Testing Warehouse Layouts Using Transportation Optimization at Land O’Lakes
When I was a kid, I was amazed on how my older brother creatively used pop cans to build a launcher that throws tennis balls (Figure 1). He did so by making some holes in the top and bottom of the cans and then taping them together. He’d make a lateral hole in the bottom can to load the fuel (alcohol) and completely open the top can where the tennis ball would sit. After igniting the alcohol with a match (and quickly moving away), the ball would get fired to heights sometimes in vicinity of 60-80 ft. (Disclaimer – the above example is for illustrative purposes only and should not be recreated or re-enacted by anyone.)
After receiving a request from a warehouse planning manager from my team at Land O’Lakes to determine the best layout for a warehouse, I eventually thought I could use the functionality of transportation optimization to compare the total distance traveled out of the current and proposed warehouse layout in the same creative way my brother used pop cans to build his launcher.
Simply put, the Warehouse Planning Manager wanted to know which of the warehouse layouts – the current one (Figure 2) or the one he was proposing – (Figure 3) was better in terms of distance traveled by an average forklift operator given a frequency of customer orders, with each customer order containing one or more order line items being requested.
The Warehouse Planning Manager built his new layout proposal based on ABC classification, meaning that each product was classified by a letter based upon commonalities or attributes of the product. He hoped that by allocating the A and B products closer to the staging area of the warehouse, the movement would be more efficient.
The measure of success was simple, the layout that yielded the least distance traveled would be likely to be the better option.
The innovative drive of using VRO to tackle this problem is that there are a lot of similarities between the order picking process that takes place in a warehouse and vehicle routing, among those:
|Order Picking||Vehicle Routing|
|Order Line||Shipment to be routed|
|Forklift / Operator constraints||Vehicle / Driver constraints|
|Staging area||Pooling Location|
Key Challenges and Solutions
Once these similarities were identified, the next hurdle becomes adapting the warehouse problem to an environment that the transportation optimization engine can understand. This entailed overcoming three main challenges:
1.Assign picking bins and a staging area as a set of coordinates (latitude, longitude) and an appropriate scale.
To do this, I leveraged the warehouse blueprints and set up a system of coordinates and proceeded to calculate the coordinates of the picking bins and staging areas (latitude, longitude) and established a scale, in this case, 1 geographic degree = 69 mi = 70 ft in warehouse. (Figure 4).
In this case, the staging area or dock coordinates are Lat -2.057, Long -0.629 (2.057W, 0.629S)
2. Make sure orders are built for their respective customer
In typical warehouse operations, orders are fulfilled by having the operator picking each item (order line) of the order. If orders for different customers arrive at the same time, we pick them separately so the orders are not mixed. To do this in transportation optimization, I made sure that Arrival Times and Due Dates for individual orders do not overlap
3.Adjust model output distance to scale when compiling results and don’t forget to adjust for distance circuitry factor that transportation optimization uses to approximate straight distance calculations to real world ones.
This illustrates an important point regarding the limitations of this approach, since VRO calculates based on straight lines between two points plus circuitry factor, the output distances are not necessarily the real ones because the warehouse layout itself places racks and bins in the warehouse area and the operator cannot go over those obstacles to get to the picking locations, so the exact route may not be reflected (Figure 5).
VRO line: straight line
Real distance: Dotted line
Despite this limitation and since we’re using the same criteria to compare both warehouse layouts, you can comfortably make decisions on which layout is likely to be better using this approach.
This are just a few use VRO case examples that we have applied at Land O’Lakes to better manage our routes and help improve overall efficiency.
Want to learn more? Click to learn how Land O’Lakes leverages optimization technology to more efficiently manage its supply chain network and transportation optimization strategy.
*Editor’s note: While the default method for distances uses Straight Line + Circuity, LLamasoft does have the option of using specific distances using the Transit Matrix table.
Erik Lopezmalo is Sr. Operations Research Analyst within the Supply Chain Planning and Analytics team at Land O’ Lakes. He and his team have leveraged Supply Chain Design techniques to uncover million-dollar savings across Land O’ Lakes three business units. Erik has 15+ years’ experience in managing Supply Chain Design projects across food, consulting and building materials industries. Erik holds a MBA from Instituto Tecnológico Autónomo de México (ITAM) and a bachelor degree in Industrial Engineering from Universidad Iberoamericana, both in Mexico City, Mexico.