Sales forecast calculator

How it works

level = α·actual + (1−α)·(level + trend) · trend = β·(level change) + (1−β)·trend · forecast(h) = level + h·trend

This is Holt's linear trend — double exponential smoothing — which is what most 'AI sales forecasting' turns out to be once you open it. Alpha decides how much the newest figure moves the level; beta how much it moves the trend. Higher values react faster and are noisier. It deliberately does NOT fit seasonality: that needs at least two full cycles of history, and a tool that silently fits a season to eight months of data produces confident nonsense. If your business is seasonal, forecast against the same period last year rather than the month just gone.

This tool runs entirely in your browser. Nothing you type is sent to a server, stored or logged.

Frequently asked questions

How much history do I need?

Two points is the arithmetic minimum and far too few to trust. Eight to twelve periods gives the trend something to settle on; below six, read the result as a direction rather than a number.

What should alpha and beta be?

The defaults — 0.3 and 0.1 — suit most monthly business data. Raise alpha if your series shifts level abruptly and you want the forecast to follow; lower it if the series is noisy and you keep chasing spikes.

Why does it not handle seasonality?

Because fitting a seasonal component honestly needs two or more complete cycles, and most people arrive with a year or less. A tool that fits one anyway returns a confident wrong answer, which is worse than a plain trend.

Should I trust a long horizon?

No. A linear trend extrapolated far enough always becomes absurd — it does not know your market is finite. Three to six periods is where this is useful; beyond that it is arithmetic, not a forecast.