You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
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.
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-up1)
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!