Home / Logistics & inventory

Demand forecasting

Paste your sales or usage history and the calculator tests six methods, picks the lowest error and projects the next months with a confidence band, ready to carry into safety stock and economic lot size.

Pre-sizing and estimating tool: results do not replace design, calculations or reports signed by a licensed professional, nor verification against the applicable codes and standards.

Preliminary estimateRevised 11/02/2026
Demand history
See what was read

Use actual sales or usage (issues), not purchase orders. Stockout months understate demand: correct them before pasting.

Methods
For the next steps

Your inputs stay only on this browser and are not sent anywhere. Untick to delete them.

—Forecast for the next period
—

History, fit and forecast

Actual demand, the chosen method’s one-step-ahead forecast and the future forecast with its approximate band.

Tap the chart to see the values.

Method comparison

Forecast by period

Errors of the chosen method

Actual demand minus forecast each period; bars above zero: the forecast was low.

Tap the chart to see the values.

Next step

Take this page’s results to another calculator with the fields already filled in.

Safety stock →Economic lot size →Production capacity →

The calculation report gathers this page's inputs, results, equations, assumptions and references, with fields for project, engineer and revision. The full report also includes the charts and interpretations, in color. In the print dialog, choose Save as PDF.

Method, assumptions and references

Equations used

Moving average: F_t = (y_(t−1) + … + y_(t−n)) ÷ n; weighted: weights 1, 2, …, n (largest on the most recent)
Simple exponential: L_t = α·y_t + (1 − α)·L_(t−1); F_(t+1) = L_t
Holt: L_t = α·y_t + (1 − α)(L_(t−1) + b_(t−1)); b_t = β(L_t − L_(t−1)) + (1 − β)·b_(t−1); F_(t+h) = L_t + h·b_t
Additive Holt-Winters: L_t = α(y_t − s_(t−m)) + (1 − α)(L + b); s_t = γ(y_t − L_t) + (1 − γ)s_(t−m); F = L + h·b + s; multiplicative: y_t ÷ s, y_t ÷ L_t and F = (L + h·b)·s
Errors (e = y − F): MAD = Σ|e| ÷ N; RMSE = √(Σe² ÷ N); MAPE = Σ|e ÷ y| ÷ N; TS = Σe ÷ MAD; band = F ± z·RMSE
For safety stock: demand/day = forecast average ÷ days per period; deviation/day ≈ RMSE ÷ √(days per period)

Assumptions and limits

  • Initialization: simple exponential with L = y_1; Holt with L = y_2 and b = y_2 − y_1; Holt-Winters with level and trend from the means of the first two cycles and seasonal indices from the detrended first cycle.
  • All methods are compared over the same periods (after the first seasonal cycle or the largest moving-average window), with one-step-ahead forecasts. The grid search tests α, β and γ from 0.05 to 0.95 in steps of 0.05: a good choice, not necessarily the continuous optimum.
  • The ±z·RMSE band is approximate and constant; in practice uncertainty grows with the horizon. Intermittent demand, promotions, stockouts and known future events are not handled.

Standards and technical documents

  • Hyndman, R. J.; Athanasopoulos, G. Forecasting: Principles and Practice, 3rd ed., OTexts, 2021 (exponential smoothing and Holt-Winters)
  • Winters, P. R. Forecasting sales by exponentially weighted moving averages. Management Science, 6(3), 324–342, 1960
  • Holt, C. C. Forecasting seasonals and trends by exponentially weighted moving averages, ONR Memorandum 52, 1957 (reprinted in International Journal of Forecasting, 20(1), 2004)
  • Silver, E. A.; Pyke, D. F.; Thomas, D. J. Inventory and Production Management in Supply Chains, 4th ed., CRC Press, 2017 (forecast errors and tracking signal)
  • Makridakis, S.; Wheelwright, S. C.; Hyndman, R. J. Forecasting: Methods and Applications, 3rd ed., Wiley, 1998

U.S. regulations

  • ASC 330 (Inventory): inventory built ahead of seasonal demand is measured at the lower of cost and net realizable value
  • OSHA 29 CFR 1910.176: handling and storage of materials

Frequently asked questions

What is demand forecasting for?

To plan production, purchasing, inventory and staffing before orders arrive. Statistical forecasting uses sales or usage history to project the coming months; then it should be adjusted with what sales knows (promotions, new customers, lost contracts).

Which method should I pick?

The calculator tests them all on the same stretch of history and picks the lowest error. In general: moving average and simple exponential smoothing for stable demand; Holt when there is a growth or decline trend; Holt-Winters when there is seasonality (peaks in certain months). Multiplicative when the peak grows with volume; additive when the peak is a fixed size.

How do I read the example result?

With 36 months of history the method picked is multiplicative Holt-Winters (α = 0.05; β = 0.05; γ = 0.25), with an RMSE of 52.2 units and a MAPE of 3.2%. The forecast for Jan-26 is 1,354 units, with a 90% band from about 1,268 to 1,440. The 12-month moving average errs almost three times more (RMSE of 150.9) because it does not follow the cycle.

How much history do I need?

The calculator accepts 12 to 60 periods. For Holt-Winters, at least two full cycles (24 months for yearly seasonality); with three or more the seasonal indices are more stable. Very old history can hurt if the market changed: prefer the last 3 to 5 years.

What are MAD, RMSE, MAPE and tracking signal?

MAD is the average error in units; RMSE weighs large errors more and is the basis for the band and for safety stock; MAPE is the error as a percentage of demand. The tracking signal (sum of errors ÷ MAD) shows bias: outside ±4 the forecast keeps missing in the same direction.

How should I clean the data before forecasting?

Use actual sales or issues, not purchase orders. Months with stockouts, strikes or unusual promotions distort the pattern: replace them with a typical value or the average of neighboring months. A one-off very large order from an occasional customer should also be removed.

What are the limits?

The methods assume past patterns continue. They do not foresee launches, lost customers, price changes or crises. The band uses ±z × RMSE, an approximation that in practice widens with the horizon. For intermittent demand items (many months at zero), use dedicated methods such as Croston’s.

Other calculators in this group

See all in Logistics & inventory ›