The Amazon Worldwide Sustainability ESO team team collects data from utility invoices in order to calculate associated greenhouse gas emissions. The team faced challenges with data latency and quality due to their reliance on the third-party provider, Urjanet. To address these issues, Futuralis developed Tesseract, an in-house data processing pipeline leveraging Amazon Textract. The implementation of Tesseract led to significant improvements in data quality and latency, enabling more accurate and timely reporting on emissions metrics and sustainability goals.
The new solution streamlined invoice processing, reduced manual intervention by the Compliance Ops (C-Ops) team, and minimized potential scalability issues. Ultimately, Tesseract supports Amazon’s vision of becoming the leading destination for sustainable shopping by providing a more efficient and reliable data processing system.
Dependency on third-party (Urjanet) for data extraction and formatting
Slow and costly process with unreliable data quality
Data latency of 65 days, with possible 30-day extension due to invoice errors
Manual data preparation by Compliance Ops (C-Ops) team for MCM and AWS security review
Easy automation of invoices using buckets, queues, and Lambda functions
Clear and simple Textract APIs for synchronous and asynchronous tasks
Verbose Textract results provide valuable information
AWS Well Archetected Concepts eliminated scaling/throttling issues for invoice load and processing.
KV relationships are easier with AnalyzeExpense API
Dependency on a third-party (Urjanet) for data extraction and formatting, a slow and costly process, and the delivery of unreliable data quality.
Data latency issues in the current process result in a total of 65 days, with possible invoice errors extending an additional 30 days.
In the current state the Compliance Ops(C-Ops) team manually preps data for Modeled Change Management (MCM) and AWS security review. This includes mapping invoices from FinOps to their corresponding site and account information.
We implemented a serverless solution that enables seamless and easy code change management with AWS CI/CD best practices. The application has the ability to run experiments using AWS services and coding languages such as Python or Node. To protect from human error, separate development and product accounts were created and safeguards were added to pause workflows until human intervention approves them to proceed. Comparative metrics can be generated using easy to configure experimentation process.
The e-commerce retailer operates in North American and International markets enabling vendors, sellers, authors, musicians, filmmakers, app developers, and others to publish and sell content via its branded websites. The packaging team is responsible for creating and choosing optimal packaging which minimizes waste, lowers cost, and delights customers.
