How it works

From a selection to a ranked forecast in seven steps — all inside Excel.

Tip: click any illustration to open it larger, and use ‹ › to walk through the whole flow.

1Select & forecast

Select date + value columns in Excel and click Forecast this selection on the WhichForecast ribbon (or load a sample from the library).

2Data quality

Blanks and outliers are flagged, never silently deleted. Usable points, zeros, and a quality verdict are shown.

3Profile

Series profile: trend, seasonal period, intermittency class, and Candidate Policy — before any method runs.

4Goal

Goal for the forecast: demand, inventory, price, FX, weather, energy, or other. One product: the wizard infers nature; this only sets constraints and the optional decision layer.

5Methods

Ranked forecast methods: eligible candidates are backtested with rolling-origin MASE — best supported among candidates. Pro methods show as locked on Free.

6Evidence

Method DNA, Learn copy, history / backtest / residual charts. Pin a method with a note if you disagree with the rank.

7Forecast & export

Forward forecast with intervals and optional safety stock / ROP. Free: JSON, watermarked note, and a summary WhichForecast_Results sheet. Pro: formatted Excel sheet, PDF of the method note, Python, and R snippets.

Educational analysis only. Not investment, meteorological, or operational advice. Past backtests do not guarantee future accuracy.