Getir Accelerates Rapid Commerce with Confluent Cloud Real-Time Streaming

Getir pioneered fast delivery by creating an entirely new category with its “delivery in minutes” model. Today, the company serves 40 million people across Turkey through an innovative network of micro-warehouses — small, strategically located fulfillment centers stocked with popular grocery items and everyday essentials.

Operating as a super app with multiple verticals, including GetirFood, GetirLocal, and GetirMore, success for Getir depends on maintaining consistent real-time data flows that enable discovery, checkout, and rewards—all high-traffic components where reliability and clean data contracts are critical.

But rapid growth revealed fundamental architectural limitations. Early systems relied on point-to-point integrations that created tight coupling, cascade failures during peak traffic, and unclear data ownership. The distributed monolith architecture made it difficult to scale efficiently while maintaining the speed and reliability customers expected from rapid commerce.

Confluent's data streaming platform transformed Getir's engineering foundation by creating a unified data backbone that connected operational systems with business applications. Now, teams can access and share information seamlessly through event-driven microservices, focusing on product innovation rather than infrastructure complexity.

A monolith stood in the way of innovation

Getir's initial architecture relied on synchronous point-to-point requests through HTTP and gRPC. While the company had many services, tight coupling meant they functioned as a scattered unit. 

"We grew very fast, and that speed meant we built a distributed monolith, tight coupling, unclear ownership, and data that wasn't fresh," explains Can Sözeri, Engineering Director at Getir.

During peak traffic periods, Getir’s legacy architecture created cascade failures as services waited for synchronous responses. Version choreography across services became increasingly complex, and any issue in one service could ripple through the entire system.

Multiple teams shared databases, blurring the source of truth across the technology organization. The data team ran batch ETL jobs against primary databases to feed machine learning models, adding extra load and cost, while creating stale downstream data.

The company had experimented with open-source Kafka and AWS MSK in limited areas like G-Locals, but usage was minimal and inconsistent across business units. Simple queueing mechanisms couldn't support Getir's vision for a true event-based architecture.

Choosing the best platform for managed data streaming

As Getir's Market Demand branch investigated managed streaming platforms, Confluent stood out for its scalability, reliability, and operational efficiency.

Transitioning from minimal, open-source Kafka usage to Confluent Cloud as a centralized data streaming platform aligned with Getir's vision to help teams focus on product innovation rather than infrastructure management. 

"Fully managed Kafka lets our teams focus on event contracts and product work, not operations," Sözeri explains. "We needed to cut undifferentiated operations – capacity upgrades, cluster hygiene, connector glue and schema registry."

Confluent Cloud freed engineering teams from cluster maintenance, capacity planning, and connector management – allowing them to focus on defining event contracts and building product features. 

The Market Demand team led the transformation by reimagining its core workflows as event-driven systems. Other business units, including G-Food and G-Locals, soon followed, adopting the same patterns and benefiting from shared learnings.

Transforming checkout and order lifecycle

One of Getir's first critical use cases was decoupling its tightly integrated order systems. Previously, payments, inventory, notifications, and logistics all depended directly on a centralized order database through synchronous HTTP calls.

Using Confluent, domain events now coordinate these services without hard dependencies. When an order is placed, events flow through the platform to trigger payments, update inventory, send notifications, and coordinate warehouse logistics—all asynchronously. Services no longer wait for each other, eliminating cascade failures and improving resilience during peak traffic.

Real-time inventory and pricing for discovery

Getir's discovery teams needed to show customers accurate prices and stock levels across hundreds of warehouses in real time. Making synchronous calls to inventory and pricing services on every client request created significant latency and backend load.

Through Confluent, Getir built a data mesh that ingests and stores real-time inventory and price streams. Discovery services now read from localized, consistent data rather than querying external systems on every request.

"This reduces latency and load dramatically," Sözeri explains. "We don't call inventory and pricing services on every request anymore. Reads are local and consistent."

Stream processing with Apache Flink

As Getir's event architecture matured, the team adopted Apache Flink® to power its advanced stream processing needs. Flink manages topic-level filtering across Getir's extensive network of warehouses and operations, ensuring that services receive only the essential, relevant data.

Flink was the optimal choice due to its proven scalability, resilience, and reliability when managing high-volume, low-latency workloads that seamlessly merge batch and stream data. While powerful, Flink is a complex technology. To simplify adoption, Getir chose Confluent's stream processing engine, which is built on Flink. This decision made advanced stream processing accessible and removed the operational burden of managing Flink directly.

"Not everyone needs every event," says Sözeri. "Flink helps us deliver exactly what each operation needs, including late event handling and deduplication."

The combination of Confluent Cloud and Flink created a scalable foundation for complex real-time processing, from enriching events to routing them intelligently across the platform.

Results

With a fully event-driven architecture powered by Confluent Cloud, Getir achieved exceptional results, including:

Improved performance

Homepage, discovery, and checkout pages are now approximately 60% faster, delivering the seamless experience customers expect from rapid commerce. "It only took one year to see the big results – faster app, lower costs, more stability, and teams could finally move at their own speed," says Sözeri.

Massive cost savings

By eliminating batch ETL jobs against primary databases and reducing redundant processing, Getir cut database costs by roughly 70% within the first year.

Enhanced reliability

Tech crisis incidents decreased by approximately 60% as the asynchronous architecture eliminated cascade failures during peak traffic. Services no longer wait for synchronous responses, keeping systems stable when it matters most.

Accelerated development cycles

Development speed increased by 50% as teams shifted focus from infrastructure management to building features. Services are now decoupled and independently deployable, allowing teams to move at their own pace.

Organization-wide adoption

Success in Market Demand inspired other business units like G-Food and G-Locals to adopt event-driven architecture, with Getir running internal training seminars to spread best practices across the company.

What’s next for Getir and Confluent?

Looking forward, Getir plans to take real-time data streaming a step further and build predictive models to forecast customer behavior and operational needs. Namely, to predict courier ETA and the likelihood of customers placing orders within specific time windows.

"We’ll be able to predict when a client is going to make an order, maybe after five or ten minutes,” says Sözeri. “If we can anticipate, we can create better architecture for our couriers, warehouses, and picking processes."

Real-time event streams from orders, basket activity, listing pages, and discovery feed directly into AI and LLM models. Data engineers can then use these events to enhance recommendations and further improve the customer experience.

For Getir, Confluent represents more than technology. The partnership, professional services engagement, and ongoing collaboration have reshaped the company's engineering culture and empowered teams to innovate. With Confluent as its data streaming foundation, Getir continues to push the boundaries of rapid commerce – delivering exceptional customer experiences in minutes.

Learn more about Getir