Before buying a $30 tube of lipstick, most buyers will want to see how it looks on them. Beauty brands, department stores, and cosmetics websites share a common pain point—how to efficiently let potential customers experience their beauty products. In 2016, Perfect Corp. introduced its proprietary software as a service (SaaS) to address this need both online and offline, using artificial intelligence (AI) and augmented reality (AR).
Perfect Corp. offers the Perfect Console platform, which allows brands and retailers to upload their own product catalogs and integrate them with online shopping sites or in-store devices that are set up at beauty counters. The AR technology then enables shoppers to virtually try on multiple products, without the need for dedicated staff. Perfect Corp.’s customers report a 10 to 44 percent increase in basket size after implementing this technology. The company also has a business-to-consumer (B2C) operation that started in 2015 with its YouCam mobile app for enhancing personal photos. The business-to-business (B2B) operation was introduced a year later due to the success of the B2C app and has been downloaded 770 million times to date.
Global scalability was Perfect Corp.’s top priority upon launching, and its primary customers today include more than 200 global beauty brands and retailers across 67 countries including Estee Lauder, Macy’s, and Target.com. To meet this priority, the business decided to build its SaaS software on an Amazon Web Services (AWS) infrastructure. “AWS gives us the ability to scale with minimum engineering effort and zero interruptions to existing customers’ operations,” says Tony Tsai, marketing communication & PR senior manager at Perfect Corp.
On the B2C side, the YouCam app processes 2.5 billion events per day. Perfect Corp. uses Amazon EMR to process such data and collect logs to analyze and improve users’ experience. For its B2B operation, Amazon EMR facilitates the processing of reports that provide valuable business insights, such as which products are being trialed most and which in-store devices receive the most traffic. The company also uses Amazon Relational Database Service (Amazon RDS) to store vast amounts of product data and streamline reporting functions.
As the Perfect Console platform is often integrated directly into brands’ websites, application security is vital. Global customers require partners such as Perfect Corp. to prove Global Data Protection Regulation (GDPR) compliance. “Being on AWS has been very beneficial in ensuring that we meet GDPR compliance as per our customers’ requests, using built-in and managed security services,” Tsai says.
Perfect Corp. uses several services to monitor and protect its data, including AWS Key Management Service (AWS KMS), AWS Identity and Access Management (AWS IAM), and Amazon GuardDuty. The company engages independent assessors to perform annual penetration tests on its architecture and shares the results with its brand customers. It must also perform security risk and compliance reviews (SRCRs) and submit reports to customers regularly.
Perfect Corp. has worked with AWS Partner Network (APN) Premier Consulting Partner eCloudvalley since its inception. “eCloudvalley has provided great ongoing support on an application level,” Tsai says. “More recently, they have served as an architecture partner, helping us review AWS cost-optimization and performance reports to make suggestions based on utilization rates and historical data.” Based on eCloudvalley’s suggestions, Perfect Corp. purchased Amazon EC2 Reserved Instances (Amazon EC2 RI) to pre-allocate resources for anticipated surges in application usage during periods such as Christmas and New Year’s, and the company has seen about 25 percent in cost savings.
Brand customers can also take advantage of Perfect Corp.’s live streaming training service for their global beauty advisors. These brands and stores hold regular sessions to inform staff advisors about the latest beauty products. With Perfect Corp.’s AR training tool, advisors gain access to an interactive platform available from wherever they are based and can virtually try on new products using the camera in their personal devices to better understand new product features. Perfect Corp.’s live streaming training service has a latency of 3 seconds, which is within the suggested range for live streaming.
To maintain latency targets, Perfect Corp. operates out of nine AWS data centers in countries across the globe. “When our customers want to expand into more countries, their request usually comes at a short notice. AWS has allowed our teams to enable instances to serve new regions very quickly within a couple of days,” Tsai explains.
One of Perfect Corp.’s latest features is called AI Style Transfer, which allows consumers to upload pictures of makeup looks they like—from personal in-camera photos or social media—into the YouCam app. Trained AI models then analyze the image based on brand-owned data sets and map the matching makeup style. Following that, the AR function allows them to virtually try on the same makeup. Training these AI models requires engineers to constantly upload more data, as there are endless variations in skin tone and features that could affect the perceived accuracy of the service.
“We built this entire service on top of our AWS infrastructure as part of the Perfect Console platform. AWS servers are highly configurable and dynamic for on-demand tasks within an automated pipeline, and we find it efficient to use AWS for the platform we have,” Tsai says.
Since launching, Perfect Corp. has been able to maintain at least 99.58 percent service availability for customers. “AWS has actually been way above that figure, so we are very happy with how robust the service is,” Tsai says.
Having onboarded 97 customers in the past 24 months—and with an eye to expanding into the Middle East and South America—a robust infrastructure is vital. “We don’t need to worry about the scalability of our AWS infrastructure, which has been very helpful. We are now able to confidently tell our customers, ‘Yes, we can serve in that area,’” Tsai concludes.
To learn more, visit https://aws.amazon.com/big-data/datalakes-and-analytics/