BayesianWitch

Bandit Algorithms are better than A/B testing for transient content - i.e., a single blog post. The reason is that A/B tests attempt to determine the one right answer, whereas Bandit Algorithms simply try to increase the number of clicks. As a result, Bandit Algorithms can increase your performance with fewer clicks than A/B testing. Bandit Algorithms are automatic - once you write the post, you can forget about it. Even if your blog has a small number of readers per post (too small for A/B testing to work), Bandit Algorithms can increase the number of clicks you recieve.

No. BayesianWitch loads it's javascript asynchronously, which is a fancy way of saying that we don't load our javascript until after your page is fully loaded.

At BayesianWitch we take reliability very seriously. We take several steps to ensure that your blog always works. First, if your readers are unable to connect to our servers (perhaps they have a slow connection), your blog post will still work. All that will happen is a random call to action will be displayed - it's not optimal, but your blog post will not be broken. You can test this out yourself - set up a wordpress blog on your laptop and create a post with a BayesianWitch bandit. Then disconnect from the internet - your blog post will still work and the call to action will still display. Second, on the server side, we use all the industry standard devops techniques to keep our servers up and running. We have automated monitoring, internal data collection, the works. We also monitor things 24/7.

BayesianWitch tracks your user's behavior on the site - pageviews, clicks, tweets, likes, etc. We analyze all that data in order to determine which call to action is performing better.