- Version
0.1.0
- Source
- Keywords
thumbnail, a/b_testing, multi_armed_bandit, python, api
thumbs up is a tool that utilizes machine learning to help you A/B test your thumbnails in production! This way you can find the best thumbnail to attract the attention of your users to your products while minimizing regret.
Submit an id to the API of a piece of content that you wish to A/B test and add some options to that content. Then add the thumbs-up-js npm package to your project. You can use the Thumbnail component as a wrapper on the content that you wish to A/B test. Whenever your users visit the page with this content, a request will be made to the API to return one of the options that you provided for this item. This option will be chosen based on the decision of the Multi-Armed Bandit algorithm.
If the thumbnail that was provided attracted the attention of the user, this information will be sent back to the API to refine the algorithm.
Over time, the algorithm will find which option is the most "clickable" and cash in on that one to help you optimize your traffic to that product.
It is possible to use this API without using the React.js library, however that will require you to build a custom solution, rather than the out-of-the-box ready solution that is provided by thumbs-up-js.
Thumbnails are often the first thing that a user will see on your page; however a bad or bland thumbnail is easily overlooked. Choosing a good thumbnail is one of the best ways to increase traffic to your products, videos or blog posts. The virality of many videos on Youtube or content on Instagram can be attributed to the ingenious marketing packaged into the content's thumbnails.
thumbs up utilizes machine learning to help you find the most clickable thumbnail out of a set that you provide based on real world interactions with your users. This can significantly increase traffic to your products!
In marketing terms, a multi-armed bandit solution is a ‘smarter’ or more complex version of A/B testing that uses machine learning algorithms to dynamically allocate traffic to variations that are performing well, while allocating less traffic to variations that are underperforming. In theory, multi-armed bandits should produce faster results since there is no need to wait for a single winning variation.
The term "multi-armed bandit" comes from a hypothetical experiment where a person must choose between multiple actions (i.e. slot machines, the "one-armed bandits"), each with an unknown payout. The goal is to determine the best or most profitable outcome through a series of choices. At the beginning of the experiment, when odds and payouts are unknown, the gambler must determine which machine to pull, in which order and how many times. This is the “multi-armed bandit problem.”
More at: Optimizely
Also: Wikipedia
So now that we know what a multi-armed bandit solution is, which implementation is used by thumbs up?
There are many variations on the multi-armed bandit (Epsilon-Greedy, Softmax, UBC-1, etc.), but we have chosen to use the Bayesian Theorem (AKA Thompson Sampling or Randomized Probability Matching). Why?
The Bayesian Theorem will help us answer that question; it states the following:
For two events Y and Z, the conditional probability of Y given Z is the conditional probability of Z
given Y scaled by the relative probability of Y compared to Z.This means that for every iteration of our trial, we scale the distribution of thumbnails based on the set of previous distributions of clicks.
Basically, after every trial we recompute the probabilities of each thumbnail's superiority, scaled by the number of trials in total. This approach will help us get the results of the experiment in as little time as possible, with less data, AND the same level of statistical validity as regular A/B testing.
If that isn't convincing enough, take Google's word for it. Google uses the Bayesian multi-armed bandit for Google Analytics!
For a detailed analysis of the Bayesian multi-armed bandit algoritm: A modern Bayesian look at the multi-armed bandit
If you are new here, these are some resources that may be helpful:
Getting started with thumbs up
A quick introduction to a simple setup with thumbs up.
-
A comprehensive overview of the API and all of its functionality.
If you have any suggestions, bug reports, or annoyances please report them to our issue tracker at https://github.com/best-bet/thumbs-up-api/issues/
Development of thumbs-up-api happens at GitHub: https://github.com/best-bet/thumbs-up-api
Please feel free to participate in development, this project is only made possible through open source contributions.
Be sure to also read the Contributing to thumbs up section in the documentation.
If you are interested in tracking development, check out our scrum board!
This software is licensed under the MIT License. See the LICENSE file in the top distribution directory for the full license text.