
TA Scoring
60 boosted trees learn who converts, and the holdout decides when to stop.
Interactive modelEN / 中文
Six consumer data models in the Fader and Hardie tradition, fitted and run live in your browser on synthetic data. Each page opens with a business question, shows the model at work, and lands on the decision it feeds.

60 boosted trees learn who converts, and the holdout decides when to stop.

k-means finds four planted customer groups live, with the inertia curve to prove it.

A BG/NBD model fitted in the browser, then your edits to one customer's timeline replayed as 200 futures.

A purchase matrix measures its own item similarity, then ranks a list for any customer.

Response curves learned from a year with dark weeks, then the mix rebuilt on sliders.

Pick target, cohort, window. The backtest report card comes first.