Thought for Food

0
858
John Parkinson, Affiliate Partner, Waterstone Management Group

A couple of decades ago I was introduced to a startup that had developed a simple device for tracking the temperature history of shipments of “perishable” items like drugs and food. By collecting this data during transit, uploading it from the device to a central location on arrival, quickly analyzing the results and sending back a report, the recipient could be sure that they were accepting a shipment that had not been exposed to conditions outside of a “safe” environmental envelope. The company also sold the data back to shippers so that they could understand what was going wrong when safe conditions were exceeded. The company did pretty well for a while and then got acquired (by United Technologies as I recall) and the founder (who had become a close friend) moved on to other things.

Much later, when discussing another business altogether, my founder friend pointed me to another food related issue, exhaustively documented in the book “Food Foolish” by John Mandyck and Eric Schultz, concerning the amazing degree of waste in the global food chain, with enough food spoiling or wasted through misallocation during the “land to table” supply chain to alleviate world hunger if the various categories of waste could be eliminated. The challenges aren’t helped by the global nature of food production and the treatment of many of the “raw materials” of food as tradable and investable commodities.

So I developed an ongoing interest in looking at how the various parts of the grocery and especially the perishable food supply chain actually worked. I was privileged to be a part of the team that helped WalMart decide to revamp its grocery business back in the early 1990s, so I had some background in the challenges of selling “perishable” food items: seasonality of supply; regional taste variations; quality of items; limited shelf life; mark downs to clear inventory before sell by dates; safe disposal of unsold items….

Consumers, at least in the developed world are increasingly focused on “healthy” food – generally interpreted as “organic” and “fresh” or “ultra fresh”, which adds additional complexity to an already difficult to manage category.

No wonder food retail has razor thin margins.

As retail has invested in better technology and data collection methods, conventional data analytics have helped some – but early forecasting approaches have not been able to cope with the many variables that are in play when trying to predict daily demand for grocery items (which may be influenced by variables such as weather, advertising campaigns and supply surpluses or shortages), balancing supply and demand to prevent out of stock or over stock situations and generally complying with a complex regulatory environment for food safety.

That could be about to change.

Applying a variety of machine learning approaches to the food retailing industry is showing encouraging initial results. Rather than constantly modifying supply chain models using frequent manual inputs as conditions change, machine learning models “learn” what adjustments to make using a combination of historic data and constraint models. This continuous adjustment approach mimic the “micro control” feedback used in complex mechanical systems to optimize performance in a way that would not otherwise be possible. Retailers that are starting to use machine-learning technology for managing replenishment of perishable items have seen a variety of initial impacts—for example: reductions of up to 80 percent in out-of-stock rates for popular items, declines of more than 10 percent in write-offs and days of inventory on hand, combined with gross-margin increases of up to 9 percent – a big win in a business where overall operating margins are often in low single digits.

Although a whole new set of data science skills and operational approaches may be needed, applying machine learning approaches can be much simpler than upgrading in house inventory management and demand forecasting software – much machine learning was born and lives in the cloud and just needs a persistent data feed to work its magic. Tie the demand management results into manufacturing capacity and even to the raw materials supply chain and we might be able to move from “food Foolish” to “Food Smart” over the next decade or so. With more mouths to feed every year we may need to.

Bon Appetit…..

John Parkinson
Affiliate Partner
Waterstone Management Group

LEAVE A REPLY

Please enter your comment!
Please enter your name here