Academic journal article The Journal of Business Forecasting

How to Use a Demand Planning System for Best Forecasting and Planning Results

Academic journal article The Journal of Business Forecasting

How to Use a Demand Planning System for Best Forecasting and Planning Results

Article excerpt

Shows step by step how to use a demand planning system for best outcomes ... the best model varies from company to company and dataset to dataset ... market research on a new product can be used as an input to determine the phase-in and phase-out periods of a newly introduced product.

The main function of a demand planning system is to provide medium- and long-term sales forecasts. The key consideration in choosing an efficient demand planning system is that it lets you forecast not only at a product level but also at product group, customer group, and regional levels, as well as for different planning horizons. The objective of this article is to show how to use a demand planning system for best results particularly in the consumer products industry.

EVOLUTION IN THE DEMAND PLANNING SYSTEM

Over time the demand planning system has gone through many changes from single client-server architecture to multi-tier client-server architecture. To understand it, one has to know client server architecture and the various parts it is composed of.

1. Client-server architecture: It is an environment in which the client machine (a PC or workstation) requests information from the supplying machine, known as the server.

2. User interface server: It displays the user's communication window for the planning system.

3. Business logic server: It manages planning administrative functions, background processing, and management's request.

4. Database server: It manages data storage in the form of database tables, rows, and other structures.

5. Web services server: It enables web services.

6. Memory resident database server: It is a server tier for memory that provides temporary storage for complex data.

As shown in Figure 1, the 2-tier clientserver architecture combines the business logic and database under one server with a separate user interface server. The problem with a 2-tier client-server or a system with all layers/tiers with a single server is that when the server runs out of space or horse power, it has to be upgraded or replaced entirely which often costs a large amount of money.

In a demand planning system, a large amount of data is organized in complex networks and this data should be available for complex computations, stored, and accessed easily. The 3-tier client server architecture has a conventional relational database management system (RDBMS) as data sources for demand planning systems. This often does not work well because complex computations bring in huge data traffic between business logic server and database server. To solve this problem, an enhancement was made to the RDBMS, enabling data structures and data flows to go through complex networks and relationships (such as data associated with product hierarchy and customer hierarchy for demand planning) so that data could be easily mapped, computed, and accessed. This gave a birth to a memory resident database whose database was based on an object-oriented method (ODBMS). This memory resident database tier was introduced between the business logic server and the database server to form a 4-tier client-server architecture, which significantly improved the system performance.

The 5 -tiered client-server architecture came into being whenthe demand planning system vendors recognized the value of enabling a demand planning system with web services. As such, a "web services" tier was added between the business logic server and the user interface server. This 5-tiered system helped to integrate web services into the demand planning system architecture.

In addition to the currently, widely used 4- and 5-tier client-server architectures, a separate data warehouse is needed to keep sales history (in units and amount) for supporting decisions, separate from the operational data and computations. Operational data include consensus forecast, bottom-up (sales) forecast, judgmental forecast, and marketing forecast. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.