How Greedy Algorithms Clean-Up American Tire Distributors Warehouses: Part 1

Take it from the analytics team of the largest tire distributor in North America: tires are difficult to store.

They come in all (well, 25,000 SKUs) different shapes, weights and sizes. The logistics around storing and moving tires involve a lot of manual work. Fully-automated warehouses using RFID chips are very helpful when it comes to the logistics inside a Distribution Center (DC), however the reality is this won’t work with tires at the moment.

Meanwhile, American Tire Distributors (ATD) operates more than 140 distribution centers that each handle thousands of tires a day. That’s a lot of moving parts, and even making slight improvements in how these distribution centers operate can lead to massive savings.

Recently, our Advanced Analytics Center of Excellence (AACoE) team used data science to improve how tires flow through our largest distribution centers. And today, we’ll let you in on how we did it.

The Idea

Our idea is to optimize our inventory in a way that would allow our trucks to be loaded faster and allow us to store inventory in a manner that coincides with their shelf life. For example, the tires that only last a few days on the shelf before being shipped out (“fast-moving”) should be moved to the front of the DC for easy loading while the tires that may last a couple weeks (“slow-moving”) are moved to the back.

The big picture here is that every truckload of tires we move is filled with different tires, and those tires have to be collected from one of our warehouses via forklift. So where do you put everything to make that process run more smoothly? With thousands of different tires, assigning the right place for the right tire is tricky. To make things even more difficult, demand for different types of tires changes many times over the year depending on the weather and other events — like after a snowstorm. The question is: How do we begin to make sense of this?

We started by looking at what we already had. The chart below shows a map of one of our warehouses where each cell represents a bin — a giant rack container that holds stacked tires.

The color is set so fast-movers are green and slow-movers are red. The bottom area labeled “DOCK” is where empty trucks arrive to be loaded with tires. As you move up this graphic, you are getting further away from where the tires are loaded and unloaded. Take a second to look at this and try to see what insights you can find. Think like a data scientist!

If you want to reduce the time spent driving around and chasing down different tires, it would make sense to have the faster-moving (greener) products closer to the dock because you need to get to them more frequently. Is that what you see here? Not really. There’s a bunch of slow-movers off to the right, and everything else is fairly random. With some smart rearranging, we can vastly improve this situation.

Out-of-the box solutions for this type of project can be extremely expensive and always require modifications. For us, the obvious choice was to just build our own from scratch. So, we built a tool that tells us where to place each tire once a truck rolls-in and tires are received.

Check back Friday to see the data and models we used to build our greedy algorithm!