Experiment with recommendation of add-ons for users without any currently installed add-ons based on other features.
Martin, I think this is a P1 (current sprint) and not a P3 (next quarter). See also https://wiki.mozilla.org/Data/Platform/General_sprint_info.
Adding dzeber to the bug to weigh in from the Metrics team.
Changed the time frame to span the quarter, but also the points to reflect that this should be an ongoing exploration. Initial notebook is finished, but there is so much more to look at in terms of exploring the different parameters and how they affect the recommendation behavior.
Prototype finished. No public gist, but ask me and I will add you to the repo. I'm going to clean this up and add documentation for presenting via screen-share.
Uploaded a gist demonstrating the basic workflow: modelling the corpus, pairwise comparisons, and new addon recommendations https://gist.github.com/mlopatka/7089ec209c6941ca860f6670b84d2a60
updated gist (13-10-2016) https://gist.github.com/mlopatka/55ae54a19ca49bdd3a22ddf0b1350649