🧠 Models — the model layer that emerged from your data
Each replaces a hand-coded heuristic with a model learned from your own data —
no LLM, holdout-validated, retrainable. They power the est-sales, momentum and category signals
across the app.
🔮 Try it all three models on one product
📦 Sales model predicts monthly units · trained on Amazon's "bought" badge
Not trained yet.
📈 Demand-momentum forecaster self-supervised · the future is the label
Not trained yet.
🏷 Category classifier audit tool — suggests, never silently overrides
Not trained yet.
💷 Price-elasticity model how demand responds to price · learned per category
Not trained yet.
Coming next (needs more data)
An opportunity-success model — once the Scout accumulates enough approve→outcome history,
the same recipe trains a model that predicts which opportunities actually win. No labels yet; it starts the moment
you begin approving.