Retail Analytics Systems Still Need a Human Touch

As more data becomes available in the grocery and retail environment, organizations are re-thinking their supply chain processes to drive increased automation, improved performance, and increased inventory turns.  They are also re-thinking the design of their supply chains, thinking about how much of their product goes through distribution centers, distributed to stores, or shipped directly from suppliers  to store locations.

I recently completed a study for the International Institute of Analytics,   to provide insights into the following questions faced by retailers confronted with the issue of how to exploit analytics to improve supply chain performance.  In doing so, I had an opportunity to reconnect with two of my former students, Jeff Behrens, who is now a demand analyst at Lowes, and Kuru Subramanian (one of my first Research Associates I worked with at NC State!) who is now a consultant at Wipro, and who worked for years as a demand forecasting analyst at Tesco’s.  I also interviewed executives at Key Foods in New York, and P&G in Europe.

All of the companies we benchmarked have a different systems and analytics capability.  In general, all retail chains use a very simplistic approach, which generally consists of looking at prior year’s sales, and running historical promotions and coupon sales based on prior year.  Only recently have organizations gotten into complex solutions and tools to forecast grocery. Grocery is an inherently relatively stable category with little fluctuation in demand, except for major holidays.  In most cases, historical sales from the previous year for that week are used to place initial orders, and integrate with POS systems with manual weekly tracking of actual sales.

In general, I discovered some important insights in reviewing these systems:

  1. Forecasting in the retail sector is still a combination of human experience and analytical systems.  There is still some level of human review and scanning required to review/revise computer generated forecasts, due to the fickleness, weather-related, and changing tastes associated with consumer buying patterns in the grocery and retail channels.  Historical seasonal factors associated with holidays and events should be reviewed annually and planned for to optimize supply chain category plans in retail sectors.
  2. Data collected form Point of Sale systems can be compiled and consolidated into categories to detect consumers’ changing tastes and needs, with but should ideally be complemented with vendor insights and reviews to understand what factors may be at play in forecasts.
  3. Automated replenishment systems must be complemented by store-level management decisions for seasonal items, items with short shelf lives, and space planning factors that may impact product presentations and order quantities.
  4. Retail forecasting and replenishment systems continue to evolve, and are being complemented by third party consumer preference panel analytical services, as well as multi-tier collaborative planning systems.