Dave Cameron
Nov 4 · 1 min read

I’m noticing this a little late. This is a very good read; thanks for putting it out there. I’ve been in the business world over 30 years; and worked in consumer marketing throughout. I’ve been aware of machine learning for 20+ years — so agree it is NOT new. I also agree that your use cases are spot on. In marketing, if you want consumer understanding — use statistics. If you want to churn through data quickly for display ads — use machine learning. I would like you do be aware of one misconception. In my last 20 years as an applied statistician; I always had a holdout sample and a modeling sample — so as to validate the model on a holdout set. Machine learning does the same thing with different names. Your writeup says (paraphrased) that statistics is unconcerned with predictive power. Not in my world; the applied statistics models had to be predictive as well. Sorry for the delay in response; this article was recently linked to one of my feeds.