1380713 - Complete analysis to determine final column (features) set that will power TAAR reccommendations for new users

Status: VERIFIED FIXED

(Data Platform and Tools :: General, enhancement)

Description

[mlopatka]

The TAAR project requires a way of making recommendations to users without any installed addons. In order to provide personalized recommendations to such users some subset of telemetry variables can be used as side-information. An analysis must be completed to select a short list of fields to be used in such cases.

Comment 1

[mlopatka]

Most valuable features seem to be:

Geo_city
Subsession_start_date (repressented as elapsed time)
Locale
Bookmarks
Tab_open_count
Total_uri
Unique_tlds

analysis report: https://docs.google.com/a/mozilla.com/document/d/1bKiaDDjj0SG_Xv7aZ1umoZ-Mu-WbkL09ytbe7gotH38/edit?usp=sharing

Comment 2

[mlopatka]

In addition to the above fields we have also decided to include the Operating system (in order to specifically target undersampled linux users). 

Also, as mentioned previously, the list of installed (both acive and inactive addons) should be included as well.

The final list is then defined as:

client_id
geo_city
subsession_start_date (repressented as elapsed time)
locale
os
bookmarks (count)
tab_open_count
total_uri
unique_tlds
insatalled_addon (vector including singed state, active state, guid, etc...)

Comment 3

[Firefox Bug Husbandry Bot]

Moved to new component, per bug 1425844.