Journal of Business Forecasting Winter 08-09 Special Issue

Journal of Business Forecasting

Many consumer product goods (CPG) companies use the Point of Sales (POS) or consumption data in their forecasting processes; Heinz is not an exception. We use the data in a variety of ways, not only for analytical purposes, but also as a communication bridge with various cross-functional partners, including Marketing and Sales, as well as various Supply Chain departments. Heinz uses syndicated data from Nielsen and IRI, depending on the data availability for various channels, and Retail Link for Wal-Mart and Sam’s Club.


We rely on the POS data to tell us the health of our businesses, the major categories in which we do business, and our business’s competitors and threats. Some of the benefits of using information from the POS data include the following.

1. They enable us to balance either too much optimism or pessimism that can come from shipment trends alone.

2. They provide more stable and consistent patterns that can be easily used, as well as helps us to identify changes in the trend over time. Shipment data are not true indicators of a change in the trend because their data bounce around.

3. They allow us to do a “sanity check” on the final forecast numbers by using trade inventory (inventory by store) in the data stream.

Providing objective third-party perspective on what’s going on in the marketplace is one of our core responsibilities as forecasters/demand planners. Therefore, we focus on training all our analysts in pulling and analyzing POS data extensively in order to understand the past performances of our brands, as well as the implications of such information on future demand. We also have annual training sessions with Nielsen that show how to use analysis trees and metrics to identify potential scenarios, such as why the base consumption of a particular brand might be growing or declining. Such a trend might be due to a change in the base price, distribution changes, competitive activities, media or marketing events, new or discontinued items, and/or unit sales off the shelf daily at the full price. Since the data exist at multiple levels and layers, we begin at the total U.S. markets level and then dig into deeper levels, such as regions and specific accounts.

Learning such analytical skills is important when deciding on a course of action across various areas of the company. Forecasting is all about understanding the past and predicting the future based on certain identifiable patterns. Knowing how consumption behaves based on our internal plans, budgets, and executions is a great place to start understanding the source of forecast error and identifying ways to improve future results.


The main benefit of using POS data is the richness of information that we find when we analyze such data. Heinz uses this data in number of ways. For Marketing, consumption information is the language of choice when it is communicated to key internal and external stakeholders. Some of these metrics may include volume and/or dollar share based on total consumption, growth or decline in consumption, incremental consumption driven by various trade promotions, pricing, distribution, and unit sales off the shelf. Marketing also uses actual consumption data to project future demand by making some assumptions about the category (Frozen Potatoes), competition (McCain Potatoes), and market share (Dollar or Unit Volume Sales). Sales also uses consumption data to analyze past promotions to determine the lift stemming from such things as discount, display, and ad support, as well as to develop most effective promotional plans for the future.

Since the use of consumption data has always been part of our culture and the basis of our forecasting process, we tied it to the future shipment plans. In fact, from Day 1, “shipping to consumption” has been one of our most important principles. Thus, POS data assumed the top role in determining the future demand. However, Marketing used the top-down approach (starting from category and brand share assumptions of products such as Heinz Ketchup), and then drove down the forecasts to an item level by using various inputs from Sales. Forecasting analysts take these assumptions and inputs and incorporate trade inventory assumptions at the major accounts to derive the final shipment forecasts.

Even after the implementation of demand planning software (Manugistics/JDA), which addresses more granular, short-term supply chain needs, the consumption-based forecasting models—developed more than six years ago—still exist and are in use. These models primarily provide the common platform to talk about the future demand. Intuitively, everyone understands the importance of selling Heinz products off the shelf to consumers. All of us in the organization, whether it’s Marketing, Sales, Supply Chain, or Operations, recognize the need to work together to increase the final sale to the ultimate consumers. As such, the common language we all speak is consumption.


Many at Heinz have asked this question: “When can we move away from using consumption in forecasting future demand?” So far, we have not found a satisfactory answer for a variety of reasons including the richness in the available POS data and their analytical power to answer various questions. Not only that, we continue to use it as a basis for cross-functional communication for discussing the status of our brands, categories, and competitiveness, as well as the status of the overall U.S. market, regions, and major accounts.

The extent to which consumption data are used in the demand planning process may vary from company to company, but one thing is certain: Consumption data enrich our perspectives and provide guidance as to what kind of volume can be expected in the future. As forecasters, we’re required to use data streams from various sources to get to the most accurate picture of future demand. Learning to analyze consumption data effectively, draw conclusions, and provide insights is very important if we are doing our job well. In so doing, we do wind up applying judgment and art to the science of forecasting.cover_w08-09