UCF metrics explained: the 6 dimensions of coordination

Beyond token counts

Token usage is a common cost signal. But what did those tokens produce? Was the output useful? Did the agents work together or duplicate effort?

Token counts help measure cost. UCF adds coordination signals; it does not independently measure output quality, factual accuracy, or safety.

The Universal Coordination Field (UCF) is Samsarix's framework for measuring how well your automation system is performing — not just "did it complete?" but "how well did it coordinate?" Think of it as the difference between "the car drove from A to B" and "the car drove smoothly, efficiently, safely, and on time."

The six dimensions

Harmony (0-100%)

How well agents collaborate without stepping on each other's work. High harmony means clean handoffs, shared context, and consistent outputs. Low harmony means duplicated effort or conflicting decisions.

Example alert: A sustained drop below your workload's tested baseline can justify reviewing role assignments and handoffs. Do not adopt 50% or any other universal threshold without calibration.

Friction (0-100%, lower is better)

Where resistance exists in the system — API failures, malformed inputs, slow external services, ambiguous instructions. Friction doesn't mean failure; it means the system is working harder than it should to produce the same result.

When to worry: Rising friction with stable harmony suggests external input issues. Check integrations and data formats.

Throughput (tasks/time)

Raw processing capacity. A spike in throughput with rising friction suggests the system is burning resources to maintain speed — a warning sign before something breaks.

Focus (0-100%)

Precision and attention quality. Are agents working on the right tasks? High focus means compute is spent on meaningful work, not busywork or re-doing failed steps.

Resilience (0-100%)

Recovery capability. When an API goes down, a model returns garbage, or a webhook is malformed — how quickly and cleanly does the system recover? High resilience means failures are brief and self-correcting.

Velocity (execution speed)

Raw execution speed, separate from throughput. High velocity with low throughput suggests bottlenecks between steps rather than within them.

Reading the signals

The power of UCF is in the relationships between metrics:

UCF vs traditional monitoring

Traditional monitoring tells you: "The API returned 200 OK."

UCF may suggest: "The API returned 200 while Friction rose near the integration boundary. Harmony also moved, so inspect both the response shape and the next handoff."

UCF complements logs, traces, task outcomes, and human review. It does not replace them or establish causation on its own.

Accessing UCF data

When live instrumentation is available, UCF metrics can update on the dashboard and are available via API:

GET /api/health/detailed
GET /api/coordination/metrics
GET /api/coordination/trends?window=1h

See UCF metrics in action. Start free → Inspect recorded coordination signals from instrumented runs.