Growth Engineering Playbook · Wave 5

A forecast without a range is just a guess in a suit.

Planning runs on a single revenue number for next quarter — but the number that matters is the range, and how fast it widens as you look further out. This fits a trend and monthly seasonality to three years of Northstar Outfitters revenue, projects it forward with prediction intervals that grow with the horizon, and — the part dashboards skip — backtests on held-out months so you can see the error (MAPE) and whether the intervals actually contain reality (coverage). Deterministic, client-side.

Fitting forecast…

Revenue: history & forecast

Solid = actual. Blue = forecast, with the shaded prediction interval widening as the horizon extends.

Actual Forecast Prediction interval

Seasonal profile

Each month's revenue vs the annual average — the shape the model projects forward.

Backtest — is the forecast trustworthy?

Trained on all but the last 12 months, then scored on the months it never saw.

Monthly forecast

Point forecast with the low/high bounds of the prediction interval.

The Decision