Academic journal article The Journal of Business Forecasting Methods & Systems

Forecasting for Short-Lived Products: Hewlett-Packard's Journey

Academic journal article The Journal of Business Forecasting Methods & Systems

Forecasting for Short-Lived Products: Hewlett-Packard's Journey

Article excerpt

Short-lived products cannot be forecasted effectively with traditional methods... demonstrates a method that works well with products with a life cycle of 9-18 months... shows the steps one should take in getting the acceptance for pilot testing for a new method.

Forecasts improve when human judgment and market data are properly combined. The life cycles of some products are too short for standard time-series forecasting methods. This article describes a method that allows forecasters to use product life cycle phases, marketing events and other known information to arrive at initial forecasts. The method proposed here not only incorporates the impact of key elements such as seasonality, price changes and promotional events but also updates the forecasts as actual market data become available. The method is most appropriate for forecasting demand for technology products with life cycles of less than one year. At Hewlett-Packard (HP) in North America, we have used it very successfully in forecasting the demand for consumer products such as ink jet printer.


Along with all its competitors in the technology arena, HP faces the challenges of high demand uncertainty, short product life cycles, steep price competition and high inventory costs. To contain the cost of inventory on the one hand, and capture all the possible sales on the other, it is important to forecast the demand of short-lived products as accurately as possible.

In the electronics industry, it is the medium-term forecasts (3-6 months) that ultimately drive the business. However, the forecasting methods in use today are often poorly suited to the consumer electronics industry, because they assume that products have a fairly long life cycle, which is often not the case in this industry. Since the speed of microprocessors is constantly increasing along with new and improved changes in functionality, standards and features, technology companies must actively plan for the end of life of their products, including capacity planning, marketing, distribution and rollover activities. Current forecasting methods do not make an assumption of the planned end-of-life scenario. Because of this shortcoming, HP's Strategic Planning and Modeling group (SPaM) developed a forecasting methodology, called the Product Life Cycle (PLC) forecasting method. The method is specifically designed to forecast products with high uncertainty, a steep obsolescence curve, and a short life cycle.


The typical electronic consumer products discussed here are those that have a life cycle ranging from 9 to 18 months. The shipment history of these products exhibits fairly well their common life-cycle characteristics. Figure 1 shows the profile of a typical product life cycle in this industry. It has a high introduction spike, gradually leveling off in the maturity phase, and then reaches the end-of-life drop-off. Although such products vary in terms of life cycle duration and the shape of curve, they all have in common:

1. Well-defined life cycle phases from introduction to maturity and then to end of life.

2. A high demand spike during the introduction phase, followed by a gradual downward leveling-off during maturity.

3. A steep end-of-life (EOL) drop-off, often caused by planned product rollovers.

The traditional forecasting methods are ill suited to such products because:

1. They require some historical data, which is often not available at the time when the forecasts have to be prepared.

2. Life cycles of technology products are generally much shorter, even shorter than many fashion products.

3. They don't consider life cycle phases such as introduction and end of life. In the technology arena, anticipating these transition phases correctly is critical to the success and profitability of a product.

4. They can't predict the end-of-life phase from the historical data of the same product. …

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


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.