Amazon Supply Chain and Logistics

AWS Simulation and Digital Twin to increase warehouse productivity

AWS customers can now leverage Amazon solutions to design more efficient Fulfillment Centers

This blog describes the processes and technology by which Amazon approaches new fulfillment center design, and how AWS is now externalizing this solution to customers through a new professional services offering called Warehouse Automation and Optimization (WAO).

Warehousing is a critical, yet often overlooked part of supply chain management. Across all industries and geographies, warehousing costs are dramatically higher than they should be. A McKinsey study found broad warehousing cost reduction potential across industries, including up to 13% in Automotive and up to 30% in 3PL and Pharma (McKinsey, 2019). Warehousing is increasingly becoming a strategic capability to support end consumer demands for faster fulfillment and deliveries. Delivery expectations are shrinking from weeks to days, and now to same-day. If you manage a business where warehouse throughput, productivity, and high customer service levels are key, you should think about how to use latest technology to optimize and automate your warehouse.

Process improvement, layout optimization, and automation improve warehouse performance to drive better business outcomes and improve customer service levels. Operational efficiency can further be improved with robotics, computer vision, and Internet of Things (IoT) capabilities:

  • Optimized pick-and-put away operations increase labour productivity by 40% (InterlakeMecalux, 2024).
  • 3D visualization and digital twin improve space utilization by 15% (Facility Executive, 2021).
  • The use of robotics in can reduce the average labor and manufacturing costs by 25 to 30% (Meteor Space, 2024)
  • Automated inventory tracking and cycle counting to achieve 99.9% bin accuracy (Stifted, 2024).

Automation and robotics are often viewed as a quick salvation for high labour costs and complex processes. At first glance, many companies may think that it is a quick win to simply deploy Autonomous Guided Vehicles (AGVs) or Automated Storage and Retrieval Systems (ASRS). Upon further inspection, the complexities in systems integration, process integration, and handling of negative scenarios immediately come to light. Warehouse automation and robotics are also costly capital expenditures. Many companies lack quantitative data to make informed decisions on what automation investments yield the biggest impact on throughput, labour costs, and service levels. In short, it is difficult to calculate the return on investment.

Amazon has developed a systematic approach for their fulfillment centers for capital investments in new automation technology. Amazon refers to fulfillment centers in terms of generations. A current fulfillment center under construction may be, for example, a “Gen 13” design. When Amazon decides to deploy the next series of, e.g., “Gen 14” fulfillment centers, these will have updated designs including new layouts, automation, robotics, and more. These are critical decisions that require extensive analysis and a scientific approach.

To make these decisions, Amazon uses digital modelling powered by Amazon Web Services, Inc. (AWS), deep industrial engineering experience, and advanced simulation on top of digital twins which represent the conceptual future state designs.

AWS Technology Offering for Warehouse Design and Simulation

AWS now provides a professional service offering to help AWS customers to design and optimize fulfillment centers, uncover increased efficiencies, reduce costs, and improve performance. Warehouse Automation and Optimization (WAO) combines AWS services, AWS Partner technology, and is built using Amazon fulfillment center experience. WAO consists of three phases based on Amazon’s experience in optimizing fulfillment centers.

Phase 1 – Survey and Modelling

Survey and Modeling is leveraged for customers who are looking to improve or re-imagine an existing (brownfield facility). For customers looking to design brand new (greenfield) sites, the process begins below in Phase 2 (Design).

In the Survey step, the Amazon Global Engineering and Security Services (GESS-IPS) team will physically walk and map your existing (brownfield) warehouse using a LIDAR scanner to capture an accurate 3D image of the current layout.

The LIDAR scanner captures highly accurate images of the entire warehouse and outputs the results in a point cloud format. Point clouds are digital files produced by 3D scanners which measure many points on the external surfaces of objects around them. The scanner used by Amazon captures items with a precision of approximately 3cm, including the physical layout elements like walls and outbound gates, and equipment such as conveyor belts, forklifts, and storage systems. The image below shows the result of a LIDAR scan, which captures the details of the warehouse layout and equipment placement.

