Three requirements
- History — at least two full seasonal cycles, ideally three years
- Stability — the patterns that held must still hold
- Volume — forecasting five orders a month is not forecasting
If two of those three are missing, the honest answer is that no technique will produce a forecast you should act on. That answer is worth giving.
Where it works
| Application | Fit | Requirement |
|---|---|---|
| Seasonal demand, established lines | Strong | Three years of history |
| Cash flow from invoice history | Strong | Consistent payment behaviour |
| Staffing against booked work | Strong | Reliable booking data |
| New product demand | Weak | No history to learn from |
| After a business model change | Weak | Past patterns no longer apply |
Use the right tool
For numerical forecasting, established statistical methods outperform language models substantially and cost a fraction. Seasonal decomposition and time series methods are mature and well understood.
Language models help with explaining a forecast in plain English and incorporating unstructured signals, not with producing the numbers.
Present ranges, not points
- A range with a stated confidence level, not a single figure
- The same model's accuracy on recent history, so trust is calibrated
- The assumptions stated explicitly
- What would invalidate the forecast
A single number invites false confidence. A range invites planning, which is what a forecast is for.
Be clear about the limits
Forecasting extrapolates observed patterns. It does not anticipate breaks in them, which is precisely when people most want a forecast.
Used for reordering, staffing and cash planning it saves real money. Used for strategy it produces spurious precision.