How I Learned to Stop Worrying and Love Supply Chains

On July 2, 1962, Sam Walton opened the first Walmart store in Rogers, Arkansas. To understand how Sam Walton ran his business it’s good to understand the man. Sam was a pilot. He loved to fly and make surprise visits to his stores. He was known for inspecting the number of cars in the Walmart parking lots as he buzzed overhead. He’d get furious if he saw an empty parking lot; obviously, something was wrong at the store. He would land at a store and go through the back door and inspect every corner to instantly get a read on how the store was run. If he detected merchandising, stocking or other shelf issues he could visualize P&L weaknesses from missed opportunities. Sam’s attention to detail, customer first approach, and operational expertise at scale have made Walmart a $600+ billion retail juggernaut with 2+ million employees.

Retail has come a long way in the last 60+ years in particular with an evolution in the expectations of the consumer. We have seen a flow from commodity one size fits all products towards a massive variety of demand. We like eating tomatoes and strawberries out of season. We like having 18 kinds of barbecue marinades to choose from. Consumer’s value choice, convenience and speed. As tastes and personalization continue their march towards total household satisfaction we are seeing social commerce that would be entirely inconceivable a couple of decades ago. With 8.1 billion people on Earth we have also shifted our attention to sustainability. End to end daily fulfillment with an eye on the future is no easy trick.??

During my time leading strategy for Walmart Data Ventures I was able to see the canon of breathtaking data available on suppliers, stores, consumer behavior and notably the underpinnings of connecting supply and demand which is supply chain data. What makes Walmart's supply chain so impressive is the sheer size, scale and depth. There are many factors at play at all times, but to start it includes 5 to 9 layers of supply chains. Even the most simple products are often touched by many carriers (ships, ports, trains, trucks). Orchestrating this network and ensuring on time and in full delivery from suppliers to stores is truly a breathtaking undertaking particularly if you consider Walmart sells roughly 160 million products. I no longer take for granted what it takes to get potato chips on to a shelf in the right four foot section of the snacks aisle.

Supply chain management processes and software are a byzantine matrix of ERP, point solutions, spreadsheets and phone calls that can be disrupted by weather, location, human error, and other idiosyncratic factors. The butterfly effect is very real. It becomes a calculus and physics problem that requires high grade, high quality data with common data formats, accurate inventory accounting, streamlined robotics and warehousing. There is no question that the benefits of scale and historical data are highly valued to compete in this often unpredictable dynamic market. You don’t see a lot of new retailers, grocery stores, and pharmacy chains showing up overnight.?What happens if you have a mosquito outbreak and the bug spray gets stocked out? This is a very real and tangible problem – SCJohnson and Google Cloud case study.

Since last November when ChatGPT was released the explosion of AI trend chasing, set off a VC chain reaction that may end up feeling like next year’s Blockchain / Web3 implosion. Removing hyperbole from the discussion, ML has seen a couple of decades of meaningful advances like the development of industrial controls, vision systems, smart refrigeration, autonomous robots, complex labor scheduling, and adaptations to meet local customer preferences. LLMs will help create and monitor the complex rules around SKU management by creating smart agents that monitor quantity and optimize labor scheduling. No matter how smart the machines get we also have to broadly apply proper change management discipline. Tracking SKUs requires stitching together carrier information by land and sea to a rail and or truck into a warehouse. Let’s keep in mind that steps in this chain like trucking are incredibly fragmented and Uncle Bob’s fleet may not even agree to be monitored and geofenced.?

What gets me personally excited about supply chains is that it is the underpinning of $ trillions of global trade and yet somehow it is so unsexy it doesn’t always attract the talent that gets drawn into a Meta or Tesla. VC activity in emergent startups may have kicked up a bit post Covid so we may see the dawn of relevant new tech land in 2030+. The industry will require a combination of technology and a mindset shift that will take time. Large corporations and SMB alike can’t risk having outages while they upgrade or change over their systems so new technology will require a step function up in utility that will likely start in the infrastructure layer before it is fully at the application layer. The good news is that LLMs, knowledge graphs, and modern tooling will allow unstructured and structured data to flow through a mix of legacy and new systems to allow for common coordination that will drive insights into actions.

So far supply chain management software has focused on visibility. Visibility is a great start, but it is a best guess. A shipment of Pepsi products to a Walmart store may be in Walmart’s ERP system for a 10 day arrival but what happens if the cargo is stuck in the Panama Canal or is in the middle of a Port with a Union worker strike? Visibility software has attracted the first wave of outside and inside capital, but predictability and true north star calibration system does not exist yet. A better predictions platform can wildly improve revenues and profits by reducing idle inventory, freeing up capital, and avoiding stockouts and nilpicks.

Current state of the art supply chain ERP is really an exception handling log instead of a proactive priority platform. What is next is the combinatory analytics platform that is a graph database with real time updates. Probability curves and managing the butterfly effect to find idiosyncratic alpha is what has been the driver of quantitative Wall Street returns for decades. If hedge funds can see patterns before they happen, maybe retail should be able to as well. The challenge is that inventory management truly has millions of features that can’t easily be sorted in a human brain or even a SQL database. Humans can handle exceptions, but can’t compute complex probabilities at scale.

For my next stage of my professional career I am focused on breaking down the supply chain industry’s data silos. Better, faster integrations across platforms will allow better visibility to answer the multi-trillion dollar question, “Where is my item?”