Turn sales history into smart, actionable forecasts fast. Test multiple models and tune assumptions easily. Download tables, share insights, and plan with confidence now.
Paste sales values, or fill the quick grid. Then choose a forecasting method and settings.
| Period | Sales |
|---|---|
| T1 | $1,200 |
| T2 | $1,350 |
| T3 | $1,280 |
| T4 | $1,425 |
| T5 | $1,500 |
| T6 | $1,610 |
| T7 | $1,540 |
| T8 | $1,720 |
Accurate sales forecasting starts with clean, ordered history. Enter values from oldest to newest, keep the same time granularity, and remove one off distortions when possible. A short series can work, but at least 12 periods helps reveal patterns. If promotions or stockouts caused unusual dips or spikes, note them separately so you do not treat them as normal demand.
This tool offers four practical approaches often used in applied machine learning workflows. Simple moving average smooths noise using a fixed window. Weighted moving average adds emphasis to recent periods through normalized weights. Exponential smoothing adapts with a single alpha that controls responsiveness. Linear trend regression fits a straight line over time and reports R2 to indicate how much variance the trend explains.
Forecasts are more useful when you measure how they would have performed on recent observations. The calculator runs a rolling one step backtest over a small holdout window and reports MAE, RMSE, and MAPE. MAE shows average absolute error in the same unit as sales. RMSE penalizes larger misses more strongly. MAPE expresses typical error as a percentage, helping compare series with different scales.
To communicate uncertainty, the results include a simple lower and upper band around each forecast point. The band is built from the backtest RMSE and a selected z value for 90%, 95%, or 99% confidence. Treat these bounds as planning ranges rather than guaranteed limits. Wider ranges often signal volatility, sparse data, or a method that does not match the underlying dynamics.
Operational decisions improve when forecasts become inputs to capacity and revenue plans. Use the horizon to align with purchasing lead times, staffing cycles, or pipeline review cadences. If seasonality is known, apply indices to scale the base forecast for repeating peaks and troughs. Re run scenarios with different methods and parameters, then select the model that minimizes error while still matching your business intuition. Pair the forecast with target conversion, average order value, and churn assumptions to estimate revenue scenarios, and document chosen settings for consistent reporting each month.
Start with Simple Moving Average for stable series, or Exponential Smoothing when recent changes matter. If you expect a clear upward or downward drift, try Linear Trend Regression. Compare MAE, RMSE, and MAPE to choose.
A smaller window reacts faster but can chase noise. A larger window is smoother but may lag. Try 3 to 6 periods for weekly data, or 3 to 12 for monthly data, then compare backtest errors.
Alpha sets how much weight the latest observation receives in smoothing. Higher alpha follows changes quickly and can be more volatile. Lower alpha is steadier and better for stable demand. Keep alpha between 0.01 and 0.99.
MAE shows average absolute miss in sales units. RMSE punishes large misses more, highlighting risk. MAPE expresses typical error as a percentage, useful for comparing products or regions with different sales scales.
Use indices when demand repeats on a fixed cycle, such as monthly seasonality. Provide one index per season position, like 12 values for months. The calculator multiplies each forecast point by the matching index.
No. The bounds are planning ranges based on recent backtest error and a confidence level. Real outcomes can fall outside the range, especially during promotions, supply constraints, or sudden market shifts.
Important Note: All the Calculators listed in this site are for educational purpose only and we do not guarentee the accuracy of results. Please do consult with other sources as well.