table after Modeling (models either one or more).

What do you think?