Contribute Media
A thank you to everyone who makes this possible: Read More

Predicting the past


In this talk I would like to present basic techniques in counter-factual predictions that are used for uplift analysis in marketing. The idea is to use time series prediction models to predict the past after certain intervention date (our marketing action) and compare what would have happened to what actually happened. With some basic black magic trickery (statistics) we can then draw conclusions about effectiveness of the marketing intervention. This basic talk would cover using simple off the shelf tools (fbprophet) and ML techniques.


Improve this page