One line of JavaScript embedded to collect UX usage data from a live site and stream it to a backend. (Existing solutions are unlikely to be collecting the level of signals needed for the ML to work at its best.)

ML models on the backend to analyze the data and identify user experiences where the user is struggling and highlight those experiences so that founders and team members can quickly find who is struggling and why.

ML modelling will show which signals are predictive but candidates are things like time on page, mouse movements, repeated visits to the same page, opening and closing dropdowns repeatedly, etc.

ML learning will be more effective if the customer defines a success criteria to train against (e.g. a URL visited that shows conversion). But, it may be possible to simply group user experiences into a number of similar groups without the need for the customer to specify success.

By aggregating UX data across a large number of sites, Tealeaves.ai will build a very large data set to continually improve the ML models, and which no individual site could hope to build. This will create a positive flywheel, building a moat.

Comments