CHANGELOG | API | current Break Version:

[com.taoensso/touchstone "2.0.2"] ; Mature/stable (basically "done")

See here if to help support my open-source work, thanks! - Peter Taoussanis

Touchstone: a split testing library for Clojure

A/B testing is great for conversion optimization. We should all be doing more of it. But traditional A/B tests can be a nuisance to setup and monitor.

Touchstone is an attempt to bring dead-simple, high-power split-testing to any Clojure web application. It uses multi-armed bandit techniques to provide fast, accurate, low-maintenance conversion optimization. The API is simple and highly flexible.

Library status

Last updated: Jan 2016

Haven't updated the lib in forever, but it's stable and works well in production. Do have some new stuff planned for a future update (particularly docs re: use with modern Cljs applications), but no ETA on that yet.

- Peter Taoussanis

Features

Tiny, simple API
Great performance backed by Redis and Carmine
High flexibility (variations are arbitrary Clojure forms)
Low maintenace (fire-and-forget, automatic-selection algorithm)
Fire-and-forget multivariate testing
Advanced capabilities like test composition (dependent tests), arbitrary scoring, engagement testing, etc.
Ring middleware

Getting started

Add the necessary dependency to your project:

Leiningen: [com.taoensso/touchstone "2.0.2"] ; or
deps.edn:   com.taoensso/touchstone {:mvn/version "2.0.2"}

And setup your namespace imports:

(ns my-ns
  (:require [taoensso.touchstone :as touchstone :refer (*ts-id*)]))

Split-testing

Traditional split-testing consists of 4 steps:

Defining content variations (e.g. possible labels for a sign-up button)
Distributing content variations to test subjects (our users)
Recording events of interest (sign-ups) by variation
Analyzing the results and adopting our most successful content (best button label)

The particular multi-armed bandit technique used by Touchstone means that we only concern ourselves with steps 1 and 3. Steps 2 and 4 are handled automatically by the algorithm.

To optimize a Ring web application

Start by adding (taoensso.touchstone.ring/wrap-test-subject-id) to your middleware stack.

One or more test selectors can then be used as part of your page content:

(touchstone/mab-select
  {:conn-opts {} ; Optional, as per Carmine's `wcar` conn-opts
   }
  *ts-id* ; Dynamic test-subject-id (assigned by middleware)
  :my-app/landing.buttons.sign-up ; Test id
  :sign-up  "Sign-up!"   ; Named variation #1
  :join     "Join!"      ; Named variation #2
  :join-now "Join now!"  ; Named variation #3
)

And relevant events (e.g. conversions) recorded:

;; On sign-up button click, etc.:
(touchstone/mab-commit!
  {} ; Same opts as given to `mab-select`
  *ts-id* :my-app/landing.buttons.sign-up 1)

Touchstone will now automatically start using accumulated statistical data to optimize the selection of the :my-app/landing.buttons.signup test variations for maximum clicks.

And you're done! That's literally all there is to it.

See the mab-select and mab-commit! API docs for info on more advanced capabilities like multivariate testing, test composition (dependent tests), arbitrary scoring, engagement testing, etc.

Contacting me / contributions

Please use the project's GitHub issues page for all questions, ideas, etc. Pull requests welcome. See the project's GitHub contributors page for a list of contributors.

Otherwise, you can reach me at Taoensso.com. Happy hacking!