What are the different user personas that use Alaveteli?

[garethrees]

Q: What are the different user personas that use Alaveteli?

We assume there are:

• Requesters
• Commenters
• Trackers (sign up primarily to follow requests)
• Classifiers (do a lot of request classification)

  Measurement

Need to do a bit of thinking about this, but we want to be able to say something like "20% of users never make a request, but have more than one track".

Requirements

Requires https://github.com/mysociety/alaveteli/issues/3283 so that we can profile existing users.

[garethrees]

RE: profiling, after a bit of experimentation in the console here's an initial sketch run over WDTK users:

ActiveRecord::Base.logger.level = 1 

def user_type(data)
  return :ghost          if data.values.sum.zero?
  return :single_request if data[:info_requests] == 1 && data.values.sum == 1
  return :commenter      if data[:comments] > data.dup.slice!(:comments).values.sum
  return :requester      if data[:info_requests] > data.dup.slice!(:info_requests).values.sum
  return :classifier     if data[:request_classifications] > data.dup.slice!(:request_classifications).values.sum
  return :follower       if data[:track_things] > data.dup.slice!(:track_things).values.sum
  #
  # Doesn't work yet because of missing association
  # return :body_updater if data[:public_body_change_requests] > data.dup.slice!(:public_body_change_requests).values.sum
  return :other
end

User.where(:ban_text => '').count
# => 97052

puts Time.zone.now
# => 2016-06-16 12:15:47 +0100

profiled_users = {}

User.where(:ban_text => '').find_each do |user|
  data = {
    :comments => user.comments.count,
    :info_requests => user.info_requests.count,
    :request_classifications => user.request_classifications.count,
    # Doesn't work yet because of missing association
    # :public_body_change_requests => user.public_body_change_requests.count
    :track_things => user.track_things.count
  }

  profiled_users[user_type(data)] ||= 0
  profiled_users[user_type(data)] += 1
end

puts Time.zone.now
# => 2016-06-16 12:51:37 +0100

profiled_users
# Manually sorted
# => { :single_request => 30306,
#      :ghost          => 25953,
#      :requester      => 20649,
#      :follower       => 11018,
#      :other          =>  7046,
#      :commenter      =>  1674,
#      :classifier     =>   400 }

• The distribution feels about right. Lots who have done nothing and made a single request; outside of that, mainly people who make requests. A fair amount of followers and less of everyone else.
• It would be interesting to look in to :other, but it probably wouldn't affect the overall picture. An example of someone that we might want to class as a commenter, but is getting caught in :other:

data = {
  :comments => 10,
  :info_requests => 4,
  :request_classifications => 4,
  :track_things => 4
}

• It takes an age to run this. Probably a lot quicker with the counter caches in place, but if not might want to look at doing this in SQL.
• It doesn't take user age in to account. I don't think this matters too much if we're only concerned about a current snapshot.

[garethrees]

Results for AskTheEU:

User.where(:ban_text => '').count                                                                                                                                   │irb(main):025:0*
# => 2176

profiled_users
# => { :ghost          => 1101,
#      :single_request =>  399,
#      :requester      =>  309,
#      :follower       =>  281,
#      :other          =>   67,
#      :commenter      =>   14,
#      :classifier     =>    5 }

Similar, except ~50% of users have done nothing at all, compared to ~30% on WDTK.

[garethrees]

Same story for IPV:

User.where(:ban_text => '').count
# => 657

profiled_users
# => { :ghost          => 234,
#      :single_request => 226,
#      :requester      => 132,
#      :follower       =>  30,
#      :other          =>  28,
#      :commenter      =>   4,
#      :classifier     =>   3 }

[garethrees]

Dumb table just for the sake of comparison.