Data mining detects signs of Lou Gehrig's disease in gene carriers long before symptoms appear

This data mining methodology of "pattern arrays" in life sciences could be adapted for business analytics. I would recommend this methodology for large retailers to examine consumer behavior in brick and mortar onsite shopping. Also, it could be used in online advertisement.

Evaluating Online Advertising: Media Measurement Framework

There is an article by Stephen Pretorius published in Adotas that correctly identifies the Media Measurement Framework for evaluating online advertisement. He correctly identifies five areas of concern:

1. Track everything – The higher the number of variables that greater the probability that you can build a robust decision support system.


2. Integrate the data – You need a data warehouse for your variables, metrics, and decision support system that is separate from your operational data warehouse.


3. Compare apples with apples – Meaning comparison requires comparing the same type of metrics.


4. De-duplicate conversions – “You need a conditional filtering system (like DoubleClick’s Floodlight tag tool) to ensure that you only ping performance media tracking tags when they were the ones sending you the customer.”


5. Make your metrics relevant to your business – Invest in creating meaningful metrics for your business.
   

Additionally, Pretorius discusses the engagement measurement within the context of some potential improvements to Microsoft’s Engagement Mapping (EMAP) methodology: assumptions by advertisers without factual prove, and more data is better at prediction than better data mining algorithms. See, http://atomai.blogspot.com/2008/04/more-data-usually-beats-better.html

Stephen Pretorius article is called Why CFOs Don’t Believe in Online Advertising and it can be found at: http://www.adotas.com/2008/07/why-cfos-don’t-believe-in-online-advertising/