Amazon F3 Saves 250 Hours a Week on Serverless Product Management Solution

One option would have been to build a new, unified system from scratch, but that would have been costly and time-consuming. There was a better option.

"On native AWS tools, it was straightforward for us to simplify the selection management process by launching a new system that consolidated the systems already in use by Prime Now and Fresh," says Isaiah Torain, a software development engineer for F3.

The new system, F3AST (for F3 Automation and Scaling Technologies), uses Amazon Relational Database Service (Amazon RDS) to centralize and store data that was previously stored ad-hoc in unstructured files in Amazon Simple Storage Service (Amazon S3). Incoming data is extracted by Amazon Athena, an interactive query service, and staged for loading as objects in Amazon S3.

For orchestration and business logic, F3 employs AWS Lambda, which executes code in response to changes in data, shifts in state, or user actions, and AWS Step Functions, which enables the coordination of multiple AWS services into serverless workflows. Status updates and approval notifications are transmitted to users via Amazon Simple Notification Service (Amazon SNS). To validate user tokens, the solution uses Amazon API Gateway and AWS Cognito.

On F3AST, F3's worldwide team of vendor managers are saving hundreds of hours each week. "The most important benefit of deploying on AWS is that we were able to easily build a one-stop shop for adding and managing Prime Now and Amazon Fresh products," says Thomas. "Because of the AWS solution, our vendor managers are saving about 250 hours a week because they no longer need to enter redundant information into different services."

A big part of the time savings results from the centralization data in Amazon RDS. "Storing data in Amazon RDS makes it easy for us to give vendor managers real-time visibility into all their products worldwide," says Amol Deshpande, a software development engineer for F3. "This is a big improvement over all the separate systems and interfaces they used to have to check before we deployed F3AST on native AWS tools, and—by eliminating the need for manual spreadsheet-based tracking—it also eliminates that category of manual error risk.”

The new solution supports much faster and more effective querying than before. “By using AWS Lambda and API Gateway, we can query our 8 million products in less than 30 seconds,” says Torain. “Before, users had to either query a downstream dependency that might not have been accurate or open a ticket for a developer to manually create a data aggregation job.”

Users also now have visibility into processes that were formerly opaque. "Before we built F3AST, our users' only interaction with the product publishing process was to initiate processing of new items, then manually identify any items that failed to complete processing successfully," says Thomas. "Now, because Step Functions tracks each step and automatically retries in the event of errors, users can have much more confidence that processing will conclude successfully. Amazon SNS keeps them posted along the way, with status updates and automated approval requests.”

The process of updating and managing product selection now also faces less risk of human error than before. "One frequent request from our users was for guardrails that would reduce the likelihood of mistakes and defects in product listings," says Thomas. "By using AWS Athena, Step Functions, and Amazon SNS, we can make certain fields mandatory, fill non-mandatory fields with default values, and send automated warnings about possible business-rule violations.”

Building the solution on native AWS technologies positions F3 well to keep responding quickly to customer demands. “On AWS, our team is moving and iterating faster than ever and can design even more robust systems, simply by taking advantage of service features available right off the shelf,” says Thomas. “Now that we are running on AWS, it feels like we can build just about anything our customers need.”