Highlights
- DDI (downstream demand inference) provides an approach to approximate downstream demand for subsequent forecasting.
- DDI allows the upstream actor to infer the demand received by the downstream actor without the need for information sharing. This provides a baseline to the upstream actors for further planning.
- It has been shown experimentally that DDI generally outperforms a situation where planning is based simply on the actual orders received, and DDI is a much-preferred approach for forecasting demand in a supply chain with disjointed information.
In this article
Supply chains
A supply chain connects multiple agents to provide goods or services to the end customer.
For example, a manufacturing supply chain will normally consist of suppliers, manufacturing units, warehouses, and distributors to finally ship the product to the end customer through retailers. Ideally, the information flows from the customer to the supplier for the demand-side data and the reverse, from the supplier to the customer, for the actual operation data related to material flow. The demand data is assimilated at the customer- or the retailer-side through interpretation from historical data and actual order bookings. Under such circumstances, demand forecasting may be achieved through a statistical approach, using available methods of qualitative techniques, time series analysis, and causal models.
Demand forecasting
The demand forecasting thus achieved may be categorized as follows:
- Passive forecasting: Conducted using the historical data available with the organization
- Active forecasting: Achieved through the data collected from market analysis, campaigns, etc.
- Short-term forecasting: Generally done for a year and mainly to manage the just-in-time supply chain
- Long-term projections of demand: Considered for four years or higher, based on a combination of sales data and market research
- External forecasting: Depends on the broader economic perspective
- Internal forecasting: Depends on the organization’s plans and aspirations, and financial capability
Thus, demand forecasting facilitates critical business activities like financial planning, sales and marketing plans, raw material planning, production planning, etc. The ERP systems enhance the provision to make inventory, logistics, and pricing decisions based on the greater visibility of information. This is the backbone for various formal coordination initiatives such as collaborative planning, forecasting, and replenishment as well as subsequent enterprise-level decision-making. Supply chains are becoming hugely dependent on the importance of such visibility of bidirectional information flow.
However, there are many supply chains where information may not be shared between its actors due to several factors such as:
- The absence of information integration between several disparate systems belonging to different actors in the supply chain due to decentralization
- The absence or partial availability of information
- Security and regulatory issues (especially in global supply chains)
- and in some cases, trust and confidentiality constraints between the various actors involved
Demand management
Thus, the demand management strategy in a supply chain can be categorized as:
- FIS (Forecast Information Sharing): Supply chain links share information on consumer demand. Optimal methods and data-driven consumer demand are used for subsequent planning, ideally resorting to an ERP product.
- NIS (No Information Sharing): Supply chain links cannot share information formally, and the optimal forecasts are based simply on the actual orders received. This obviously has a delay effect and may lead to opportunity loss, inaccurate planning, and budgeting.
- DDI (Downstream Demand Inference): Supply chain links cannot share complete information and hence, the upstream link is not aware of the demand at the downstream link. However, information, partly or wholly, may exist in pockets and isolation within the actors of the supply chain. DDI allows the upstream actor to infer the demand received by the downstream actor without the need for complete information sharing. This will provide a baseline to the upstream actors for further planning.
In the case of DDI, there are two feasibility options based on the assumption that an upstream member is not aware of the demand values of the downstream member.
Option I: The upstream member can infer or is aware of the demand determinants (explained later) of the downstream member; however, the demand values of the downstream member cannot be exactly calculated for a lack of understanding of the process to determine the downstream demand values.
Option II: The upstream member cannot infer all the demand determinants of the downstream member; however, it may be possible for the demand values of the downstream member to be approximated if a certain demand process is applicable for a restricted subset of the available determinants. For example, the determinants of a downstream demand are factors like the price of goods or services acceptable to the customers, buyer’s income profile, consumer preferences, advertisements conducted, future price profile prediction (wait or buy now), etc. The upstream member is aware of a few determinants only and may adopt a process to determine the approximate demand from a restricted data set.
Taken together, the above two options imply that it is not possible for the upstream member to infer both the demand determinants and the demand process and thereby know the demand values of the downstream member accurately as possible in an ideal situation (FIS). Thus, the dependability of inferring demand processes and demand values on the sub-optimal forecasting method increases.
Due to the lack of a complete information chain, DDI is not possible with optimal forecasting methods, such as the Single Exponential Smoothing. This strategy, however, allows an upstream actor to infer the demand arriving at his formal downstream actor using the simple moving average (SMA) method in the forecast instead of using the optimal method like minimum mean squared error (MMSE). The DDI principles are applicable for an autoregressive integrated moving average (ARIMA) demand process with a non-optimal forecasting method that allows the upstream actor to improve forecasting and average inventory levels which directly lowers inventory costs.
Inference
It is thus possible to understand and approximately infer the downstream demand based on the above principles of DDI. It has been shown experimentally that DDI generally outperforms NIS and is a much-preferred approach for forecasting demand in a supply chain with disjointed information.
