The aim of a new study by Terra Technology was to develop comparative metrics across industries for benchmarking forecast accuracy, according to Terra chief executive officer Robert Byrne. The company drew on data from the demand and sensing systems of nine multinational suppliers of consumer packaged goods. It included all of their North America-derived information on items and locations over a period of two years.
The conclusions were anything but heartening. "Forecasting remains very challenging," says Byrne, "even in an industry where some products are 100 years old." The average error was 48 percent on a weekly basis, with "best" performers cutting that number to 42 percent.
Product and market volatility make it extremely difficult to develop accurate forecasts. "I like to say that marketing is an entire department hired to make sure that history doesn't repeat," says Byrne.
Demand sensing is a valuable exercise that differs sharply from conventional forecasting, Byrne says. Traditional demand planning yields an average for sales, failing to reflect what's occurring in the business at present. Demand sensing adds channel inventory and items at point of sale - "whatever's relevant now" - and fine-tunes the forecast.
Byrne offers a number of tips on how companies can improve their forecast accuracy. Much of forecasting today focuses on the short term, guaranteeing that long-range results won't be correct. Demand sensing "automates a lot of that and frees people up," he says. "They can spend more time on what's really going to happen."
Byrne also urges companies to cooperate with downstream partners and upstream suppliers. Point-of-sale data, he says, is less valuable than information on inventory in the channel, including warehouses.
