A/Bingo Version 1.0.3 ====== Rails A/B testing. One minute to install. One line to set up a new A/B test. One line to track conversion. For usage notes, see: http://www.bingocardcreator.com/abingo Installation instructions are below usage examples. Key default features: -- Conversions only tracked once per individual. -- Conversions only tracked if individual saw test. -- Same individual ALWAYS sees same alternative for same test. -- Syntax sugar. Specify alternatives as a range, array, hash of alternative to weighting, or just let it default to true or false. -- A simple z-test of statistical significance, with output so clear anyone in your organization can understand it. Example: View <% ab_test("login_button", ["/images/button1.jpg", "/images/button2.jpg"]) do |button_file| %> <%= img_tag(button_file, :alt => "Login!") %> <% end %> Example: Controller def register_new_user #See what level of free points maximizes users' decision to buy replacement points. @starter_points = ab_test("new_user_free_points", [100, 200, 300]) end Example: Controller def registration if (ab_test("send_welcome_email"), :conversion => "purchase") #send the email, track to see if it later increases conversion to full version end end Example: Conversion tracking (in a controller!) def buy_new_points #some business logic bingo!("buy_new_points") #Either a conversion named with :conversion or a test name. end Example: Conversion tracking (in a view) Thanks for signing up, dude! <% bingo!("signup_page_redesign") > Example: Statistical Significance Testing Abingo::Experiment.last.describe_result_in_words => "The best alternative you have is: [0], which had 130 conversions from 5000 participants (2.60%). The other alternative was [1], which had 1800 conversions from 100000 participants (1.80%). This difference is 99.9% likely to be statistically significant, which means you can be extremely confident that it is the result of your alternatives actually mattering, rather than being due to random chance. However, this doesn't say anything about how much the first alternative is really likely to be better by." Installation ======= 1) REQUIRED: You'll need to generate a DB migration to prepare two tables, then migrate your database. (Note: slight edits required if you use the table names "experiments" or "alternatives" at present.) Note: if you are upgrading to A/Bingo 1.0.0, you'll want to do this again. ruby script/generate abingo_migration rake db:migrate 2) REQUIRED: You need to tell A/Bingo a user's identity so that it knows who is who if they come back to a test. (The same identity will ALWAYS see the same alternative for the same test.) How you do this is up to you -- I suggest integrating with your login/account infrastructure. The simplest thing that can possibly work #Somewhere in application.rb before_filter :set_abingo_identity def set_abingo_identity if (session[:abingo_identity]) Abingo.identity = session[:abingo_identity] else session[:abingo_identity] = Abingo.identity = rand(10 ** 10).to_i end end 3) RECOMMENDED: A/Bingo makes HEAVY use of the cache to reduce load on the database and share potentially long-lived "temporary" data, such as what alternative a given visitor should be shown for a particular test. You SHOULD use a cache which is shared across all Rails processes -- that probably means MemcachedStore or RedisStore. You PROBABLY SHOULD use a persistent cache in case you need to restart your machine. This is an amazingly good use case for MemcacheDB, so if you want to try playing with that, Google it. (Sets up VERY easily on the newer Ubuntu distros.) If you can't use a persistent cache, you're probably still OK if Memcached very rarely needs to be restarted. If the cache gets flushed, you will double-count entrants to a particular experiment and possibly double-count conversions, but that may not be the worse thing in the world. A/Bingo defaults to using the same cache store as Rails. If you want to change it #production.rb Abingo.cache = ActiveSupport::Cache::MemCacheStore.new("cache.example.com:12345") #best if really memcacheDB Copyright (c) 2009-2010 Patrick McKenzie, released under the MIT license