Optimizing E-Commerce Warehouse Operations

The global e-commerce sector has seen 320% growth in five years, and demand is increasing following the COVID-19 pandemic. To reduce costs and remain competitive, DHL Supply Chain ran a warehouse operations optimization project. The project involved developing a smart and robust warehouse simulation tool for testing wave picking strategies.

For a warehouse with 500,000 SKU (stock-keeping units) and 249 staff, the warehouse optimization project produced strategies that reduced the time required for order completion by 8.2% and the number of required staff by 66.

DHL Supply Chain is a division of Deutsche Post DHL Group with a global network and extensive logistics portfolio, including warehousing, transport, and value-added services.

Problem

DHL Supply Chain determined that two goals were essential to staying competitive in the growing e-commerce sector: meeting customer SLA (Service Level Agreements) and reducing operational costs.

To meet the goals, the organization chose to optimize its e-commerce operations, including the inbound activities of receive, stage, sort, and put away; and the outbound activities of pick, sort, pack, stage, and dispatch.

Jigar Panot, a specialist at DHL Supply Chain’s Global Solutions Design Center, worked on a solution for a large-scale warehouse:

• Area: 111,000 sq. meters
• Products: ~500k
• Pick zones: 12
• Daily volume: ~171k
• Staff: >3000

The project objectives were to develop a robust and smart system for testing different wave strategies and determine optimal warehouse throughput and resource utilization.

Warehouse operations and pick path optimization

Large warehouses use batch picking when collecting orders because individual order picking and cluster picking may involve an item picker spending time traveling several kilometers to fulfill an order. The aim is to find an optimal pick path.

Large DHL warehouses use batch picking with wave release. Batches group orders together and waves group batches together for periodic release. Typically, a batch contains 14 orders. Waves help coordinate shop floor activities by time — allowing other operations, such as stocking and cleaning, to take place efficiently.

On release, batch items are divided for picking by zone so that the distance between items for a picker is minimized. After all the items from a zone are picked, they are combined into whole batches in a process called staging. Complete batches move to sortation where orders are brought together in put walls before being sent for packing and dispatch.

Batch pick operations and put wall activities are grouped into several key processes:

1. Wave release — orders are grouped into batches and sent for picking.
2. Picking — the collection of items from storage in warehouse zones. Picking in zones prevents long distances between items. Zoned picking can be for items in different orders and batches.
3. Staging — all items in a batch are brought together. Batches are made up of several whole orders. When a batch is complete, it is sent to a put wall.
4. Put Wall — orders are assembled from batches and sent for packing and dispatching.

Solution: Warehouse modeling and testing

Phase one of the solution involved modeling the warehouse processes as they were and back testing the model with real-world data for calibration. This ensured model accuracy and provided a baseline to compare against order picking strategy proposals. The modeling in this phase was done using AnyLogic’s built-in Process Modeling Library. The library is specially designed to simplify and speed up the accurate capture of business systems and workflows.

After creating an accurate representation of the warehouse, the project began phase two to test different wave release strategies. The strategies were dynamic, based on metrics such as staging and put wall occupancy, the number of batches in the pipeline, and so on.

Phase two of the project had three stages:

1. Dynamic waving — where different wave strategies were created, and bottlenecks were identified.
2. Scenario analysis — to understand how a wave release strategy affects completion time and the average queue size at staging.
3. Comparative analysis — using KPI to compare the strategies, including the baseline.

During phase two, the engineers investigated the trade-offs between scenarios. These investigations helped them understand resource constraints and find an optimal balance of resources and speed. The result was a dynamic waving model that increased resource utilization and decreased cycle time when compared to the ‘as-is’ model of the then current warehouse operations setup.

Results: Optimized warehouse throughput and resource utilization

From phase one, with the as-is model, it was possible to see an opportunity to increase resource utilization and reduce task completion time because staging was not being used at, or close to, its maximum capacity. For staging, there were instances of idleness at some times and long queues at others. This can be seen in the Staging Statistics chart where the occupied time drops very low at some times and rises to a plateau at others.

Phase two of the warehouse optimization project produced a dynamic wave release model that optimized resource utilization and minimized cycle times. While put wall processing times remain roughly the same, the put wall utilization was increased by evening out the staging demand. The effect was to reduce overall order cycle time.

In comparison to the model of the original warehouse operations, the dynamic wave release model reduced order and batch cycle times, increased resource utilization rates by almost 10%, and delivered an overall completion time saving of 8.2%. The savings meant either 66 fewer staff were needed, or process completion could be two hours quicker.

The e-commerce warehouse process optimization project showed DHL Supply Chain how to deliver significant savings for their large-scale warehouse operations. The supply chain engineers used simulation modeling to accurately capture warehouse operations that could be checked and verified using real-world data from operations. Confident in the model’s behavior, the engineers then designed and tested a dynamic wave release strategy to deliver the operational gains needed to compete in modern global e-commerce.

This case study is from a presentation given by Jigar Panot, Consultant at DHL Global Supply Chain’s Global Solutions Design Center, at the AnyLogic Conference 2021:

Learn more about using AnyLogic for warehouse operations modeling and optimization.