Deliveroo Finds Ingredients for Success with AWS
With the flexibility of AWS, we can say we want to scale not just the entire operation up or down, but on a feature-by-feature basis. Being able to say how we want each one to respond to change in demand is pretty huge"
Founded in 2013, Deliveroo works with more than 80,000 restaurants, delivering meals in over 500 cities across Europe and Asia. The company serves a three-sided marketplace of customers, restaurants, and 60,000 delivery riders. Success means understanding the needs of all three parties and balancing them so that all are satisfied.
Deliveroo relies on Amazon Web Services (AWS) in every part of its core business: accepting orders, transmitting them to restaurants, and delivering meals to customers.
AWS’s scalable infrastructure helps Deliveroo meet the fluctuating demands of delivering food in 12 markets worldwide, while machine learning (ML) and data analytics services provide the intelligence to manage delivery logistics and offer customers personalized restaurant recommendations.
Don’t Let Technology Get in the Way of the Business
Deliveroo’s demand fluctuates, with order volume spikes at lunch and dinner times, as well as special days, such as Valentine’s Day or the Game of Thrones finale, which can see demand increase by up to 400 percent. And because Deliveroo operates in multiple countries, it deals daily with multiple, sometimes overlapping, peaks. During World Cup tournaments, for instance, order numbers spike in different countries on different days, depending on who’s playing, as well as on Muslim feast days such as Eid al-Fitr.
When Deliveroo turned to AWS in 2017, this was a key challenge. Using AWS has allowed Deliveroo to meet demand and improve service quality, reducing food delivery times by 20 percent while also cutting costs.
Dealing with the spikes is just one of its challenges though. Deliveroo turned to AWS to prepare for the future and differentiate with a great customer experience, as well. To meet demand on backend and frontend apps, it moved to a containers architecture based on Amazon’s Elastic Container Service (ECS) and Elastic Kubernetes Service (EKS) for the load balancing, monitoring, and storage it needed. It also migrated its PostgreSQL databases to Amazon Aurora, with its improved throughput and scalability. A serverless architecture using Amazon DynamoDB to publish into Managed Streaming Kafka (MSK) clusters now meets Deliveroo’s variable demand on backend and customer-facing apps.
These improved capabilities save money, too. After migrating to AWS, Deliveroo saw a 56 percent reduction in compute and database costs.
With those concerns addressed, Deliveroo has been able to look at other improvements. Using AWS’s ML capabilities, Deliveroo now does more with the customer data it gathers.
Machine Learning for Personalization and Delivery Drivers
Deliveroo uses Amazon SageMaker to recommend restaurants, products, and features to users based on their past orders. When a customer starts to browse options, serverless functions publish ML models into MSK clusters so that the user sees restaurants in the order that the customer is most likely to choose.
Efficient rider dispatch means filling customers’ orders while allowing drivers to deliver more orders per hour and earn more money. “We monitor driver satisfaction pretty closely—how drivers feel about the experience and app,” says Will Sprunt, CIO at Deliveroo. “That ties back into metrics on drivers’ likeliness to stay loyal to us, and predicts the amount of orders they will deliver in a given shift.”
Deliveroo previously used simple geo-positioning data to determine the closest driver to a restaurant. Now, thanks to AWS ML, algorithms built using Amazon SageMaker provide a sophisticated system that predicts restaurant meal prep time, as well as driver pickup and delivery times. It also considers other factors—such as predicted future orders in an area—to ensure efficient use of drivers.
Amazon SageMaker’s capabilities—including helping with hyperparameter tuning and parallelizing model training—increase the productivity of Deliveroo data scientists, who previously used time-consuming and complex proprietary tools for machine learning. SageMaker gives them a more streamlined, powerful way to create and test new ML features, reducing the time of each experiment they conduct by 4–5 hours.
Every day is a test for a growing business. But Deliveroo was truly tested when 80 percent of Europe’s restaurants closed during COVID-19 lockdowns and demand for food delivery as much as tripled.
Food delivery is a highly competitive, low-margin business at the best of times. Deliveroo needed a solution that balanced speed and customization with the need to make money, especially during the uncertainty of the pandemic lockdown. Mike Rogers, senior engineering manager of Platform Group at Deliveroo, says this is where the AWS experience shone.
“From my viewpoint, this is where the relationship really came in. We said, ‘we’re not interested in growth right now, we’re interested in cash flow.’ AWS gave us different cost-reduction models and brought in people who could think about different things. It was really valuable, when things were down, to minimize the cashflow out, and then as we came back out of lockdown, we had different challenges.”
One way Deliveroo saved money was by scaling down little-used features. For example, the personalized list of suggested restaurants—auto-generated per user—was shorter as many restaurants closed, so it consumed fewer resources. Those resources were shifted to meet increased demand for the home page.
“With the flexibility of AWS, we can say we want to scale not just the entire operation up or down, but on a feature-by-feature basis. Being able to say how we want each one to respond to change in demand is pretty huge,” says Sprunt.
That flexibility lets Deliveroo keep innovating.
The company introduced new services, such as grocery delivery, during lockdown. Afterward, it scaled up deliveries and launched a new table service that allows customers to order food and pay for meals in a restaurant without interacting with staff.
Going All In
Deliveroo has gone all in on AWS, using it for compute and database, creating and testing ML functionality, personalizing features, customer contact centers, and intelligently dispatching drivers. It now plans to migrate services currently with other providers.
“It simplifies things from an architectural point of view and means we don’t have to manage the data flow between systems managed by different vendors, and all of the headache that comes with that,” says Sprunt.
AWS also frees up development staff on the backend.
By using AWS, Deliveroo has ensured it can continue to innovate without risking operational excellence. With 650 percent year-on-year growth, Deliveroo has found a balance that works.
The key, Rogers says, is being able to focus on business, not technology. “As you move to a model where everything is infrastructure as code, and the ability to just treat service as something you can move and spin them up elsewhere, it just makes our ability to move fast and grow that much simpler.”
Deliveroo is a data-driven food-delivery platform with the stated aim of bringing the best local restaurants to customers’ homes. It uses data to balance the needs of customers, restaurants, and drivers, ensuring all are satisfied.
Benefits of AWS
- Reduced food delivery times by 20%
- Flexible response to surges
- Reduced costs
- Better personalization
- More efficient dispatch
- Improved ML training
AWS Services Used
Amazon SageMaker is a fully managed service that provides every developer and data scientist with the ability to build, train, and deploy machine learning (ML) models quickly. SageMaker removes the heavy lifting from each step of the machine learning process to make it easier to develop high quality models.
Amazon Elastic Container Service
Amazon Elastic Container Service (Amazon ECS) is a fully managed container orchestration service. Customers such as Duolingo, Samsung, GE, and Cookpad use ECS to run their most sensitive and mission critical applications because of its security, reliability, and scalability.
Amazon Elastic Kubernetes Service
Amazon Elastic Kubernetes Service (Amazon EKS) is a fully managed Kubernetes service. Customers such as Intel, Snap, Intuit, GoDaddy, and Autodesk trust EKS to run their most sensitive and mission critical applications because of its security, reliability, and scalability.
Amazon DynamoDB is a key-value and document database that delivers single-digit millisecond performance at any scale. It's a fully managed, multiregion, multimaster, durable database with built-in security, backup and restore, and in-memory caching for internet-scale applications.
Companies of all sizes across all industries are transforming their businesses every day using AWS. Learn more about Machine Learning on AWS and start your own AWS Cloud journey today.