It’s a big moment for our team—one that comes after months of research and a shared conviction that we’re tackling a problem the industry can no longer afford to ignore. Observability is essential, yet its costs are spiraling out of control, forcing engineering teams into an impossible choice: maintain full visibility or keep costs in check. My co-founders, Amir Jakoby and Erez Rusovsky, and I started Sawmills because we believe they shouldn’t have to choose.
How the journey began
We didn’t start with a product—we started by building the right team. Between us, we have spent years in DevOps, AI Ops, and engineering, working with teams pushing the boundaries of innovation while managing growing complexity.
Erez and I worked together at Rollout and always knew we wanted to do it again. When we met Amir through a mutual friend, it quickly became clear he was a missing piece. The more we talked, the more it became obvious—we had the right mix of skills, experience, and ambition to take on something big.
Amir has tackled observability challenges at scale as VP of AI Ops at New Relic, driving automation and intelligence in operations. Erez has built and sold developer-focused startups and led next-gen CI/CD products at CloudBees. I’ve scaled multiple developer-first companies, each acquired by Splunk, CloudBees, and Tricentis.
With the team in place, we shifted to finding a big problem to solve.
The hunt for the right problem
Instead of chasing random ideas, we applied clear criteria to the search. We sought a massive market with room to scale, focusing on a painful, unsolved problem rather than a mere “nice-to-have.” By targeting a well-defined buyer persona, we ensured a clear focus while entering a space without a dominant winner, where competition primarily came from small startups. Our approach was to deliver a fundamentally better solution—not just an incremental improvement—while engaging a buying group that also serves as the implementing group, making adoption easier.
We explored multiple ideas, including code instrumentation, but every time we tested these concepts with VPs of Engineering and DevOps leaders, the response was the same. "Interesting, but this isn't our biggest problem." So, we asked the natural follow-up: "What is your biggest problem?" Again and again, we heard the same answer: Observability costs are out of control.
The observability cost explosion
Observability costs are directly tied to the volume of telemetry data being ingested. Modern architectures—microservices, containers, distributed systems—generate an explosion of telemetry data. And yet, most teams lack a system to intelligently manage or reduce it.
We spoke with over 100 engineering and DevOps leaders and asked a simple question: How much of the telemetry data you send to observability platforms is actually useful? We expected answers in the range of 40%–70%. Instead, the response shocked us: only 10%–30% of telemetry data is useful. That meant 70%–90% of observability data was noise—wasting budget, slowing performance, and reducing the effectiveness of observability itself.
Where is the waste coming from?
The sources of this waste are broad. Developer mistakes lead to verbose logging, debug logs being accidentally sent to production, and chatty libraries creating unnecessary data. There is also a lack of ownership, as DevOps teams "own" observability, but developers generate the data without feeling the pain of the costs. Even when they care, figuring out what’s unnecessary is difficult, and fixing it is even harder.
Out-of-control telemetry data isn't just an efficiency problem—it has real consequences. Skyrocketing costs lead to budget overruns, unpredictable spikes, and year-over-year cost increases that spiral out of control. The lack of control and governance makes it difficult to enforce data collection policies, resulting in more noise and making it harder to find useful signals in observability platforms. For on-prem observability, excessive telemetry data can even impact observability system availability, adding risk instead of reducing it.
Observability tools are supposed to help engineering teams, but they're only as good as the data they ingest. If 70%–90% of that data is noise, then we don't just need better observability tools—we need a smarter way to manage telemetry data itself.
The solution: smart telemetry data management
We set out to attack the problem with a solution that would eliminate waste without adding complexity for engineering teams. It had to be smart, leveraging AI to optimize telemetry data without manual intervention. It had to be automated, detecting inefficiencies and acting on them instantly. It needed to provide control, allowing engineers to fine-tune automation as needed, ensuring critical telemetry data remained intact. And it had to be OpenTelemetry-native, built on the OpenTelemetry (OTel) Collector for seamless integrations and vendor flexibility.
Sawmills delivers on this vision with AI-driven telemetry management that refines your pipeline in real-time. The platform continuously analyzes data streams, identifying cost-saving opportunities, improving data quality, and preventing observability system failures. With a single click, organizations can implement Sawmills’ recommendations and enforce automated policies to keep costs predictable and systems running smoothly. Intelligent data routing ensures telemetry is stored efficiently while providing the freedom to switch observability vendors without disrupting existing monitoring workflows.
Sawmills ensures high-quality telemetry while keeping costs in check by detecting missing data points, normalizing inconsistencies, and applying smart sampling techniques. For on-premises environments, built-in policies safeguard against availability issues while maintaining system stability. This intelligent approach gives enterprises full control over their observability strategy, ensuring long-term flexibility as their needs evolve.
We're humbled by the extraordinary support from our design partners, customers, and investors. With $10M in funding from Team8, Mayfield, and Alumni Ventures, we’re on a mission to make observability work for you, not against you. Huge shoutout to our team, our early customers and my amazing co-founders, Amir Jakoby and Erez Rusovsky, for building this together.
Let's fix observability together. Schedule time to see Sawmills in action.
Thanks,
Ronit Belson and team
