Sales Planning

Forecasting Beyond Statistics

In the era of big data and collaborative planning, the final prediction of future demand involves much more than the statistical forecast.
In the era of big data and collaborative planning, the final prediction of future demand involves much more than the statistical forecast.

A forecast is a prediction of the future. Sometimes the future prediction can be based on what has happened in the past and sometimes it cannot. For many years statistical forecasts based on trends and seasonality have been at the heart of demand planning. But increasingly in the era of big data and collaborative planning, the final prediction of future demand involves much more than the statistical forecast.

For sure statistics have a role in forecasting. Extrapolating the demand trend with the application of seasonal and other causal factors is a good forecast starting point in many situations. Oracle Demantra does an outstanding job creating such statistical forecasts. The Demantra tool determines valid statistical forecast models to apply based on characteristics of the demand history and then tests weighted combinations of the valid models to arrive at the statistical forecast. Oracle Demantra innately uses seasonal indices and will also apply causal factors if the values for the underlying causal variables can be provided for past and future time periods.

But a lot of the time statistical forecasting is not a very good predictor of future demand, even with the use of seasonal and causal factors. Instances where this may be the case include:

  • Lots of variation in past demand or lots of intermittency in past demand
  • No past demand
  • New large customer or loss of large customer
  • Changes in marketing (price, promotions, similar products)
  • New sales channels and differential growth in existing sales channels
  • Demand is related to service intervals and units in the field

Increasingly other techniques can be used by demand planners to estimate future demand in situations where statistical forecasting alone is not sufficient:

Demand Driven Supply Chains:

Existing real time sales can be used to replace the near term forecasts. Oracle Advanced Supply Chain Planning (ASCP) has a forecast consumption feature which uses sales orders to discount the effect of near term forecasts and increase the effect of near term sales on future supply needs. There is standard integration between Oracle ASCP and Oracle Demantra to use this consumed forecast as a distinct data element visible to demand planners.

Service Intervals and Units in Operation:

For spare (service) parts planning units in operation and failure or maintenance rates can be used to directly plot the future demand of products. These features are available in Oracle Demantra when implemented with Oracle Service Parts Planning.

Bills of material can be used to derive demand for components where scheduled services involve standard operations and parts.

Collaborative Planning, Forecasting, and Replenishment:

Many customers are more than happy to provide estimates of future needs, often using industry standard data frameworks. Customer provided forecasts can be brought into Oracle Demantra as distinct data elements.

“In the era of big data, the statistical forecast can be used as one source of forecast but the final forecast can often be improved by incorporating additional information.”

Oracle Demantra also supports multiple collaborative direct forecast inputs by various stakeholders such as sales representatives or project engineers.

The demand planner can review the demonstrated past accuracy and future validity of the statistical, customer, and stakeholder forecasts to arrive at the best forecast estimate of future demand. Oracle Demantra can assist this overall review by defining a specific forecast source or weighted combination of forecast sources to derive the final demand forecast. The planner can then adjust this final forecast where appropriate.

In the era of big data the statistical forecast can be used as one source of forecast but the final forecast can often be improved by incorporating additional information. Sales person forecasts, project engineer forecasts, customer forecasts, real time sales data, and service intervals may be relevant to the final forecast determination. In instances where causal factors influencing demand can be identified in the past and predicted in the future (such as promotions) causal factors can be incorporated into the statistical forecast model.

The Demantra platform is well suited to meet these requirements. Inspirage has significant experience configuring Oracle Demantra to not only create excellent statistical forecasts but also to incorporate current demand, collaborative forecast inputs, service intervals, marketing, and demand management activities in order to assist demand planners to create the best possible final forecasts.

Please contact Inspirage to learn more.

Mark Graham

Key Contributor: Mark Graham

Mark Graham is a Senior Principal Consultant at Inspirage and has implemented Oracle Demantra and Oracle Advanced Supply Chain Planning in the oil and gas and transportation sectors, focusing on configuration to meet business requirements, BI, and end user support. Mark has also held leadership positions in the supply chain and marketing areas for two large transportation companies.