The spreadsheet extrapolates the trend and misses the turns. Weather, prices and the calendar sit in other files. One number, no range, so planning guesses the buffer.
A spreadsheet extrapolates the trend. It cannot see the promotion, the holiday shift or the new competitor until the month has already gone wrong.
Weather, prices, calendar, the customer's own forecast. The factors that move your numbers sit in other files, unread.
A forecast of 1,200 units with no idea whether 900 or 1,500 is plausible. Planning cannot size the buffer, so it guesses.
We build a forecast from your own series and the external signals that matter, tested on the months it never saw, and deliver it with a range into the planning tool your teams already use.
Three parts: the data, the model, the delivery.
Your series, and what moves them.
Tested on the months it never saw.
A forecast that runs itself.
Four phases from the current spreadsheet to a forecast in production.
What is forecast, at what grain, for which decision.
History and factors, refreshed on schedule.
Baseline, candidates, held-out test.
Deliver, track, retrain.
Fewer surprises, buffers sized to the uncertainty, and a forecast people use.
Forecast against actuals, every period, against the baseline it replaced. Improvement is a number the planning team reports.
Stock, staff and cash planned for the interval. Less over-stock in the calm weeks, fewer shortages in the uncertain ones.
It lands where the plan is made and refreshes on its own. The team spends its time on the decision.
| Side by side | Spreadsheet | Model |
|---|---|---|
| Inputs | Last year's numbers | History plus the factors that move it |
| Patterns | Trend and a guess at season | Season, promotions, calendar, lead times |
| Output | One number | A number and a range |
| When things change | Someone edits the sheet | Retrained on the new data, error tracked |
| Accuracy | Rarely measured | Measured every period, against the baseline |
Anything with a history and a decision behind it: demand by product and site, sales by segment, stock and replenishment, cash flow, call or ticket volume, energy and machine load. The first series we pick in discovery, by the value of the decision it feeds and the quality of its history.
Enough to cover the patterns you want captured: two or three full seasons for a seasonal series, less for a stable one. Short histories can borrow from similar series. We say in discovery what the data supports, and measure the baseline before promising anything.
Every forecast comes with an interval, and we check on the held-out periods that actuals fall inside it as often as the interval claims. Planning then sizes buffers to the range. Weeks with wide intervals are flagged so the team looks at them first.
Yes. The forecast is delivered into the ERP, the planning tool or the sheet your team uses, on the schedule they plan on. The data comes from your systems through their APIs or exports, and the refresh runs without anyone pressing a button.
We are a small team of senior specialists. We pick the right model and the right layer, and we build the least machinery that does the job. You get a call with an engineer, not a sales deck.
