Organizations that experience exponential growth can be held back if their IT infrastructure is unable to scale quickly and deliver enhanced services as their customer base expands.
One Asia-based company knows this well. The online marketplace for non-fungible tokens was seeing significant growth as public interest in NFTs increased and more consumers flocked to the site to buy and sell digital goods.
Built on top of Ethereum, its platform allows users to create, distribute and trade digital files. The company provides smart contracts, powered by blockchains, that facilitate the minting or creation of digital items without the need for users to know how to code or program. Its platform plays a key role in helping artists and creators monetize their work by enabling users to easily browse, mint, and buy and sell tokens. It also provides assurance that purchased goods are authentic.
When the company’s inventory of NFT items spiked from 7 million to 30 million in less than a year, the company realized it needed a way to scale quickly and extract critical insights from its data to push more personalized items to its users. Without such insights, its customers would find it difficult to sort through a large database of NFT items and identify items they might be interested in. This was clearly necessary, as visitors to the e-commerce site did not stay long and were uninterested in the vast array of NFT items, most of which showed little relevance to their personal interests.
A robust recommendation engine was needed to capture and analyze user interaction data, such as browsing patterns, so more relevant NFTs could be pushed to potential buyers and increase the likelihood of a sale. The marketplace operator turned to Amazon Web Services and AWS partner AspireNXT for help building that out.
Data Model That Continues to Learn
The proposed solution comprises a data-logging layer that is integrated with the existing e-commerce platform, so the data captured can be used to power a product recommendation engine running Amazon Personalize.
A fully managed machine learning service, AWS Personalize enables developers to easily deliver personalized experiences to users. For instance, Amazon Personalize can generate product recommendations based on a consumer’s preferences and browsing behavior. It also can deliver personalized email content or create targeted marketing campaigns based on customer segments.
Because it does not require developers to have machine learning experience, Amazon Personalize allows companies to get started quickly and with use cases that are optimized for their business domain or configured to their own custom resources.
Amazon Personalize helped the marketplace operator build a layer of business rules to push recommended NFT items attuned to the consumer’s personal preferences and profile. The solution generates access log records to capture item data and user interaction information, including customer demographics such as gender, age and location. It uses these logs as training data sets to power the recommendation and personalization machine learning model. The model is retrained each month as new data sets are generated, which ensures the algorithm continues to generate insights based on updated user interaction data.
Amazon Personalize also provides real-time APIs to extract relevant product recommendations based on a user’s prior interactions on the NFT e-commerce platform. In addition, it generates batch recommendations that can be used to drive email marketing campaigns that are personalized to the consumer’s profile.
Because there are no minimum or upfront fees associated with Amazon Personalize, businesses pay only for what they use. Charges depend on the company’s data processing and storage requirements, training and number of recommendation requests.
AWS Lambda was also deployed to display details of the NFT items listed for sale on the marketplace as well as facilitate data logging and extraction. The data then is temporarily published to Amazon CloudWatch log groups and periodically retrieved by AWS Lambda or an equivalent service, such as AWS Batch. Extracted data is saved in Amazon S3 (Simple Storage Service) buckets for further processing.
AWS Lambda is a serverless, event-driven compute service that enables you to run code for any application or back-end service without provisioning or managing servers. It can be triggered from more than 200 AWS services and software-as-a-service applications.
With Amazon CloudWatch, developers and engineers at the NFT marketplace can access data and actionable insights to monitor applications and resolve performance issues. CloudWatch gathers monitoring and operational data in the form of logs, metrics and events, providing a unified view of a company’s operational health as well as visibility of AWS resources, applications and services running on premises and on AWS.
Without the AWS solutions working together to deliver richer user engagement, the growing NFT e-commerce site would have struggled with a high customer churn rate and decreasing customer footprint. It also might have lost revenue opportunities, since potential buyers would not be able to view NFT items aligned with their personal preferences.
With the help of AWS and AspireNXT, the Asian marketplace improved the exploration or reach rate of NFT items by 20% since deployment. More important, it now has a platform running on AWS that is highly scalable and able to support the company’s high growth mode.
