March, 2016 San Francisco, California
Demand Analytix users can now take advantage of an Auto Forecast method, which automatically chooses the most appropriate forecasting calculation method for any given time series. In addition, the method produces a 95% certainty range for the forecast, also known as a Prediction Interval. The Prediction Interval provides both a high and low range with 95% certainty that the actual number will be within the range. Based on the Prediction Interval, a forecasted product's Safety Stock and Reorder Point is calculated and presented to the user as a function of the supply lead time.
In addition to the new Auto Forecast method, Demand Analytix has added an Intermittent Forecast method with Prediction Intervals. The Intermittent method calculates the forecast for products that are sold sparsely, or with a significant number of periods of zero demand between spikes of non-zero demand. Products that typically exhibit this kind of demand behavior include accessories or options and service parts. The Intermittent Method is based on the proven Croston's Method that has become widely used at a number of discrete manufacturers.
Also, Demand Analytix has re-coded its exponential smoothing forecasting methods to generate Prediction Intervals. They include Single Exponential Smoothing, Holt Linear, Holt Exponential, Holt-Winters Additive with Seasonality and Holt-Winters Multiplicative with Seasonality. All Holt methods can be used with or without a damping factor. Parameters for all methods can be input manually or computed based on an optimization scheme that minimizes forecast errors within the in-sample range.
The methods described above are collectively grouped within the "Exponential Calcs" group that is shown within Demand Analytix's forecast pull-down menu. The Auto Forecast algorithm selects the appropriate Exponential Calcs method based on the number of samples in the time series history, the nature of the time series (e.g. whether it is intermittent or not), minimization of forecast errors within the in-sample range, and number of parameters in the forecasting method. Forecasting methods with larger numbers of parameters are "penalized" relative to those with fewer parameters to prevent overfitting.
The new Exponential Calcs are available immediately at no additional charge to Demand Analytix users. Auto Forecast is yet another example of Demand Analytix delivering upon its strategy to bring sophisticated but easy to use Business Analytics directly to product line managers to help them analyze, predict and create demand shaping product, marketing, and sales strategies.