Forecasting Methodology
How Our AI Forecasting Works
A comprehensive guide to understanding Finacast's forecasting models, mathematical foundations, and how to interpret results with confidence.
18+ Forecasting ModelsFull Mathematical FormulasDetailed Worked Examples
The Big Picture
Finacast analyzes your historical financial data and automatically selects the best forecasting method for each account. Think of it as having a team of expert forecasters who:
- 1Examine your data - Look for patterns like seasonality, growth trends, and volatility
- 2Test multiple approaches - Try 18+ different forecasting methods on your data
- 3Pick the winner - Select the method that performs best for your specific data
- 4Generate forecasts - Produce predictions with confidence ranges
- 5Maintain consistency - Ensure your P&L relationships stay realistic (e.g., COGS scales with Revenue)
What Makes This Different from Spreadsheet Forecasting?
| Traditional Spreadsheet | Finacast |
|---|---|
| One method for everything (usually "last year + X%") | Optimal method selected for each account |
| Manual pattern identification | Automatic seasonality and trend detection |
| Fixed assumptions | Data-driven, self-adjusting |
| Single point estimate | Range of outcomes with probabilities |
| Easy to break P&L relationships | Maintains Gross Margin consistency |
Quick Reference
When to Use Which Model
- âĒ Flat, stable â Moving Average or Naive
- âĒ Steady growth â Linear Trend (OLS)
- âĒ Seasonal pattern â Holt-Winters or ETS
- âĒ Complex seasonal â SARIMA
- âĒ Many features â XGBoost or Random Forest
- âĒ Holidays matter â Prophet
- âĒ Lots of zeros â Croston or SBA
Key Formulas
- âĒ MAPE = Avg(|A-F|/|A|)
- âĒ CI = Åķ Âą 1.96ÃÏÃâh
- âĒ ETS: Åķ = (L+hT)ÃS
- âĒ Croston: Åķ = Z/P
- âĒ OLS: Åķ = Îą + Îēt
Ready to see these forecasting models in action?