How Amazon Worldwide Returns & ReCommerce enhanced analytics capabilities with Amazon Q in QuickSight

Amazon’s Worldwide Returns & ReCommerce (WWRR) organization manages the lifecycle of returns, from customer returns to warehouse processing and determining the most optimal disposition solution. Our mission is to set the global standard for excellence in reverse logistics while building a future that prioritizes customers, the planet, and our pursuit of zero costs, zero defects, and zero waste.

In this post, we explore how our team transformed our analytics capabilities using Amazon QuickSight, resulting in improved performance, enhanced self-service analytics, and significant cost savings while serving over 4,000 users across business, finance, operations, and tech teams generating over 475,000 views across our organization.

The challenge

The R&R Business Intelligence (BI) team operates a comprehensive data landscape managing 6.1 PB of data. In our organization’s analytics journey, we encountered significant limitations with our previous reporting systems. The system required dedicated resources for infrastructure maintenance, struggled to handle increasing user loads, and offered minimal programmatic capabilities. Additionally, its limited integration with AWS services created operational inefficiencies, preventing us from fully taking advantage of the cloud landscape. Our use of multiple reporting tools created a fragmented reporting environment.

Performance issues were a constant concern, because limited caching capabilities forced repeated computations of frequently accessed data. As our reporting data volume grew from 1.7PB to 6.1 PB, scalability became increasingly problematic. Furthermore, business users struggled with limited self-service capabilities, including lack of natural language analytics. These limitations forced them to rely on technical experts, making data analysis both slow and resource intensive.

Solution overview

We chose QuickSight for three primary reasons: fully managed cloud infrastructure with its seamless integration with different AWS services like Amazon Redshift, Amazon Athena, Amazon Simple Storage Service (Amazon S3), and Amazon Timestream; advanced analytics capabilities with Amazon Q in QuickSight along with SPICE caching; and superior scalability combined with cost-effectiveness.

We aligned our migration to QuickSight into our top-down vision, which aims to consolidate tools and simplify the user experience across R&R. Our migration started in June 2023, with QuickSight setup and development beginning in August 2023.

The project was implemented in three phases: Phase 1, completed in December 2023, focused on migrating high impact frequently used dashboards for Return Center Operations serving 2,300 users, and Phase 2, launched in January 2024, tackled dashboards with greater complexity, including weekly business reviews that were critical for leadership decision-making. We migrated 16 weekly business review (WBR) reports containing more than 1,000 metrics and created 50 team-owned dashboards.

Throughout 2024, we significantly expanded our QuickSight usage and enabled self-service for the whole organization by introducing over 80 SPICE-based datasets and over 200 live datasets. Our users independently created 993 dashboards for daily performance tracking and monitoring. Entering 2025 with Phase 3, we began exploring natural language processing capabilities with Amazon Q in QuickSight and working towards making business metrics queryable using Amazon QuickSight Q topics.

Our Customer Returns Problem Solve dashboard, as shown in the following screenshot, exemplifies how we track customer returns processing performance in return centers.

Our Plans Shown and Selection WBR at Carrier dashboard analyzes different return options presented to customers and their selection patterns.

Technical implementation

Our QuickSight implementation integrates with multiple data sources through a sophisticated architecture. We primarily pull data from Andes (Amazon’s data lake) using Athena for SPICE-based aggregated datasets, Amazon Redshift for detailed datasets, and Amazon S3 for ad-hoc data sources. The following diagram illustrates this architecture.

Security and access management are handled through single sign-on (SSO) with ANT connectivity using a redirector service. We extensively used the QuickSight API for programmatically migrating code across alpha, beta, and prod environments; creating backups and performing restores of QuickSight assets in our QuickSight account; and orchestrating SPICE refreshes and automating paginated report exports and delivery using Apache Airflow.

Innovation with the R&R Data Assistant

One of our most innovative implementations is the R&R Data Assistant, an AI-assisted chat interface powered by large language models (LLMs) with generative BI capabilities. In 2024, we started to work on a bespoke solution, R&R Data Assist, by integrating LLMs and Amazon Q in QuickSight for self-service reporting, data exploration, metric and content discoverability, and SQL generation capabilities. Amazon Q embedding in the AI assistant enabled instant insights through natural language conversations for over 1,000 business metrics. We enabled automatic redirection to relevant Amazon Q topics based on user questions by programmatically referencing historically verified questions and using internal metadata.

