Delivery Hero faced a challenge in May 2022, when it received a request from the United Arab Emirates (UAE) that would require it to update its verification of rider identities within a 6-week compliance deadline. It was already planning to implement an automated solution for facial verification in 3 months to reduce its delivery rider fraud—for example, where multiple individuals would share an account to defraud Delivery Hero, avoid labor and immigration laws, or achieve a financial advantage over other users. It now needed to accelerate the process dramatically. In addition, Delivery Hero wanted to streamline the collaboration between its fraud detection and compliance teams for faster remediation in cases of fraud.
Delivery Hero began using Amazon Web Services (AWS) in 2011 and, in response to the new requirement, decided to build a solution using Amazon Rekognition, which automates and lowers the cost of image recognition and video analysis with machine learning. Delivery Hero met the 6-week deadline and reduced fraudulent facial verification attempts from 80 percent to 0.1 percent, which constitutes 99 percent, while saving around €250,000 per month on account fraud and preventing €430,000 per month from civil and criminal liability.
Opportunity | Meeting UAE Regulations Using Amazon Rekognition
Based in Berlin, Germany, Delivery Hero operates on four continents worldwide, working with a network of brands to deliver goods such as groceries, flowers, and medicine to over 70 countries. It started as a food delivery business with a bicycle fleet of riders, and the term stayed as it expanded.
Initially, local fleet managers had to manually verify submitted rider photos, and if a photo didn’t match, they would then manually issue a warning or suspension. This impacted the effectiveness and accuracy of the solution. Of the photos submitted to Delivery Hero, 80 percent were faceless, and 4 percent showed faces that didn’t match the rider’s identity. It lacked a quick process to take account actions, such as suspension, based on poor photo submissions. To meet the deadline shared by the UAE government, Delivery Hero had 6 weeks to implement an automated solution to reduce fraud and meet compliance needs for facial verification.
To power its business, Delivery Hero had been using AWS services including Amazon Simple Storage Service (Amazon S3), object storage built to retrieve any amount of data from anywhere, to store the photos collected for identity verification of its riders. Therefore, Delivery Hero chose AWS for the new solution because it already trusted AWS to store its riders’ personal data. Because Delivery Hero needed to accurately and efficiently reduce fraud on the more than 50,000 images that were submitted daily, it wanted to automate its facial verification solution.
Amazon Rekognition is especially nice because it’s plug and play. All the time we dedicated to this project was spent 100% on development.”
Software Engineering Manager, Delivery Hero SE
Solution | Reducing Fraud Attempts and Saving Money Using Amazon Rekognition
Delivery Hero’s own facial-verification solution was developed by a team of four engineers, consisting of front-end and backend engineers. The team successfully implemented Amazon Rekognition in 4 weeks. With 2 weeks to spare before the 6-week deadline, Delivery Hero focused on seamless internal integration and rigorous testing to improve user experience. And the efforts paid off—Delivery Hero met the swiftly approaching deadline. “We had biweekly calls talking to the AWS team that were extremely helpful,” says Roberto Ferraz, software engineering manager at Delivery Hero. “AWS reviewed our proposal and recommended improvements on it, which definitely helped us deliver on time.”
With Delivery Hero’s solution, riders can start their shifts quickly and easily by submitting a selfie through the mobile app. Using Amazon Rekognition, Delivery Hero automatically compares the photo with the rider’s stored data in up to 0.7 seconds—a significant improvement over the previous manual verification process, which could take a few days. With seven million photos processed each month, these time savings add up quickly. Thanks to the solution’s intelligent design, riders are prevented from starting their shift if they submit a photo without a face—reducing faceless photo submissions from 80 percent to 0.1 percent. “We are making sure that we are compliant with laws and regulations everywhere, which impacts the safety of our riders and our customers,” says Ferraz. “Customers can be sure that the person coming to their door is someone that we hired and that we trust.”
If a photo does not match a rider’s stored data, the facial-verification solution seamlessly notifies the compliance service, which then automates remedial actions based on country-specific regulations. Delivery Hero’s compliance service automatically applies consequences for breaking the rules based on configurations created by local teams. With this streamlined process, the rider can receive a notification of suspension or termination, prioritizing the safety and security of riders and customers. “Amazon Rekognition is especially nice because it’s plug and play,” says Ferraz. “All the time that we dedicated to this project was spent 100 percent on development. There was no need to manage infrastructure or configurations.”
Using the facial-verification solution, Delivery Hero is reducing the impact of fraudulent behaviors and saving money, an estimated €8 million per year. By meeting the 6-week deadline, it avoided fines of €13,000 per rider caught by the UAE with documents not matching their account profiles, which could have resulted in €430,000 per month. And because Delivery Hero can prevent account-sharing fraud, it is saving €250,000 per month in two countries.
Outcome | Creating Innovative Solutions Using AWS
The success and speed of implementation of this identity-verification solution raised developers’ interest in using more AWS artificial intelligence services within Delivery Hero. It plans to improve the automated facial-verification solution by using Amazon Rekognition Face Liveness, which verifies that only real users, not bad actors using spoofs, can access services. This will discourage riders from taking photos of photos to bypass selfie verification. It also wants to make the verification process smarter by reducing the frequency of photo submissions for successfully verified riders and making the requests of the verification random. It also plans to include helmet detection to contribute to its goal of a safe environment and will continue to use AWS to improve its business.
“Using AWS is more than a vendor-to-customer relationship,” says Ferraz. “We work alongside the AWS team to discuss problems and how to resolve them using AWS solutions.”
About Delivery Hero
Delivery Hero is a local delivery service based in Berlin, Germany, working with a global network of brands. It delivers goods such as prepared meals, groceries, flowers, and medicine to its customers’ doors in 70 countries across four continents.
AWS Services Used
Amazon Simple Storage Service (Amazon S3) is an object storage service offering industry-leading scalability, data availability, security, and performance.
Automate and lower the cost of your image recognition and video analysis with machine learning.
Learn more »
Organizations of all sizes across all industries are transforming their businesses and delivering on their missions every day using AWS. Contact our experts and start your own AWS journey today.