OllyGarden raises $4M betting less telemetry beats more: delete noisy data at the source

Developer platform that continuously evaluates telemetry quality and turns findings into fixes at the source, built on OpenTelemetry.

Observability's cost problem is data quality, not volume tooling: fix instrumentation at the source and delete the rest, in a market where vendors bill by data stored.

$4 million.

OllyGarden launched in 2025 to attack a side effect of the observability boom: telemetry that is unreliable, unnecessary and expensive. Its platform evaluates data quality continuously and fixes problems at the source, so bad data never reaches observability backends — cheaper to store and safer for engineers and AI agents to act on.

Over its first year it helped customers identify log-volume cuts of up to 85%. In October 2026 it raised $4M backed by Next Frontier Capital, Grand Ventures and ACTAI Ventures with existing investor DIG Ventures participating — alongside strategic investments from Datadog, Grafana Labs and Dash0 — and launched Minimum Viable Instrumentation for Rose: spot missing instrumentation, guide a baseline, keep practices consistent across teams, repositories and pull requests.

Telemetry bills scale with volume, so teams hoard noisy data; deleting up to 85% of logs cuts cost and sharpens signals at once.

AI agents now act on telemetry, so unreliable data no longer just confuses dashboards — it drives wrong actions.

AI-generated code multiplies instrumentation of uneven quality, exactly the gap MVI is built to close.

Datadog, Grafana Labs and Dash0 took strategic stakes, tying the volume-billing incumbents to the quality layer their platforms depend on.

When an industry bills by volume, subtraction can be the product: charge for making data smaller. 'More instrumentation is not better instrumentation' — no more, no less.

As of 2026-10-09 the round was closed and MVI had shipped for Rose, with the platform expanding for customers in retail, financial services and enterprise software. Founder Juraci Paixão Kröhling says AI-generated code is already pushing telemetry volumes up, which the company positions as demand rather than threat.