Use case: Write an AI style Lambda function on AWS or Google Cloud, along with other supporting services and send your impression and click rates for each of your variants on your unbounce page. So far so good. All very doable today on unbounce with javascript on a page hitting a REST API to store impressions and clicks as they happen. Next I want to gather those impressions and clicks and send them to my AI function and do something with the results to tell me as fast as possible which of my variants is the winner.
The problem is, how do you act on this information returned from your AI function? The idea is to write a multi-armed bandit style of function you can use to automate the finding of your winner variant quickly.
You can’t do this today (unless I’m missing something) because you would have to manually go into unbounce and adjust your weights for each variant based on your return values from your AI function. Being able to adjust the weights for each variant via the API would allow you to automate this process and integrate whichever AI/bayesian model you prefer to allow you to make your A/B decisions faster and smarter. I know there are webhooks, and Zapier but they trigger off a form submission event which is not sufficient for all landing pages, hence the Javascript solution. Now if I could just adjust the weights via the unbounce API, I’d be all set.