Academic journal article The Journal of Business Forecasting Methods & Systems

Improving Sales Forecasts by Improving the Input Data

Academic journal article The Journal of Business Forecasting Methods & Systems

Improving Sales Forecasts by Improving the Input Data

Article excerpt

Although often ignored, data quality may be one of the biggest stumbling blocks to creating accurate sales forecasts ... evaluation of sales and governmental data may reveal both forecasting flaws and forecasting solution ... keeping a log of product changes, competitors' activities, promotions, and internal politics will prove helpful in future forecasting efforts.

For many years sales forecasters have looked at improving the accuracy of their forecasts by improving the models used to make forecasts. The other way to improve the accuracy of forecasting is to improve the quality of the data used to forecast.

The November 7, 1994 Business Week cover story illustrates the problems with forecasting with bad data. The article is entitled "The Real Truth About the Economy. Are Government Statistics So Much Pulp Fiction? Take A Look." The article focuses on the errors in the data that the government provides and how the data do not represent what they intend to represent. At the beginning of the article it states, "The economic statistics that the government issues every week should come with a warning sticker: User Beware... the government is pumping out a stream of statistics that are nothing but myths and misinformation." Problems also surfaced during the 1990 census, in spite of a large effort to gather accurate data, some states and organizations sued the U.S. Census Bureau over the accuracy in the reported statistics.

The Manager's Journal column in the January 9, 1995 Wall Street Journal reports that, "The flaw in Intel's Pentium chip underscores an already pervasive business problem - bad data. Faulty information generates poor decisions, which in turn can cost a company its market. Most often, however, hardware isn't to blame. Consider how much information we gobble up without double-checking its origin or validity." Apparently, gathering and identifying accurate data is a problem for governments and businesses.

The purpose of the article is to highlight several sources of data error and to provide ideas for helping forecasters recognize and adjust for inaccurate data. We first discuss pitfalls associated with company-generated sales data, then turn to problems with government data that may affect sales forecasts.


There are many possible sources of data error that complicate sales forecasting. Most companies use a time series approach to forecast sales for production planning. The philosophical basis of time series forecasting is that the measured values that constitute the series are generated by an underlying process that remains stationary over time. As a result, past data patterns can be thrown forward to produce a forecast of the time series. All time series models, such as exponential smoothing or BoxJenkins, use this notion. The primary difference in time series models is in the manner in which the past data are statistically partitioned, aggregated, and weighted prior to the projection and the method of projecting. Obviously, time series forecasting depends heavily on gathering accurate past data that reflect the underlying process that generates the data.


A time series can be broken down into components or patterns. Most researchers refer to four components in time series, that is (1) trend, (2) seasonality, (3) cycles, and (4) noise or residuals. In many time series, particularly those which measure a product's sales, a fifth component is present which are called "outliers."

The fifth component, outliers, can be the result of marketing mix expenditures such as a price promotion, consumer promotion, or extraordinary advertisement that have an effect on only a few of the data points. One important goal of marketing managers is to create positive sales outliers through manipulating the marketing mix. Identifying and adjusting for marketing mix outliers is a recurring and difficult process in sales data analysis that is critical to sales forecasting. …

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.