We embedded key dashboards in our R&R Data Central system, which serves as a centralized gateway for BI and data needs within our organization.

We used a suite of LLMs and Amazon Q in QuickSight to create a conversational chatbot for R&R users. Currently, it features 38 tailored Amazon Q topics to support 29 business domains, enabling users to view metrics, conduct detailed analyses, and benefit from automated data narratives. Users can directly access using natural language inputs. The following screenshots show some examples.

When users discuss specific metrics in the chat window, the system automatically identifies and suggests relevant Amazon Q topics for instant insights and facilitates deeper analysis of the metric.

When asked for SQL, the chatbot provides SQL code and suggests query topics related to the requested metric.

The embedded auto-narrative feature presents the data in a descriptive text, with straightforward insights on the underlying data.

For continuous metric monitoring, users can create personalized metric decks (XBRs) using the pinboard feature.

Benefits and results

The implementation of QuickSight has delivered substantial benefits across our organization. We’ve achieved a significant reduction in visual latency from 45 seconds to 7.44 seconds (P90), and alleviated the previous 2-hour monthly downtime for maintenance. The solution is poised to save an estimated 30,000 business user hours and 2,000 technical hours annually through the integration of Amazon Q in QuickSight.

From a financial perspective, we’ve realized $700,000 in annual licensing cost savings and reduced technical bandwidth requirements by 55 person weeks per year previously needed for supporting third-party infrastructure. The system has also driven increased self-service adoption, with 55,000 self-serve analysis sessions conducted in 2024 alone.

Looking ahead

We’re continuing to expand our QuickSight implementation with plans to extend Amazon Q topic coverage to more business domains and experiment with new APIs like PredictQAResults for enhanced insights in our R&R Data Assistant chatbot. We’re also exploring scenarios with Amazon Q in QuickSight to analyze complex business problems using natural language and enabling dashboard Q&A capabilities.

Conclusion

Through our journey with QuickSight, we’ve not only addressed our initial challenges but also discovered new opportunities to enhance our data analytics capabilities. The service has become integral to our mission of setting the global standard for excellence in reverse logistics while maintaining our focus on customers, sustainability, and operational efficiency.

For additional information about Amazon Q’s integration with QuickSight, please refer to our documentation on Amazon Q in QuickSight:


About the authors

Dheer Toprani is a System Development Engineer within the Amazon Worldwide Returns and ReCommerce Data Services team. He specializes in large language models, cloud infrastructure, and scalable data systems, focusing on building intelligent solutions that enhance automation and data accessibility across Amazon’s operations. Previously, he was a Data & Machine Learning Engineer at AWS, where he worked closely with customers to develop enterprise-scale data infrastructure, including data lakes, analytics dashboards, and ETL pipelines.

Lakshdeep Vatsa is a Senior Data Engineer within the Amazon Worldwide Returns and ReCommerce Data Services team. He specializes in designing, building, and optimizing large-scale data and reporting solutions. At Amazon, he plays a key role in developing scalable data pipelines, improving data quality, and enabling actionable insights for Reverse Logistics and ReCommerce operations. He is deeply passionate about enhancing self-service experiences for users and consistently seeks opportunities to utilize generative BI capabilities to solve complex customer challenges.

Karam Muppidi is a Senior Engineering Manager at Amazon Retail, where he leads data engineering, infrastructure and analytics for the Worldwide Returns and ReCommerce organization. He has extensive experience developing enterprise-scale data architectures and governance strategies using both proprietary and native AWS services, as well as third-party tools. Previously, Karam developed big-data analytics applications and SOX compliance solutions for Amazon’s Fintech and Merchant Technologies divisions.

Sreeja Das is a Principal Engineer in the Returns & ReCommerce team at Amazon. In her 10+ years at the company, she has worked at the intersection of high-scale distributed systems in eCommerce & Payments, Enterprise services, and Generative AI innovations. In her current role, Sreeja is focusing on system and data architecture transformation to enable better traceability and self-service in Returns and ReCommerce processes. Previously, she led architecture and tech strategy of some of Amazon’s core systems including order & refund processing systems and billing systems that serve tens of trillions of customer requests everyday.


This is a companion discussion topic for the original entry at https://aws.amazon.com/blogs/business-intelligence/how-amazon-worldwide-returns-recommerce-enhanced-analytics-capabilities-with-amazon-q-in-quicksight/