A plain-English guide to forecasting for finance teams

A forecast is a structured guess. The aim is not to be right every time. It is to be wrong by a known amount, so you can plan around it.

Start with a baseline

Before anything clever, try the simple options: last year’s figure, or the average of the last three months. If a complicated model cannot beat the baseline, it is not worth having.

Ask for the range

A single number hides risk. A good forecast says “between 4.1 and 4.6 million, most likely 4.35”. That range is the useful part, because it tells you how much buffer to hold.

Check it against the past

Ask to see a back-test: how would the model have done on last year’s data, had you used it then? If nobody can show you one, treat the forecast with caution.

Keep a record

Write down each forecast and what actually happened. After six months you will know which numbers to trust, and the model’s makers will know what to fix.