How Circles built its observability stack on Grafana Cloud and launched a $12M ARR travel lifestyle startup without a DevOps team

Circles didn’t just modernize its observability stack with Grafana Cloud—it changed the economics of how it builds and scales products. As a result, the company reduced tooling costs, eliminated DevOps bottlenecks, and proved it could launch and grow a $12 million ARR business without a dedicated DevOps team.

“We avoided getting a dedicated DevOps person and we launched that into a $12 million ARR business without a single DevOps,” said Kelvin Chua. “And that’s unheard of.”

Previously, the picture looked very different: fragmented tools, rising costs, and increasing headcount just to manage observability infrastructure. The gap between what the tools promised and what the business needed was widening.

When Kelvin Chua joined Circles in 2014, he was the first employee to join the founders… and the only engineer.

The three founders needed someone to build their stack, so he did it. And over the next decade, as Circles expanded from a Singapore telco startup into a multinational operator with markets spanning Asia, Europe, Africa, the Pacific, and Latin America, Kelvin held nearly every technical role the company ever needed: infrastructure lead, data strategist, chief architect, head of DevOps and SRE, and eventually the executive who built the corporate venture engine that produced some of Circles’ most ambitious new products.

That breadth of experience gave him a perspective most technology leaders don’t have: he had seen, firsthand, what it cost—in dollars, headcount, and missed business outcomes—to get observability wrong.

The cost of fragmentation

By 2018, Circles was happily running their legacy  observability platform. The original rationale was sound. At the time, that proprietary platform was the strongest application performance monitoring product on the market, and Circles needed visibility into an increasingly complex microservices architecture. APM was the draw.

But over the years that followed, Kelvin noticed something troubling. The promise of their legacy platform’s APM—faster incident resolution, tighter feedback loops, engineers drilling down to root cause in minutes—was not being realized. “We were paying a lot of money without hitting our business outcomes,” he said. “To me, it’s not acceptable.”

Then, around 2021, there was change in their legacy vendor’s pricing model, moving from negotiated contracts to consumption-based billing. For a company operating multiple products across multiple geographies, the math no longer worked.

At the same time, their commercial ELK (Elasticsearch, Logstash, Kibana) stack presented a different kind of problem. It wasn’t about pricing they paid to license it. It was about people. Every new market launch and every new Kubernetes cluster required more engineers who knew how to manage Elasticsearch. “For every launch that we made,” Kelvin said, “we had to multiply X-fold how many people we had at that moment just to manage Elasticsearch.”

Circles evaluated proprietary observability tools. But for a company that wanted to move quickly and stay lean, the economics didn’t shift enough to justify a migration.

A different kind of solution

Kelvin had been watching Grafana for years. Originally, he’d had Grafana OSS running alongside their legacy APM tool and appreciated the open-source ecosystem, but the maintenance burden of running the OSS version was a dealbreaker. “We actually figured out that at that point, in 2018, Grafana in itself was not able to give us what we need because we had to maintain it ourselves.”

That changed when Grafana Labs launched its hosted cloud offering, Grafana Cloud.

The entry point for Circles wasn’t metrics or tracing. It was logs. In 2021, as Kubernetes was becoming foundational to Circles’ architecture, log aggregation had become critical. Running multiple Kubernetes clusters without centralized log management was untenable, and every major logging vendor Kelvin evaluated was still building on Elasticsearch – the same infrastructure problem he was trying to escape.

Loki, the open source log aggregation system from Grafana Labs, was different. Lightweight, indexed by labels rather than full-text, and now available as a hosted service through Grafana Cloud, it offered a path out. “When Loki came out and I tried it myself, I was amazed at how lightweight it was,” Kelvin recalled. Once Grafana Labs was hosting it, the last barrier disappeared. “That was actually the main driver on why it made sense for me to actually move all my logs to Grafana Cloud.”

From there, the stack expanded incrementally. Prometheus replaced InfluxDB for time-series metrics. Grafana’s Kubernetes Monitoring feature dramatically simplified cluster onboarding. The CloudWatch integration for Amazon EKS environments went further, enabling Circles’ engineers to query metrics directly from AWS without pulling data into Grafana Cloud first. “I don’t need to duplicate data,” Kelvin noted. “I just need to directly query from CloudWatch, which I think is actually quite cool.”

The cultural shift

The technical migration from Circles’ previous platform to Grafana Cloud was significant. But Kelvin argues the cultural change it enabled mattered more.

Under the previous model, developers at Circles were largely dependent on SRE and DevOps teams for observability. Building a dashboard required looping in specialists. Responding to an alert required routing through the right people. The model created friction in sprints, bottlenecked incident response, and put a ceiling on how fast teams could move.

Grafana Cloud changed the dynamic. “Grafana is so easy to use that developers can do it themselves,” Kelvin said. “We are now moving things back towards the developers, meaning they write the code, they write the instrumentation, they write the alerts, they respond, they know what’s going on.”

This wasn’t a theoretical organizational design. It was tested in practice. 

Building without a safety net

Jetpac, Circles’ consumer venture, was built and scaled entirely on Grafana Cloud. No legacy APM vendor. No ELK. No dedicated DevOps team.

That last point bears emphasis. Jetpac reached $12 million in annual recurring revenue without a single dedicated DevOps or SRE hire. 

Part of what made this possible was the Grafana community – something Kelvin cited repeatedly as a strategic advantage that proprietary tools can’t replicate. 

When engineers hit unfamiliar territory, the answers are already out there. “The availability of information is quite better compared to our previous solution. If we need answers, it’s going to be easier for us to find as someone’s probably already been there and shared their experience.” 

With their prior observability platform, the team was dependent on the vendor’s customer success resources. On Grafana Cloud, the open source ecosystem meant no single point of support dependency.

What comes next

The lessons from Jetpac are now being applied to Project Black, Circles’ newest corporate venture as it will be 100% Grafana Cloud supported, and it is built AI-native from the start.

“Project Black is AI-native, so we don’t want to have any incident at all,” Kelvin explained. “AI needs to be able to trigger workflows to autofix itself. So Grafana Cloud will be used in tandem with how we built this AI-native product so that any issues that happen underneath can easily be solved by the AI itself.” Tempo and distributed tracing will play a larger role here, given the depth of the service dependency chain.

For Kelvin, the question of why Grafana Cloud is the default for every new venture at Circles has a straightforward answer. It’s already proven. The headcount math works. The community is there. And when use cases evolve—as they always do—the platform evolves with them.

“There would be no Jetpac,” he said. “There will be no Xplore. There will be no Project Black. We would not have been able to launch these corporate ventures without Grafana Cloud.”