🍁 A/B testing for Rails

• Designed for web and email
• Comes with a dashboard to view results and update variants
• Uses your database for storage
• Seamlessly handles the transition from anonymous visitor to logged in user

Uses Bayesian statistics to evaluate results so you don’t need to choose a sample size ahead of time.

Add this line to your application’s Gemfile:

gem "field_test"

Run:

rails generate field_test:install
rails db:migrate

And mount the dashboard in your config/routes.rb:

mount FieldTest::Engine, at: "field_test"

Be sure to secure the dashboard in production.

Add an experiment to config/field_test.yml.

experiments:
  button_color:
    variants:
      - red
      - green
      - blue

Refer to it in controllers, views, and mailers.

button_color = field_test(:button_color)

To make testing easier, you can specify a variant with query parameters

http://localhost:3000/?field_test[button_color]=green

When someone converts, record it with:

field_test_converted(:button_color)

When an experiment is over, specify a winner:

experiments:
  button_color:
    winner: green

All calls to field_test will now return the winner, and metrics will stop being recorded.

You can keep returning the variant for existing participants after a winner is declared:

experiments:
  button_color:
    winner: green
    keep_variant: true

You can also close an experiment to new participants without declaring a winner while still recording metrics for existing participants:

experiments:
  button_color:
    closed: true

Calls to field_test for new participants will return the control, and they won’t be added to the experiment.

You can get the list of experiments and variants for a user with:

field_test_experiments

For JavaScript and native apps, add calls to your normal endpoints.

class CheckoutController < ActionController::API
  def start
    render json: {button_color: field_test(:button_color)}
  end

  def finish
    field_test_converted(:button_color)
    # ...
  end
end

For anonymous visitors in native apps, pass a Field-Test-Visitor header with a unique identifier.

Any model or string can be a participant in an experiment.

For web requests, it uses current_user (if it exists) and an anonymous visitor id to determine the participant. Set your own with:

class ApplicationController < ActionController::Base
  def field_test_participant
    current_company
  end
end

For mailers, it tries @user then params[:user] to determine the participant. Set your own with:

class ApplicationMailer < ActionMailer::Base
  def field_test_participant
    @company
  end
end

You can also manually pass a participant with:

field_test(:button_color, participant: company)

To get variants in jobs, models, and other contexts, use:

experiment = FieldTest::Experiment.find(:button_color)
button_color = experiment.variant(user)

By default, bots are returned the first variant and excluded from metrics. Change this with:

exclude:
  bots: false

Exclude certain IP addresses with:

exclude:
  ips:
    - 127.0.0.1
    - 10.0.0.0/8

You can also use custom logic:

field_test(:button_color, exclude: request.user_agent == "Test")

Keep track of when experiments started and ended. Use any format Time.parse accepts. Variants assigned outside this window are not included in metrics.

experiments:
  button_color:
    started_at: Dec 1, 2016 8 am PST
    ended_at: Dec 8, 2016 2 pm PST

Add a friendlier name and description with:

experiments:
  button_color:
    name: Buttons!
    description: >
      Different button colors
      for the landing page.

By default, variants are given the same probability of being selected. Change this with:

experiments:
  button_color:
    variants:
      - red
      - blue
    weights:
      - 85
      - 15

To help with GDPR compliance, you can switch from cookies to anonymity sets for anonymous visitors. Visitors with the same IP mask and user agent are grouped together.

cookies: false

If the dashboard gets slow, you can make it faster with:

cache: true

This will use the Rails cache to speed up winning probability calculations.

If you need more precision, set:

precision: 1

You can set multiple goals for an experiment to track conversions at different parts of the funnel. First, run:

rails generate field_test:events
rails db:migrate

And add to your config:

experiments:
  button_color:
    goals:
      - signed_up
      - ordered

Specify a goal during conversion with:

field_test_converted(:button_color, goal: "ordered")

The results for all goals will appear on the dashboard.

You may also want to send experiment data as properties to other analytics platforms like Segment, Amplitude, and Ahoy. Get the list of experiments and variants with:

field_test_experiments

You can configure Field Test to use Ahoy’s visitor token instead of creating its own:

class ApplicationController < ActionController::Base
  def field_test_participant
    [ahoy.user, ahoy.visitor_token]
  end
end

authenticate :user, ->(user) { user.admin? } do
  mount FieldTest::Engine, at: "field_test"
end

Set the following variables in your environment or an initializer.

ENV["FIELD_TEST_USERNAME"] = "moonrise"
ENV["FIELD_TEST_PASSWORD"] = "kingdom"

Assign a specific variant to a user with:

experiment = FieldTest::Experiment.find(:button_color)
experiment.variant(participant, variant: "green")

You can also change a user’s variant from the dashboard.

To associate models with field test memberships, use:

class User < ApplicationRecord
  has_many :field_test_memberships, class_name: "FieldTest::Membership", as: :participant
end

Now you can do:

user.field_test_memberships

A huge thanks to Evan Miller for deriving the Bayesian formulas.

View the changelog

Everyone is encouraged to help improve this project. Here are a few ways you can help:

• Report bugs
• Fix bugs and submit pull requests
• Write, clarify, or fix documentation
• Suggest or add new features

To get started with development:

git clone https://github.com/ankane/field_test.git
cd field_test
bundle install
bundle exec rake compile
bundle exec rake test