Delivery Hero helps customers navigate more than 100 million products with MongoDB Search on Atlas

INTRODUCTION

Multinational leaders in quick commerce

Delivery Hero, the world’s leading local delivery platform headquartered in Berlin, Germany, started as a food delivery service in 2011. By 2022, the company was serving up to 2.2 billion customers in over 70 countries under household names such as PedidosYa in South America, Glovo in Europe, Central Asia and Africa, foodpanda in Asia-Pacific, and talabat in the Middle East, among others. The company has evolved from a takeaway delivery service into one of the global pioneers of quick commerce, bringing small orders such as groceries, medicine, flowers, and more to customers’ front doors in under an hour, and often in 20 to 30 minutes.

“Delivery Hero operates across four continents. We serve customers under local brands with a shared mission of always creating an amazing delivery experience that’s fast and convenient,” said Andrii Hrachov, principal software engineer at Delivery Hero.

While the company has strong in-house IT skills, it relies on strategic partnerships with cloud providers to manage day-to-day maintenance, leaving staff to focus on more strategic tasks. It has a multi-cloud environment predominantly based on AWS and Google Cloud, and has a democratic attitude to adopting new technologies.

“We give staff a choice in the technologies they use. It’s a balancing act to combat bloat while empowering developers, so we’re working on a consolidation effort to narrow down our stack to fewer, more valuable partnerships,” added Hrachov.

THE CHALLENGE

Creating great experiences for 2.2 billion customers without impacting performance

Delivery Hero designed its customer-facing app from scratch. It empowers customers to browse shops, filter search results, and place orders in just a few taps, but creating the best possible user experience can be challenging for companies like Delivery Hero. Slow loading, poor response times, or difficulty finding what they’re looking for could cause customers to turn to a competitor.

The app uses a low-latency API to access product information for more than 100 million items and needs to be able to handle 10-12,000 requests per second. This includes random access queries, sorting by category, and the requirement to support personalization without exceeding 30-40 milliseconds of latency, which is important for both daily operations and to successfully run promotions.

“We need to meet the expectations of our B2B customers and the people at home ordering their goods. We want to deliver a variety of everyday items to customers' homes, for example, we need that item to appear near the top when the customer is browsing for a fresh salad,” explained Hrachov. “But that same customer also needs to be able to filter search results by their own criteria — price, rating, or distance from their location, for example — without impacting app performance.”

To meet these requirements, Delivery Hero needed a flexible database with indexing capabilities that could handle various access patterns, however its existing SQL database would have resulted in 20-25 indexes per country and impacted performance.

THE SOLUTION

A scalable, high-performing database platform

Hrachov has been a fan of MongoDB for more than 10 years and was delighted when Delivery Hero replaced its SQL database with MongoDB Atlas on Google Cloud in 2021. “MongoDB Atlas lowers management overheads, it’s great at handling large volumes of data, and lots of people on our team were familiar with it,” he explained. “As a general-purpose database, it fits nearly all use cases and offers flexibility and horizontal scaling. It was the natural choice.”

To handle random access requests at scale, the team considered using various options, but were apprehensive to add another database system for them to maintain. When MongoDB Search on Atlas was announced, Delivery Hero realized it provided the same functionality without adding another vendor to its tech stack. The flexibility of the solution also removed a lot of the restrictions developers encounter with other technologies, which aligns with the ethos of empowering them with the best tools for the job.

“MongoDB Search on Atlas was a game changer. We ran a proof of concept and discovered how easy it is to use. We can index in one click, and because it’s a feature of MongoDB Atlas, we know data is always up-to-date and accurate,” recalled Hrachov. “We don’t need to maintain additional infrastructure as the new features were implemented in less than two weeks, compared to the three months it would have taken to roll out another solution.”

This makes it much easier for Delivery Hero to support shops with promotions and personalization. And because it’s low maintenance, the team has plenty of time to refine the customer experience like improving the sorting and retrieving functionality, for example.

THE RESULTS

Making more than 100 million products searchable without latency issues

When a customer browses their local delivery app, they enjoy fast, seamless access to a range of shops. They can view and search menus, filter by location or popularity, and see the latest personalized deals clearly displayed on the page.

With MongoDB, Delivery Hero is confident it can scale this experience to more customers, more menu items, and more regions. “We have more than 100 million products in our catalog. We use sharding to spread that data across clusters and ensure high performance,” said Hrachov. “My team is responsible for serving category data, but because we can operate so much faster than colleagues who don’t have MongoDB, everyone comes to us when they need data.”

This speed and agility comes down to having a flexible database with the right functionality as a managed service. It’s more cost effective, takes less effort to maintain, and enables the company to move quickly when it needs to respond to new or changing customer requirements.

“We can implement things that would take six months on other platforms in just a few days. That means we can experiment with different sorting strategies and earn more commission,” said Hrachov. “The order that products appear has a huge impact on purchasing decisions, the more ways we have to sort and organize that data, the greater the impact will be for everyone. That’s how MongoDB helps us retain our competitive edge.”