It’s no secret that getting the right products to the right stores during extreme weather events—including in the recent hurricanes Harvey, Irma, and Maria—poses severe supply chain and merchandising challenges. Further, the stakes for solving the problem, from both a humanitarian and revenue perspective, are very high. How do you predict the extreme spikes in demand, at what quantities, on which product categories, for which geographic areas, and know when demand will hit relative to an event? Further, how do you easily funnel this new demand into an organization’s existing and rigid supply chain process? Even the best automated systems cannot react fast enough to these changing conditions, causing retailers to revert to pen and paper, phone and email, and a lot of back-of-the-napkin math to get products to stores quickly.
The Unanswered Questions
When facing these events, retailers ask themselves:
- How can we predict extreme spikes in demand?
- Which quantities are needed to accommodate this demand?
- Which product categories will be impacted?
- When will demand spike?
- Which geographic areas will be impacted?
Empty shelves don’t help anybody. Before and after a major natural disaster, people need to purchase preventive and repair items. When a store doesn’t have what is needed, customers are either left in despair or go to a competitor. One company we worked with wanted to ensure that customers could get the products they needed, but relied only on memory and gut instinct to ship products to stores. Due to the incomplete information they had on hand, the client sent too many (or not enough products), taking up excess space or forgoing sales. Further, their process required them to re-ship products repeatedly to get them to the right location. Due to this inefficiency, they ultimately lost money on every transaction. Despite increased overall sales, the organization lost millions from their bottom-line in their effort to help communities re-build.
While we believe that the well-being and recovery of communities should always take precedence over the bottom-line, data & analytics presents a compelling solve to the emergency response supply chain challenge—ensuring the delivery of products to those in need, while upholding product profitability.
In our work with this client, we designed a machine learning system that could efficiently meet customer demand. To effectively deploy this system, we merged three components:
- A rich data history of past product supply and demand during similar weather events
- Dozens of weather metrics measured multiple times per day, going back a decade, for the entire continental U.S. and Caribbean
- A platform that could run machine learning models and crunch through terabytes of data with ease
We then focused on understanding the full business context and built a tool to automatically collect data and make demand predictions in real-time based on weather events. Our team brought this together through several inputs: Over 74 billion weather data point served from NOAA, flowing through a distributed computing framework called Kubernetes, ten years of sales and inventory history in a distributed Enterprise Data Warehouse tool called Google BigQuery, and a deep learning algorithm developed in the Google Machine Learning Engine.
Operating with a bias for execution, we’ve focused on producing recommendations in a format that the inventory and replenishment team can quickly review and import into their ordering systems to cut an emergency Purchase Order, getting the trucks packed and rolling efficiently to those in need. And knowing that dashboards should always be designed with the business problem in mind, our dashboards provide visibility into the recommendations and order execution progress.
The Power of Data & Analytics
Natural disasters are tragic. They strip families and communities of many of the resources needed to survive. At North Highland, we believe we can fight this challenge with a new resource: data. Through analytics, data yields the insight needed to solve supply chain challenges and ensure necessary products are delivered to those who need it, when they need it—all while upholding profitability. While it’s not possible to control the weather, there’s power in analytics in defining the path forward during times of crisis.
For more details, watch the North Highland team present at Google Next 18 in San Francisco.