What makes a demand forecast best-in-class

Forecasting is the process of estimating future demand in order to help make better-informed business decisions today. Across consumer goods brands, teams use forecasts to more intelligently allocate resources, optimize inventory management and maximize revenue.

However, companies vary widely in their forecast accuracy. According to the Institute of Business Forecasting & Planning (IBF), the retail industry averages a 30% error rate when forecasting products one month in advance, and “Best-in-Class Forecast Errors” is even a term in the IBF glossary.

They define it as “The average forecasting errors of ‘best-in-class’ companies, which may be the average of top 25 or 30 performing companies.” That’s accurate (no pun intended), but doesn’t provide any actual clues as to how to achieve a best-in-class forecast error. In this post, we’ll build on the principles of modern demand forecasting that we’ve outlined and share some key steps for a best-in-class forecasting process. These best practices will help you build a more accurate forecast to support short-term (S&OE) and long-term (S&OP) planning. 

1. Assemble the necessary inputs

Compared to predicting the future, data preparation might seem like the easy part of the process. But if you thought it’s actually harder than it sounds, you’re right.

You don’t use sell-in data or even actual sell-out to build a best-in-class forecast, but rather unconstrained demand: how much product you would sell if you never ran out of stock in any channel. To properly calculate this number that we call “true demand,” you must look at both sell-through and out-of-stocks.

Additional important inputs for a value-add forecast are channel inventory and shipment units; these data points give you a full picture of where your product currently is. Lastly, keep in mind your target goal for channel weeks of supply, taking into consideration working capital needs, storage costs, out-of-stock costs, shelf life and other factors.

An accurate and actionable forecast also requires the most up-to-date data available. If you’re trying to forecast next month’s demand using data from two months ago, your error rate is going to be higher than if you had data up to yesterday, reflecting the latest consumer trends.

Fig. 1: Sales forecast broken down into its component parts
2. Create a baseline

As you assess true unconstrained demand, understand the outliers in your data and account for them to create a baseline forecast. For example, a certain spike in sales may be the result of an influencer advertising your product. This event probably won’t happen again with regularity, so you don’t want to let it influence your baseline. The goal of a baseline forecast is essentially to serve as your “normal” forecast, not counting promotions or other special events.

The key part of your baseline is seasonality一the regular fluctuations in demand that occur quarterly, monthly, weekly or even daily. You can determine seasonality patterns by bringing in multiple years of sales data to identify trends, and of course, the more historical data you have at your disposal, the easier it will be. With new products, you will need to look to alternative data, like similar products, to extract seasonality.

3. Fine-tune with adjustments

For demand planning, you’ll need to adjust the baseline forecast to reflect internal and external changes that impact the business. The last part of best-in-class forecasting is thus sizing the impact of these events, so you can use data to adjust the forecast accordingly.

Changepoints and events to think through include:

  • Where is the product in its lifecycle? Has it just been introduced or are you sunsetting it?
  • Will any strategic business decisions have an impact on the product sales?
  • Will larger industry trends spillover into the performance of a product?
  • Are there any promotions or brand marketing activity planned for the product?

By adding these factors on top of your baseline, you’ll achieve a more realistic view of predicted demand. Each company and industry is unique, and this step really allows you to tailor your model to your specific needs.

At Alloy, we’re committed to making demand forecasting easier by automating much of the data preparation work and facilitating continual forecast updates based on the latest sales and inventory information. If you’re interested in a more in-depth look at best-in-class forecasting, check out The No-Nonsense Guide to Forecasting!

Related resources


Talking better product launch and allocation decisions with Ferrero USA

The global confectioner mitigates waste, improves service levels and controls costs by connecting digital supply chain visibility with POS analytics.

Keep reading

Say goodbye to constant supply chain firefighting: A guide

How to take an iterative approach to digital supply chain transformation with real-time alerts that motivate teams to collaborate on issue resolution

Keep reading

New white paper exposes the gap between planning and execution

Understand how gaps between systems, teams and processes are keeping you constantly firefighting and hurting your supply chain resilience

Keep reading