Demand Forecasting

Qualitative Forecast Methods: Forecasts that use subjective inputs

Trends

Seasonality

Cycles: Wavelike variations lasting longer than a year

Random Variation: Residual variation that remains after all other behaviors are accounted for

Irregular Variation: Don’t reflect typical behavior, but rather a specific event

Near Term Forecasting

  • Naive Forecast: Forecast utilizing a single previous value of a time series as the basis

    • Forecast for a time period is equal to the previous time period’s value when stable

    • Forecast using previous period’s demand plus the difference in the last 2 periods’ demand

    • Use when

      • Time series is stable

      • Trend

      • Seasonality and you can match to the prior year’s seasonal value

  • Moving Average

    • As new data becomes available, update the data with the newest information to continue producing a forecast

  • Exponential Smoothing

    • Ft=Ft−1+α(At−1−Ft−1)F_{t}=F_{t-1}+\alpha\left(A_{t-1}-F_{t-1}\right)

  • Weighted Moving Average: Most recent values in a time series are given more weight in computing a forecast

Forecast Accuracy

  • MAD = ∑∣Actual−Forecast∣n\frac{\sum\left\vert Actual-Forecast\right\vert}{n} → weights all errors evenly

  • MSE = ∑∣Actual−Forecast∣2n−1\frac{\sum\left\vert Actual-Forecast\right\vert^2}{n-1} → weights errors more when they are larger

  • MAPE = MAD/Average Actual Demand