TreaterResearch from Treater.
How we build agents that work an opportunity for weeks, and what we learn from the conversations they have.
How merchandising visits, treated as Bayesian evidence alongside POS and orders, reveal the true state of a product's performance.
The √n pooling law assumes stores move independently. Promotions and demand shocks break it, and the failure cascades through the supply chain.
Why modern replenishment systems miss breakout CPG demand, how left-censored sales data deepens the problem, and what brands need to do to close the detection lag.
Why retail demand swings, how DCs absorb it with safety stock, and the √n pooling law that explains why one DC needs ~10× less buffer than the stores it serves.
A technical research note on building evaluation systems for LLM-powered workflows in production environments.
A historical analysis of retail execution, from ancient marketplaces to today's tech-enabled field operations.