Community Maintenance Support

Some analytics to investigate community activity within the AgileVentures Slack community (and associated systems)

How to run tests

python run_tests.py from the command line at the root of the project should suffice

How to train and predict using basic RandomForest model

You first need the slack archive unzipped into av subfolder. Inside that av subfolder, you also need to place a users.json file FROM the users.list api. You also need a stripe_customers.csv with stripe signup data and a PayPalPayments.csv, and you also need in that same folder av_members.csv indicating current premium members.

1. av/av_members.csv <-- current premium members (from google drive spreadsheet + extra)
2. av/PayPalPayments.CSV <-- PayPal payments from AV members (from PayPal) for date of signup
3. av/stripe_customers.csv <-- Stripe payments from AV members (from Stripe) for date of signup
4. av/unresolved_emails.csv <-- hand created to match people up
5. Agile Ventures Slack export Sep 13 2017.zip <-- from https://agileventures.slack.com/services/export, e.g. unzip "Agile Ventures Slack export Sep 13 2017.zip"
6. grab users.json from https://api.slack.com/methods/users.list/test

Then run python data_dump.py to export the historic user signup and user activity data to data.csv, which should be git ignored.

Also, run, python export.py to export the current week's data to to_be_predicted.csv. The script as currently constituted will export data for the week ending on July 8th 2017 and the preceding 3 weeks of activity. To change this to accomodate future archive dumps and dates, change the dates as appropriate in export.py

Now with both data.csv and to_be_predicted.csv in the main folder, open up R or R studio. Set the current working directory (Session --> Set Working Directory) to the main folder for CommunityMaintenanceSupport and run the code contained in random_forest.R (note, all code needs to be selected in order to run, and the following packages need to be installed:

install.packages('caTools')
install.packages('DMwR')
install.packages('ROCR')
install.packages('randomForest')

). You should get some output predicting those likely to signup.

Objective

How to identify community members who are contributing more and provide them with additional support.

As an AgileVentures admin
So that I can ensure the AV revenue stream
I would like to know which members are likely to be downgrading in order to offer them additional help and support

As an AgileVentures admin
So that I can ensure the AV revenue stream
I would like to know which members to approach to suggest premium upgrades to

As an AgileVentures admin
So that I can ensure we have a healthy community with lots of inspiring projects with great contributors
I would like to know which members to offer additional help and support to

Completed so Far

• Analysis of Slack activity in the three weeks leading up to Premium signup -> saw a possible trend of increasing activituy
• Ranking of Slack users by number of posts in a given week, and associated email domains

ToDo

• Weighted ranking over multiple weeks (looking for increasing activity)
• Normalise metric over activity levels to see growth at all levels
• Use weighted ranking on previous successes, to see if they show up
• Logistic regression (identify predictors)
• Automated contacting? (send "how are you doing" to anyone in the top ranked users, who has not had a DM with a mentor that week)
• refactor code to short methods, more self documenting?