Great summary - there's lots of subtleties in this space and most of our home-brew solutions would miss a lot. This company is an example of experts leveraging their knowledge to benefit us all. Well done!
Thanks! Based on my understanding of Sift Science as an outsider, we're focused on a different market. We're explicitly trying to stay away from traditional e-commerce applications (shopping carts, standard transactional businesses, etc) because the space is so crowded, preferring instead to focus on trust and safety applications (i.e. two untrusted users interacting with each other on some service vs 1 untrusted user interacting with a merchant).
Our developer docs at http://docs.smyte.com/ should have the answers here, but basically you can drop a script tag on your page to get a basic view of what's going on, and use our REST API server-side to annotate that basic view with additional data. Ideally, you'd serialize most of your models to JSON and send them over to Smyte so it can detect patterns and relationships.
Definitely check the docs as Pete suggested, but would also like to point out that we'll take just as much data as you're comfortable with and/or can send for privacy reasons.
We have a ton of different signals that make our system tick, but based on whatever you send us it'll function as best as it can.
