The forecast that is really last year plus ten per cent
You're planning demand, cash and headcount on a hunch with a spreadsheet around it.
Demand, cash-flow, inventory and capacity forecasting — rebuilt on your actual history and the drivers that move it, with the accuracy of every forecast measured rather than assumed.
You will recognise at least three of these.
- A forecast built once a year in a spreadsheet, and quietly abandoned by March.
- Method is last year's actuals with a growth percentage applied.
- Seasonality, promotions, weather and pipeline live in people's judgment, not the model.
- Nobody scores the forecast afterwards, so nobody knows if it was any good.
- Stock-outs and overstock are both routine, and treated as bad luck.
Signals it is costing more than anyone has measured.
- You hold safety stock because you don't trust the forecast, and it still runs out.
- Cash surprises happen, in both directions.
- Hiring and capacity decisions are made reactively, at the worst possible price.
- Two departments plan off different assumptions and neither knows.
- Nobody can tell you last quarter's forecast error.
What it becomes.
Forecasting becomes a running system rather than an annual document. It learns from your own history — including the seasonality and the promotions and the weather effects people carry in their heads — and it forecasts at the level decisions are actually made: this SKU at this location, this week; cash by week rather than by quarter; capacity by skill rather than headcount. It publishes a range, not a single number, because the range is what tells you how much slack to hold. Human overrides are first-class, recorded and scored alongside the model. And every forecast is graded against what happened, so accuracy is a tracked number that improves rather than an article of faith.
What the system does
- Forecasting at the level decisions are made — SKU, site, week, skill
- Ranges and confidence, not a single misleading number
- Driver-aware: seasonality, promotions, pipeline, weather, lead times
- Human override, recorded and scored against the model
- Forecast accuracy tracked over time, by product, region and horizon
What it connects to
- Sales, order and POS history — the raw material
- Inventory and purchasing, so a forecast becomes a replenishment suggestion
- Finance, for cash-flow and budget alignment
- CRM pipeline, where demand starts as an opportunity
It augments the systems of record you already run. Nobody is asking you to replace your accounting package.
Where we would start, and how long it takes.
A forecast for one decision — replenishment for a product family, or thirteen-week cash — backtested against your own history so you can see how it would have performed before you rely on it. Usually two weeks, most of it spent on the data rather than the model.
Others in data & decisions
The report someone rebuilds every Monday
“Four exports, one pivot table, every week, forever.”
AI agents & copilotsThe judgment call that happens two hundred times a day
“Not the decisions that need a person — the ones that only got one because there was no alternative.”
Knowledge & internal searchThe answer that exists, in a document nobody can find
“Somebody already solved this. In 2023. In a PDF.”
Let's build
What's your AI Nirvana?
Tell us where you want to go. We'll bring the team, build the product, and grow it with you — and you own it.
- You own the IP
- US-based team
- Reply within 1 business day