EXCLUSIVE: Target optimizes middle mile with in-house digital twin tool
Target Corp. is using proprietary modeling technology to improve the flow of goods from warehouses to stores.
In an exclusive interview with Chain Store Age, Jake Krings, VP technology, global supply chain & logistics, Target, explained how and why the discount giant developed a digital twin (interactive, AI-based virtual models that serve as digital replicas of real-life objects and/or environments) solution called Proxima to improve decision-making and performance in the “middle mile” of its supply chain, where goods travel between warehouses or from a warehouse to a “last mile” pickup point.
What type of supply chain challenge was Target facing?
Target is always trying to improve our in-stocks for customers in the store and the inventory position for our online inventory. We think about in three ways. We need to sense what's happening operationally in the supply chain, simulate our operational responses to what's happening and then ultimately, we need to solve.
There are some great capabilities around sensing and solving, but we saw the simulation capability as missing. Target has so many variables in what we can do and how we position our inventory. Being able to simulate what would happen with a high degree of accuracy allows us to continue to advance our inventory position capabilities to drive better in-stocks for our customers and ultimately, better customer experience.
What made Target select digital twin technology?
We evaluated a few different options but landed on digital twin because of the accuracy that it was able to provide. Given the many different variables in a supply chain, such as lead time, demand forecast, promotional impact, vendor schedules and transportation schedules, we needed a high degree of accuracy in driving a simulation that we can rely on.
Our digital twin solution leverages Monte Carlo analysis (editor’s note – repeated random samplings), which uses the same algorithms that financial planners use with a high degree of accuracy.
What made Target decide to develop something in-house as opposed utilizing a third-party solution?
Target didn’t find a lot of packaged solution providers being able to support the breadth of our categories, but we have the in-house capability to build these capabilities. We have patterns in place and an architecture in place that allows us to rapidly scale up capabilities like this, and so it allowed us to iterate, test a simulation, see how accurate it was, and ultimately realize that this has been a real unlock for Target, which we have continued to scale for different activations.
Why did Target focus on the middle mile?
The middle mile is ultimately how we get products to our stores, and ensuring we have the right in-stocks is mission-critical for us. The best way that we can influence this process is via our middle mile capabilities which deliver products from our regional distribution centers into our stores.
What benefits has digital twin technology delivered?
We made the decision to add a receive center node earlier this year, an upstream supply chain node from our regional distribution network. Adding a new node like that introduces a whole bunch of variability into how inventory flows and we needed to understand that.
Months ahead of launch, we leveraged Proxima to run different simulations. One of the things that we found was our trailers weren't going to be as full as we would have liked, leading to inefficiency in transportation. Because we had identified that issue literally months ahead of launching the receive center, we were able to tweak the inputs and variables and logic to ensure that we balanced that trailer inefficiency with those store in-stocks.
Doing that allowed us to get everything tuned and then predict with 98% accuracy what product was going to flow through the distribution center when we launched. Instead of like learning as we went and reacting to things days and weeks after launch, we were able to anticipate exactly what was going to happen months in advance and be well prepared.
Another example is around food, which is incredibly complex with the perishability factor. We are always trying to improve our in-stocks for food, especially with the criticality of having our menu for customers when they need to make their evening meal, so we look for ways to balance demand, lead time, perishability and transportation schedules for food.
[READ MORE: Target supports food category with specialized distribution center]
If we tweak any one of those variables, we need to know what is going to happen. With Proxima, we are able to test improvements of the algorithms to understand exactly what's going to happen. For the food categories we have activated on this new inventory positioning system, we are already seeing a 2.5% lift in those in-stocks.
How have employees responded?
They are absolutely loving it! As we make changes to inventory positions or inputs, inventory analysts have a ton of confidence in being able to understand what will happen through their network – what’s going to move to the distribution centers and what the store should anticipate receiving. They can plan ahead of time instead of reacting in real time.
Is there anything else in terms of possible future development you can discuss?
Proxima fits into Target’s ‘sense simulate solve framework’ - what are the events happening in our supply chain, how do we capture those, and then how do we prevent potential disruptions or leverage potential opportunities, and then simulate that response with capabilities like Proxima, and then ultimately drive all that ahead of time versus reacting.
Digital twin technology positions supply chains like Target’s to be much more reactive in what we're able to do compared to executing on a plan that is static. I’m excited to see the possibilities unlocked.
