The basic ingredient of any demand plan is a statistical forecast. Statistical models and resulting forecasts are the building blocks of the planning process.
Although consensus and collaboration are key ingredients of a successful demand management program, statistical forecasting is the first step to creating the baseline plan. To this end, a good processes and software technologies become important. One of the key things you look for when you prepare a Request for Proposal (RFP) is to ensure that you cover all of the modeling algorithms and techniques which are relevant to your process. This depends on your industry and your specific business model.
Forecasting techniques can be broadly classified as:
- Time Series Forecasting models consisting of exponential smoothing, Holt-Winters Multiplicative Smoothing, ARIMA models and Box-Jenkins Models, Logarithmic regression models
- Promotional Planning Models that typically use event modeling methodologies and indicator variable models
- Causal models that include a variety of Multiple Linear Regression Models and transfer function models
- Probabilistic Models that often forecast the probability of a particular event happening in the future and these include Logit, Probit, and Tobit, models
- Croston’s Models to forecast intermittent demand. Here is a link to a semi-technical explanation of Croston’s Method.