Kroger Instore 0015 Anderson Township March 20219450 Produce M 1

Taming the beast of multiscale retail forecasting

November 01, 2021
By: Andrew Cron, Senior Vice President, chief scientist; Joe Puchala, Director, Forecasting Center of Excellence; Kiran Ravulapati, Senior Research Scientist

When it comes to retail forecasting on an enterprise scale, the academic models fall short—particularly when that enterprise is as large as Kroger. There are thousands of Kroger stores, each stocked with tens of thousands of items, and daily, weekly and monthly decisions to make, often up to a year in advance. With trillions of data points moving through our forecasting systems daily, maintaining the consistency needed to drive smart business decisions is a colossal challenge.

Enter the 84.51° Forecasting Center of Excellence, an initiative bringing together 19 of our top engineering, science and business minds to build the best sales forecasting science possible across Kroger’s business areas, from merchandising and finance planning to supply chain and operation, labor planning and manufacturing. Here’s what makes it work:

  • The cross-functional nature of the forecasting team supercharges the path from idea to impact. Business consultants provide the framing of the problem, project managers bring in their customers’ requirements, and data scientists, statisticians and engineers work together to find solutions.
  • The team employs a variety of modeling approaches depending on the granularity of a particular forecasting need—from high-level insights on the impact of a promotion on department sales, for example, to how many units of a particular item a store can expect to sell on a given day.
  • Because forecasting at different levels with different models inevitably leads to discrepancies, the team built a mechanism, Unified Forecasting, that brings its forecasts at every level into alignment. This ensures that everyone from supply chain manager to store manager is working with the same numbers.
  • Cloud computing enables faster experimentation and time to market and makes the team’s insights shareable with customers both inside and outside the business.
  • By centralizing 84.51°s sales forecasting team, lessons learned in one business domain can be quickly applied to another.

To learn more about what Andrew, Joe and Kiran do at 84.51°, visit our Science page.

Visit our knowledge hub

See what you can learn from our latest posts.