AI Model Data Requirements
Each AI model trains on specific input tables. This section describes the User and Event tables each model needs — the mandatory fields, the recommended enrichments, and the minimum data volumes for reliable results.
This is the data side of AI models. For how to configure and run a model once your data is ready, see AI Models → Configure a Model.
:::info Field names are flexible Across all models, field names are examples — identity resolution stitches users across tables regardless of column naming. What matters is the grain (1 row per user / 1 row per event), the presence of a user identifier, and a UTC timestamp on event rows. :::
Models
| Model | Primary inputs |
|---|---|
| Action Prediction | User table (+ optional event table to derive the target) |
| pcLTV | Transaction event table (+ optional user table) |
| RFM | Transaction event table |
| Interest | Event table with event_label or page_url |