MongoDB Atlas powers global engagement for HBO Max

THE CHALLENGE

Managing global recommendations at scale

As a global leader in premium TV streaming services, Warner Bros. Discovery (WBD) aims to deliver a flawless, personalized experience to the 128 million subscribers that use its HBO Max platform.

WBD needs to process immense volumes of user and content data across 155 territories and 15 distinct device types, with users typically maintaining between three and five individual profiles per account.

“We handle content for users of iPhones, Android phones, Apple TV, Xbox, PlayStation, whatever,” explained Mohammed Bakri, Senior Engineering Manager at Warner Bros. Discovery. “Overall, we manage between five and seven terabytes of active data every single day.”

Delivering timely, highly personalized push notifications to HBO Max subscribers involves capturing viewership insights and marrying them with live content updates on a 24/7 rolling basis. Critically, HBO Max needs to do this at times when viewers are most likely to be receptive to the content delivered.

“We’re not here to spam users,” said Bakri. “We are really keen to discover and act on what really matters for them.”

WBD initially chose DocumentDB for the infrastructure that handles these push notifications. It worked well on a small scale, but as HBO Max expanded, Bakri’s team began to suffer from severe scalability constraints and rigid operating models.

“Pricing was also a pain point,” said Bakri. “A single region can have millions of read-write operations each day, so a standard pay-as-you-go model wasn’t for us.”

To further complicate matters, the metadata associated with streaming content changes daily; new attributes and fields are constantly introduced, making traditional SQL schemas and rigid data architectures impossible to maintain. WBD needed a highly scalable, flexible, and cost-effective database that could keep up with its ambitious demands.

“We were growing and needed to scale,” Bakri added. “We also needed managed scaling at the storage and cluster levels. And the more our data grew, the more power we needed. That wasn’t easy.”

OUR SOLUTION

A fully managed, flexible data architecture

MongoDB Atlas’ flexible document model was the perfect fit for WBD’s canonical data model, allowing its engineering team to seamlessly accommodate daily additions of new content fields without being constrained by rigid schemas.

WBD adopted a ‘medallion’ architecture on MongoDB Atlas to categorize data into Bronze (raw viewership events), Silver (aggregated user insights and sports affinity), and Gold (final recommendations and signals). This allows it to process two main data streams: viewership and content. While viewership data involves massive volumes of simple models, the content library involves mid-sized data with highly complex modeling for over 35 million active documents and images.

“Our complexity is in the content. Imagine that the data we handle every day was in single messages; new fields need to be added every time,” added Bakri. “That’s just one reason why MongoDB Atlas works for us.”

The migration to a fully managed service was also transformative. By using MongoDB Atlas, WBD gained automatic scaling capabilities at both the storage and cluster levels. This proved crucial for handling the unpredictable, massive traffic spikes generated by live sports events or hit TV shows.

“If we’re streaming a big football game, everything goes up. If we have a new episode of a very popular show, everything goes up,” Bakri noted. “Even very localized shows or events can create peaks that are difficult to anticipate.”

To interact with this complex data environment, WBD’s engineers rely heavily on MongoDB Compass and the MongoDB Atlas Console. The integrated LLM in Compass allows developers to use natural language to instantly generate complex data queries—such as asking for the “next seven football games in Mexico”. It not only saves valuable time but also accelerates onboarding for new team members.

“And with the MongoDB MCP Server, you don’t need to jump from one application to another,” said Bakri. “You have everything on the same platform, on the same IDE. Everything is managed.”

OUTCOME

Empowered teams, optimized costs, global reach

With MongoDB Atlas, WBD now successfully executes up to 100 million read-write operations per region daily. Crucially, it has bypassed the restrictive limits of its previous database, allowing the team to run incredibly heavy, server-side data aggregations without incurring high and unpredictable costs.

Perhaps most impressively, this vast global architecture is fully maintained by a lean team of just seven engineers, thanks to the streamlined management tools provided by MongoDB Atlas.

“The Online Archive and Data Federation tools are also great to use,” added Bakri. “And the alerting and monitoring show us exactly how the data storage has increased or decreased.”

With a highly efficient infrastructure in place, the team can now confidently deploy multiple parallel pipelines that process viewership metrics alongside live content updates, ensuring relevant, hyper-personalized engagement. Defined engagement rules, such as frequency capping and time-zone awareness, also ensure that notifications reach users when they are most likely to be watching.

“Thanks to MongoDB, we can reach more people with more sophisticated and more granular recommendations,” said Bakri. “And we can cover around 4,000 users per second in a single batch.”

Ultimately, MongoDB has allowed WBD to turn a chaotic mass of global streaming data into a highly reliable, cost-effective engine for user engagement. It has also benefited from robust, continuous support from MongoDB's dedicated global teams, enabling Bakri and his team to explore advanced technologies such as AI and Retrieval-Augmented Generation (RAG).

“MongoDB Atlas has given us many efficiencies from multiple perspectives, especially as we’re a relatively small team,” Bakri said. “We needed the power of a managed service to maintain our expansion, and that’s what we have in MongoDB Atlas.”