An intelligent observability platform that helps enterprise teams understand system performance, investigate anomalies, and act before issues become business risks.
Metric OS unifies telemetry from applications, infrastructure, services, and critical user journeys. Instead of asking teams to interpret disconnected monitors, it prioritizes meaningful anomalies, explains likely causes, and connects every incident to an accountable owner.
Led product strategy, research synthesis, information architecture, interaction design, prototyping, validation, and design-system direction.
1 product lead, 2 designers, 5 engineers, 1 data scientist, and enterprise subject-matter partners.
Reduce time-to-detection, increase telemetry trust, and standardize investigation workflows across teams.
tools used to investigate a critical service issue
of incident time spent reconciling conflicting telemetry
average delay between detecting and assigning an incident
Modern cloud teams rarely lack telemetry. They lack a coherent way to determine which signals matter, how failures are connected, and what action is safe to take. Metric OS correlates logs, traces, metrics, and service dependencies into an opinionated incident workspace.
The platform compresses hundreds of noisy alerts into a single operational narrative, exposes the blast radius, and keeps investigation and remediation in one governed flow.
“I do not need another alert. I need to know what changed, who is affected, and the safest action to take.”
Frankin jumps between logs, traces, infrastructure metrics, and Slack while manually rebuilding the timeline of failure.
Production errors arrive as dense, unorganized payloads with weak connections to services, deployments, and user impact.
Five hundred downstream alerts compete for attention even though they originate from the same upstream regression.
Metric OS correlates temporal proximity, trace ancestry, deployment changes, and service dependencies. Instead of paging Frankin 500 times, it creates a single incident with a confidence score, suspected origin, and affected SLOs.
The topology canvas brings the failing API gateway forward, dims healthy infrastructure, and maps the propagation path across checkout, payments, and identity. Selecting a node pivots the log grid to the exact service and trace context.
Frankin reviews the recommended rollback, validates affected dependencies, and executes it without leaving the workspace. Guardrails show permissions, expected impact, rollback progress, and post-action health verification.
A single-screen response workspace combines spatial service topology with high-density debugging data and governed remediation.
Started 09:41 UTC · Revenue checkout flow degraded · 18.4K users affected
service:api-gateway level:error⌘ ↵ Runupstream connect error: TLS handshake timeout; reset reason=connection_terminationtr_8f2c91aPOST /checkout failed: upstream request timeout after 1800ms; retry_budget_exhausted=truetr_8f2c91acircuit_breaker state=OPEN dependency=tokenization-v2 failure_rate=0.42tr_4bd109eupstream reset before response headers; transport_failure_reason=TLS_errortr_73ac011request completed status=200 duration_ms=68 region=us-east-1tr_a013bd2connection pool saturation=0.91 active=182 idle=4 pending=63tr_1cc81faFirst error occurred 47 seconds after deployment. Rollback is expected to restore checkout traffic within 4–6 minutes.
The extended observability workspace supports system-health scanning and detailed investigation without forcing teams into separate monitoring tools.
Color communicates status and confidence, never meaning alone.
Supporting context / 14
Every visual choice is tied to the cognitive demands of high-stakes incident response.
During an incident, an SRE compares timestamps, service names, statuses, payload patterns, and trace IDs across multiple events. A compact table preserves more causal evidence in the viewport, reduces scrolling, and makes repeated patterns visible. Density is controlled through strict column alignment, monospace payloads, restrained borders, and semantic badges.
UX principle · Maximize comparable evidence per glance.Healthy infrastructure recedes into lower-contrast background layers. Anomalous services rise forward on crisp glass surfaces with elevation, glow, and high-contrast error badges. This reserves saturated crimson and amber for genuine risk while depth and focus communicate hierarchy across the rest of the topology.
UX principle · Use depth before adding more color.The 40% topology canvas preserves system relationships while the 60% debugging pane exposes the supporting logs, KPIs, and traces. Selecting a service pivots evidence without opening another tab. Keeping detection, investigation, and rollback togethis eliminates context reconstruction and reduces working-memory load.
UX principle · Preserve context across every investigative pivot.Prototype validation with enterprise users showed that combining signal prioritization, transparent AI explanations, and governed telemetry reduced time spent interpreting dashboards and increased confidence in incident response.