A calibrated twin of your category — watch thousands of agents shop, then try any price before you set it.
Frozen Pizzaa living market — shoppers walk to the brand they pick · calibrated on real data
Try it:demo:
or upload — needs product, price, units. Your file is processed in memory and not stored.
Prices — drag to run a what-if
vertical mark = actual market share (validation anchor) · NEW = predicted from positioning, no sales history (simulation mode)
Your brand
Your unit cost $
Category share
–
Revenue index
–
Raise your price 10% → who wins the shoppers:
Your price → share & revenue curve
Promo simulator — is this promotion worth it?
Discount25%
Incremental = promo units minus the model's no-promo counterfactual. Same held-out-validated engine as our public promo backtest.
Have real promo history? Measure the true incrementality (ex-post)
Upload a CSV of your promotions — product, price, units (+ optional base_price, feature, display, store, week) — and we compute the incremental units each promo really drove vs the model's no-promo counterfactual, with held-out validation and honest bounds. Processed in memory, not stored.
The agent population
Substitution matrix — who wins when a rival raises price
each row = that product raised price +10% · greener cell = the column product captured more of the freed-up share (cross-price substitution)
Compare scenarios (A vs B)
Set prices / assortment / promo, pick your brand, then save the plan — compare two side by side.
Foothold · agent-to-agent deal — real, grounded in the calibration (not a mockup)
A buyer-agent sources the category; each brand's seller-agent quotes a price from the calibrated model; they negotiate; the winning deal is logged. Every number comes from the demand model.
The outcome ledger — every agent deal, its result & would-be fee
Every agent deal is logged with its outcome and the fee it would earn — a record that compounds as the network grows.