BigPanda slashes observability costs without losing visibility

About BigPanda

BigPanda is the kind of company that quietly powers the internet’s uptime. Its platform is built for agentic automation, a smarter, faster way to run ITOps without drowning in manual effort. The system doesn’t just detect incidents; it responds to them, prevents them, and helps teams stay ahead of failure in real time.

Used by some of the world’s most iconic brands, BigPanda sits at the center of modern IT Operations. Its AI Detection and Response uses real-time signals and automation to detect, diagnose, triage, and resolve issues quickly. BigPanda can also surface hidden knowledge across siloed teams and systems to quickly uncover what’s happening, why, and how to fix it.

A telemetry firehose with no off switch

“Sawmills helped us cut observability costs by 63%, without sacrificing visibility. And ingestion was slashed by an incredible 93%. Now, we index only what matters, and our engineers have reclaimed time that used to be spent chasing noisy telemetry.”

When Ben Neumann joined BigPanda as Team Lead of DevOps, the company was already in the midst of a strategic shift to proactively scale its telemetry strategy to support growth. This included consolidating observability tooling to Datadog, and navigating the ever-growing flood of telemetry data, especially logs.

The team had taken early steps to control spend by indexing only 10% of ingested logs, but with Datadog charging for ingestion rather than indexing, there remained a disconnect: rising costs without a proportional gain in actionable visibility.

“We were still paying for every log we ingested. Even if we didn’t index them, they still cost us.”

^{Ben Neumann, Sr. Director of DevOps}

Support teams rely on full access to logs to troubleshoot customer issues. Developers needed live access for debugging. Balancing these needs in a fast-paced environment required careful coordination, and the manual effort to manage log volume across teams often introduced unnecessary overhead.

“Fixing log volume manually was like playing whack-a-mole,” Neumann explained. “You’d fix it in one place, and a week later, someone else would unknowingly flood the system again.” BigPanda saw an opportunity to accelerate its observability evolution, preserving visibility and agility, while dramatically reducing cost and complexity.

Say hello to smarter telemetry management

Enter Sawmills. BigPanda began routing all observability data through Sawmills’ smart telemetry management platform. Instead of filtering after ingestion in Datadog, they extracted those same filters and implemented them upstream before the logs were ingested.

“Sawmills let us keep the exact same filters we’d built in Datadog, but move them earlier in the pipeline. We’ve stopped paying for logs we don’t need.”

^{Ben Neumann, Sr. Director of DevOps}

That initial shift made an immediate impact. Only about 10% of the log volume continued into Datadog for the R&D team. The rest was routed to low-cost storage in S3 and queried via Snowflake. Ingestion costs dropped dramatically, and support teams retained access to 100% of the logs, without expensive rehydration or delays.

“Before, logs that weren’t indexed were not accessible,” said Neumann. “If a big customer issue came in, we had to rehydrate logs, which slowed everything down. Now they’re always available, without the cost.”

To support real-time debugging workflows, Sawmills shipped their own Live Tail tool, enabling BigPanda to explore telemetry data at any point within their pipeline, before or after any processor, and per pipeline or destination. Now, they can inspect raw data exactly where it's sent or transformed, providing real-time visibility into their logs without triggering ingestion-based charges.

Continuous optimization with Sawmills’ AI-powered insights

With ingestion under control and visibility maximized, BigPanda’s next focus is sustainability to ensure the improvements they’ve made persist and continue to evolve. That’s where Sawmills AI-powered telemetry management comes in.Sawmills continuously analyzes incoming telemetry to identify high-impact inefficiencies in real time. It detects patterns like:

1. Duplicate or redundant log lines
2. Excessively verbose debug messages
3. Schema mismatches and formatting issues
4. Repeated attributes or multi-line logs that break parsing
5. Redundant metrics and unused time series

“Sawmills doesn’t just help us filter data. It tells us exactly what needs filtering or transforming, we’re not just cleaning up after the fact anymore. We’re preventing telemetry bloat before it starts.”

^{Ben Neumann, Sr. Director of DevOps}

Each AI-generated insight includes sample data, a severity score, and a breakdown of volume impact, pinpointing which services are generating excess and what to do about it. Suggested actions, such as dropping, sampling, or aggregating, are surfaced directly in the UI.For BigPanda, these capabilities represent the next frontier in observability, moving from reactive cleanup to proactive optimization.

“We’ve already seen huge gains just by reducing what we ingest,” said Neumann. “But the potential to automatically detect waste and resolve it with a click? That’s where we’re headed, and Sawmills is taking us there fast.”

As these insights are activated, teams will be able to take one-click actions to clean up telemetry before it causes downstream issues, without engineering overhead or code changes. And with automated tracking, resolved issues will stay resolved. The goal: telemetry data that optimizes itself.

Fix noisy telemetry at the source with a single click

Fixes are fast and frictionless with Sawmills. Users can take in-context actions, such as routing, dropping, or modifying telemetry data, with a single click. No manual labor, no code changes, and no waiting on developer cycles. And because the system tracks results over time, it automatically marks issues as resolved once they fall below defined thresholds. The result is a self-reinforcing loop where waste gets flagged, addressed, and prevented automatically.

“We won’t just be reacting to spend anymore — we’ll be proactively improving data quality,” said Neumann. “Sawmills will show us what’s noisy, where it’s coming from, and how to fix it with a single click, saving us significant time and money as we scale.”

63% lower observability costs, and 93% reduction in ingestion with full visibility and AI insights

Sawmills delivers instant impact for BigPanda by combining AI-powered insights with one-click actions that eliminate waste, improve data quality, and reduce engineering toil.

• Now, BigPanda’s teams rely on a smarter, faster approach:
• ‍AI automatically flags noisy, redundant, and high-cost telemetry
• Fixes can be applied with a single click—route, drop, or modify data at the source
• No code changes. No dev bottlenecks. No manual cleanup.

• The immediate results
• 63% reduction in observability spend — immediate ROI
• 93% reduction in ingestion costs - only ingesting quality data
• Full visibility maintained — support teams access all logs without rehydration or indexing fees
• Real-time insights across the org — engineers get live-tail access, leaders monitor data waste, and Sawmills handles implementation and verification of fixes automatically

What used to take weeks of investigation, coordination, and code changes will now be resolved by Sawmills AI upstream with a single click. And it keeps getting smarter, continuously surfacing new savings opportunities and validating fixes over time.