A/B testing - Better student experience without losing Revenue

A/B test performed on Udacity's student database to improve student experiment i.e. Reduce the number of students who cancel out of their first month free-trial without losing net-conversion(Revenue related metric).

The Process involved :

• Identifying and choosing invariants
• Identifying and choosing evaluation metrics.
• Calculating Expected , empirical standard deviation in the dataset.
• Choosing size.
• Choosing power.
• Duration vs Exposure trade-offs.
• Experiment Analysis :
• • Sanity checks
• • Result analysis
• Recommendation

Follow-up Experiment :

Decrease the number of students who cancel after they enroll in their nanodegree without reducing net-conversion. (Improve retention).

Full report in Udacity Report.pdf

Project summary and instructions : Instruction