A takehome exercise for predicting user engagement.
This repository contains the data in two csv's, an ipyNotebook detailing my analysis, and a recommendation report. It also includes the initial instructions/details of the challenge as a pdf.
In order to perform EDA on this dataset, I had to create a function that allows you to create labels for defining an 'adopted_user'.
Interesting findings
- Adopted Users accounted for ~12% of the dataset
- Source Creation - Adopted Users were mostly invited by Org_Invite and Guest_Invite
- Marketing Drip and Mailing List Opt didn't show anything to interesting
- The second highest adoption feature was the month of May
- The highest correlated feature were users invited to personal projects
Most Interesting Finding:
- The most interesting finding is that the least amount of registration occurred on May for adopted users, but it's the most registered in non-adopted users.
Further Investigation: It would be great if we can have more data about user activity. With the features given it was difficult to find any correlation with adoptive users.
Defining an "adopted user" as a user who has logged into the product on three seperate days in at least one seven-day period, identify which factors predict future user adoption.
Please send us:
- a brief writeup of your findings (the more conscise, thebetter -- no more than one page)
- summary tables
- graphs
- code
- queries that can help us understand your approach.
Please note any factors you considered or investigation you did, even if they did not pan out. Feel free to identify any further research or data you think would be valuable.