Ads, demand and a plan people use.
Retail and distribution · multi-channel operations. One layer across systems that don't reconcile, ad spend that follows demand instead of trailing it, and a demand plan a planner actually signs off on.
Three systems, one spreadsheet, last month's picture
What we walked into
Orders in one system, inventory in another, spend in a third, and a team reconciling all three in a spreadsheet every week. Marketing therefore spent against last month's picture: the fast movers went out of stock mid-campaign while cash sat in stock nobody was pushing.
What we built
A layer that unifies those systems and encodes the reconciliation rules the team carried in their heads, puts SKU-level demand on top, and reallocates spend across channels and creatives against what is actually going to sell. Anomalies — a channel that stopped reporting, a SKU whose returns spiked, a vendor drifting off agreed rates — go to the person who owns that number, as an alert, not another dashboard.
Spend follows demand; cash stops sitting in stock.
Figures from the live engagement; client name withheld.
A demand plan a planner will actually use.
Anyone can fit a model to two years of shipments. Whether the plan gets used is decided by workflow, not modelling — so that is what we built.
The history is dirty, and nobody tells you
A week with zero sales because the SKU was out of stock is not zero demand — it is a censored observation, and training on it teaches the model to under-forecast exactly the SKUs that keep selling out. Cleaning the stockouts, promos and launch ramps out of the history is the product.
A planner who can't see why won't use the number
The baseline arrives decomposed — trend, seasonality, promo, event, price — as a waterfall a planner can argue with one bar at a time, instead of a single forecast line they can only accept or reject whole.
The model never publishes a number on its own
Every figure that leaves the system was accepted or set by a named planner, with a reason code. The cells the model is unsure about are routed to a person; the rest clear automatically.
Overrides are scored against what happened
Forecast value-add — did the planner's change beat the model or make it worse — is measured by name. It is the metric that turns a forecast into a discipline, and the reason the plan is still trusted in month nine.
Built on your warehouse.
Is your forecast a plan, or a sales target with a different name?
The first deliverable is often just measurement — a fast, unarguable win. In a five-day discovery we run your own history and show you what the stockouts have been hiding.
Book a 30-minute call