Do we have metrics, logs, and traces? This is the question most teams use to assess their observability. It is answered with an inventory: count the instrument types, and when all three are present, the assessment is complete. The outputs are counted; what the outputs let anyone do goes unexamined.
This post works from another question: What interpretive function are we performing here?
This question is answered by naming an operation and placing it on a scale. The scale runs from acknowledging that something happened to projecting where a situation is heading. Each rung is an interpretive operation, and each operation determines what can be inferred from an output and where that inference takes place.
The rest of this post walks the rungs in turn.
Three terms
The argument presented here rests on three concepts.
A record is an emitted or collected artifact: a metric point, a log entry, a span, a profile sample. A record holds what was observed at the point of emission.
An operation is the interpretive function performed on or by a record: recording, measuring, correlating, localizing, diagnosing, interpreting, judging, projecting. The operation sets the interpretive reach of an output.
A determination is a meaningful statement about system condition: degraded, stalled, violating, converging, saturated, recovering. A determination carries a reading of the system rather than evidence awaiting a reading.
Records are artifacts.
Operations set interpretive reach.
Determinations are where observability begins.
The ladder
Eight levels, ordered by interpretive reach. Each level is defined by the operation its output performs.
1. Recording — something happened
The output asserts an occurrence. A log entry sits here: a discrete statement committed at a specific point in execution. It captures the fact of the event and the emitter’s description of it.
2. Measuring — something has a value
An output assigns a value to a quantity on a defined scale. A metric quantifies this value, positioning it against the scale. This value invites comparison to a threshold or a prior value.
3. Correlating — these things belong together
Outputs are grouped by a shared identity. Trace context propagation sits here, as does an exemplar linking a metric sample to a span. The operation is association: the output declares that records across boundaries belong to a path.
4. Localizing — the issue appears to be here
A condition is positioned within a grouped structure. Reading a trace against a question sits here: the structure built at level 3 is searched, and the condition is assigned a place within the path — this span, this boundary, this hop.
5. Diagnosing — this mechanism likely produced the failure
A located condition is attributed to a mechanism. A human reading traces and logs together sits here. The operation is attribution: a cause is proposed for the condition placed at level 4. This is the ceiling of telemetry as telemetry. Tools built above the records may carry interpretations of their own; the records themselves reach level 5.
Where the frame lives
Levels 1 through 5 share a property: they operate on records, and the records hold facts, values, groupings, and positions. The frame that turns those into a diagnosis is supplied by the reader, after collection.
- A trace holds spans in relation. What the relation means for the service rests with the reader.
- A metric holds a value. Whether the value is acceptable rests with the reader.
- A log holds a statement. The significance of the statement rests with the reader.
At each of these levels, the output waits for a reader to apply a frame of reference to give meaning and make some sense.
The rungs above five rest on a different arrangement. The system applies the frame before emission, and the determination travels with the output.
6. Interpreting — this emission has this meaning
An emission is assigned significance. The operation is signification. A measured or located condition is bound to what it means for the system: this latency rise is degradation, this gap is a stall. A value becomes a sign that carries a reading.
7. Judging — this component is keeping or breaking a commitment
A component is appraised against a commitment. The operation is appraisal. The output states whether an expectation is held: this dependency is honoring its promise, this caller is exceeding its agreed rate. The judgment is the output’s content, not a downstream derivation.
8. Projecting — this situation is likely to evolve in this direction
A present condition is extended into a trajectory. The operation is anticipation. The output gives the situation a direction of travel: this saturation is approaching a limit, this pattern resembles a prior failure run.
The governing condition
A level is defined by the interpretive operation it performs, which determines the application of the frame of meaning.
From level 1 to level 5, the frame is external. The system emits records, and a reader supplies the question, the expectation, and the significance after collection. The interpretive act happens downstream of emission.
From level 6 to level 8, the frame has moved into the system. The output carries a determination made against an expectation before emission. The interpretive act happens at the source.
Observability pertains to the degree to which a system’s internal state can be deduced from its outputs. Determinability is attained when the output explicitly communicates the system’s condition, which is achieved at level 6 when the system’s output directly states its state.
CLIMBING The Ladder
OpenTelemetry (OTel) standardizes instruments whose outputs reach level 5. It is a toolkit for generating records and carrying them on a common wire, and its reach ends where the reader’s interpretive work begins.
Our reframing replaces the inventory question with the function question.
The diagnostic becomes: which interpretive operation is being performed? The answer places the system on the ladder.
A system with trace context propagation operates at level 3 — correlation.
A system that determines its own condition operates from level 6, where outputd carry determinations made at source.
The field’s slogan holds that observability and monitoring are distinct practices. Beneath the slogan sits a stack of instruments (metrics for monitoring, traces for diagnostics, logs for troubleshooting) and every instrument in the stack performs an operation at level 5 or below. The gap traces to a backward-looking definition: the field labeled its existing instruments “observability,” and the term inherited the limits of those instruments.
A forward definition begins from the work of understanding (determining internal condition from outputs) and places that work at levels 6 through 8, where the output carries the determination the work requires.