Finding a Demand Forecast You Can Trust

For years, bad forecasting prevented companies from reaching their revenue and profit potential and caused supply managers to scramble to deal with last-minute exceptions and problems. Now, new technologies, combined with human- and process-oriented software, are making accurate forecasting possible for the first time.

To understand the future of forecasting, consider its evolution. For decades, forecasting tools relied on sales history and 50-year-old statistical models to create a starting point for a computer-generated forecast number. Forecasters or planners could choose 20 or 30 models that would attempt to predict demand for each product.

Some software let a computer suggest the best model, while others allowed users to make manual custom adjustments regarding specific product properties. Forecasting accuracy improved, but it was still way too far off, lacking the human involvement needed to make it successful.

Next came the ability for companies to see and share forecast data across the organization. The goal was to correct computer miscalculations and to account for factors such as promotions, new products, and supply constraints. Forecasting improved but remained hampered by the poor initial computer forecasts that formed the basis of the collaboration.

Today, bad forecasting remains a big problem even though companies have more data available than ever before. So, a new generation of forecasting technology is welcome news.

Advanced “smart” technologies improve accuracy remarkably by learning about demand over time. Further improvements come from innovations that enable easy human input and interaction with the forecasting process. Why do you need human input? Because no matter how good a computer system can forecast, it can’t predict unknown events such as new customer acquisitions, major customer defections, plant maintenance issues, or component part shortages.

Human Interaction

Supply managers can reduce the amount of time they spend dealing with forecast mistakes by getting more involved in the forecasting process.

Here are some steps companies can take to get the most accuracy out of new technologies and human input in demand forecasting:

Create the best computer-based starting point. Next-generation forecasting tools are interpreting increasingly complex data. Point-of-sale data, promotional histories, distributor data, and causal factors are analyzed to give companies more insight into actual demand and inventory levels. Artificial intelligence tools circumvent the decades-old reliance on statistics and dynamically model demand for products on their own.

Forecast at the lowest level needed. The explosion in data and the autonomous nature of the systems analyzing it enable companies to forecast at the very lowest level needed. For supply managers, that means seeing forecasts detailing product characteristics such as color, style, and options.

Enable cross-functional teams to collaborate easily. If everything always went according to plan, supply chains could run by computers alone. But, in the real world, the human element is critical for success. Progress in this area comes from the combination of computer-based forecasts and people, with people focusing on what the computer cannot know. With buy-in and participation by cross-functional teams, employees will be less likely to throw the forecast out the window.

Use one-number forecasting. Companies that collaborate and work from one forecast number have less opportunity to make mistakes. Also, if someone makes changes, the system can tell you why. The key to one-number forecasting is to make sure everyone can see what is going on, speak the same language, and make appropriate adjustments.

Companies should look for big gains in their abilities to forecast effectively. New tools can help reduce forecast error by 25 to 50 percent or more. With better information, companies can make better decisions, leading to greater profits and revenues.