LIDAR scan image

The next step is Modeling, which generates a digital twin. A digital twin is an accurate digital representation of a physical space, which includes the physical structure, digital assets, and data related to the physical space that enables simulation and scenario modeling. Digital twins are commonly categorized by four maturity levels: (1) a basic digital representation of a physical space e.g. a .pdf or CAD drawing, (2) a digital model which is spatially accurate enough to use as a foundation for simulation, (3) a digital model linked to real-time data, e.g., IoT sensors, (4) a digital model allowing bi-directional communication to control equipment or systems in the real world remotely.

This AWS offering delivers a digital twin at maturity level 2. This can later be enhanced to a level 3 or 4 digital twin by enriching it with real-time data. This can occur with or without AWS support at a future point in time.

During Modeling, the point cloud file is converted into a highly detailed and accurate 3D model. The use of a LIDAR scanner to generate the point cloud file accelerates the modeling by several weeks, improves accuracy, and decreases the labour involved. The result of the LIDAR scanning and the subsequent modelling process results in a digital twin at maturity level 2 (see Phase 3 – Simulate and Analyze). The below image depicts the 3D model base file of the warehouse in AutoDesk Construction Cloud.

3D model base file

The final step in the Modeling phase involves the creation of a digital asset library, which defines and identifies relevant individual assets inside the facility. Common examples include workstations, storage systems (e.g., pallet racking), forklift charging stations, conveyor, and other automation and material handling equipment. Each asset is given a unique ID number, and upfit with meta-data. The meta-data can include design-relevant elements such as performance data, carbon footprint, cost data, geolocational data, vendor information, and more. These assets can be re-positioned to review and test different design ideas and optimize layout options. They can also be connected to real-time data from IoT sensors (when installed) or other sources like warehouse applications to advance the digital twin to maturity level 3. This allows for real-time remote monitoring of the physical space from the digital twin.

Phase 2 – Design

AWS customers requiring entirely new facility design start at the Design phase. The Design phase leverages AWS, Amazon, and AWS partners to create cutting-edge conceptual warehouse designs. These conceptual designs represent the art of the possible for what the warehouse could look like.

The key output of the Design Phase is a future-state conceptual design package in line with the AWS customer’s vision, including best-in-class technology and automation, and compliant with the customer’s operational and financial goals. Examples of future state scenarios are the design of a new warehouse layout or potential process changes and equipment investments for existing warehouses. Our team incorporates customer constraints such as budget, timeline, security, infrastructure, safety and regulation during the Design Phase. The results are conceptual layouts in AutoCAD which include all design aspects such as detailed engineering and performance data for any proposed MHE, automation, robotics, and storage solutions. For existing facilities, these designs are quickly upfit into a new revision of the digital twin created during Modelling. For new facilities, the AutoCAD files are converted into 3D Revit files – another product from the AutoDesk Construction Cloud suite. In both cases, these serve as the foundation for future state design simulations. This is covered in detail in Phase 3 (Simulate and Analyze) below.

AWS partners with SpinnakerSCA and Amazon GESS-IPS to complete the Design Phase. The Design phase work includes a detailed analysis, engineering, and AutoCAD design for facility layout, automation, robotics, material handling equipment, operational processes, material flow, storage solutions, safety, security, and more.

We consider and review all generally available solutions that offer the best value for AWS customers in the automation and robotics phase of the design. We also support request for proposal (RFP) creation and vendor selection for our customers as an output of the design and simulation process. Typical solutions here may include, but are not limited to: ASRS, AGVs, conveyor, sortation, storage solutions, Cobots, computer vision solutions, and automated packing and label application.

Phase 3 – Simulate and Analyze

In this phase, simulation models are created using ProModel software to allow users to assess the quantified impact of new design and automation. Together with our AWS Partner, we collect the most metrics which are most important to each customer. Current state performance is converted into these measurable metrics and used as a benchmark for performance measurements when analyzing envisioned future designs. This allows deep understanding of impacts to cost, throughput, labour, equipment, and space. Additionally, we can track sustainability data from utility usage or CO2 emissions. Ultimately the model guides return on investment (ROI) analysis for different solution options.

Simulation models in ProModel are built to cover the entire warehouse end-to-end from receiving to shipping. This provides for accurate simulation of an entire operation, instead of one process at a time. This allows us to overcome potential over-estimation of performance for automation, robotics, or other stand-alone processes. For example, a vendor may advertise a certain performance level of an automated storage and retrieval system (ASRS) which assumes perfect and static inputs (putaways) and outputs (picks). In the reality, this is rarely the case as inputs and outputs are dynamic and influenced by warehouse downtime, order drop schedules, maintenance windows, and other variables. End-to-end simulation models allow us to account for these variables, thereby improving accuracy of prediction and removing bias from vendors or other stakeholders.

The image below shows the process flow for receiving in a future state simulation, i.e., after process and/or equipment changes have been implemented. You can see how workers are moving between each process step, which together with the equipment performance is the input to calculate future throughput performance.

process flow for receiving in a future state simulation

Well-constructed simulation models allow AWS to build customer input/variable dashboards to meet the customer’s requirements and key performance indicators. Detailed data is collected including actual labour efficiencies, engineered labour standards, shift data, equipment performance data, maintenance schedules, dock door availability, zone/location data, travel times and distances, and more. The models are custom built to ensure the highest possible degree of flexibility and accuracy against current state, and thereby generate confidence for conceptual design simulations. The following image provides an example on how the ‘picks per hour’ will change in an optimized, future state setup compared to the current performance.

'picks per hour’ change example

The image below provides an aggregate view across all activities and informs the user how much a potential future state scenario ‘add 1 picker and 2 push carts’ would increase usage costs. Both the performance and cost data points for a future state scenario are important to inform a data-driven business case prior to making any investment or process change decisions.

aggregate view across all activities

AWS customers can continue using the models built in their day-to-day operations. For example, to run simulations on how an unforeseen increase of inbound trucks will affect end to end throughput, and how to best reallocate labour.

The most common areas analyzed are cost, labour, space, equipment, and throughput, although there are countless other KPIs and variants to these elements. Prior to the construction of any simulation model, AWS will work to define and document the exact outputs and data which are required to support customer goals. This may include quantitative financial data to derive return on investment (ROI), payback period, total cost reduction, and similar metrics. Reports and data outputs from ProModel simulation are customized for each user, to ensure that the results allow the deepest insights and best decision making available.


Warehouse Automation and Optimization (WAO) design and simulation provides cost reduction and opportunities to improve efficiency at any warehouse, regardless of industry. Rather than the traditional benchmark-based approaches, we outline a structured approach and provide the technology capabilities to create a digital twin of a warehouse to run advanced simulations. This enables users to improve processes and make informed, data-driven investment decisions for automation.

WAO provides access to experts from AWS,, and the AWS Partner Network. This process has decreased the time it takes Amazon to build digital twins by up to 80% over traditional approaches. Through this offering, AWS provides cutting edge facility design and digital twins while bringing additional core features including simulation. AWS and Amazon have access to industry benchmarking, best practices, suppliers, and partners that nobody else does, and we leverage these to the benefit of our customers through WAO. This is the same methodology that Amazon uses for our own Fulfillment Centers, and where we have proven success. Customers retain ownership of the simulation model(s) and can leverage these for daily operational value over the long-term.

AWS also offers solutions to improve supply chain management, and introduce new capabilities like generative AI and sustainability reporting. Please visit AWS Supply Chain to learn more. Please also reach out to your AWS account manager to set up a discovery workshop with our industry experts for warehousing if you wish to explore how AWS can support you in cutting through the complexity of your warehouse operations or supply chain.

Eldon Travers

Eldon Travers

Eldon Travers is a Principal Supply Chain Consultant in the AWS Professional Services (ProServe) Global Practices Organization (GPO) based in AMER. Eldon and the GPO team are responsible for the creation and adoption of Supply Chain solutions leveraging AWS and Amazon technology and advisory. Prior to joining AWS, Eldon served as Vice President of Business IT for both Kuehne + Nagel and CEVA Logistics, and worked in the 3PL space for over 20 years supporting Supply Chain IT, Engineering, and Operations across all regions and industries.

Manuel Baeuml

Manuel Baeuml

Dr. Manuel Baeuml is the Head of AWS ProServe Supply Chain & Logistics in Asia Pacific and Japan. Manuel and his team are responsible for sharing leading digital supply chain practices and for solving our customers’ most pressing supply chain problems using AWS cloud technology and offerings. Over the last 15 years, he has had the privilege to work with industry leaders in Asia Pacific and Europe in mining, energy, retail/CPG, as well as transportation and logistics. Manuel is based in Singapore.