Skip to main content
kellerai.blog

Observability Theater

Empty telemetry fields train operators to trust signals that carry no information.

KellerAI White Paper · Observability & Drift · May 2026

Context

In production compliance platforms, governance telemetry events carry fields named to promise specific content. When those fields emit permanent empty lists on every event—an obligations-referenced field that stays empty regardless of which compliance packs were applied or which obligations were evaluated—they train operators to treat empty as the expected state. After enough exposures, the field becomes invisible despite being visually present. This is worse than emitting nothing, because it produces false confidence in monitoring coverage that does not exist.

The Finding

The remedy is a field-level liveness contract: a schema commitment that asserts not only the type of a field but the conditions under which it must carry content. A counter that increments every time an obligations field is emitted empty converts a silent failure into a visible debt counter, making the business case for fixing the wiring legible to leadership for the first time.

Tags:
Observability & TelemetryCompliance Audit TrailsData Quality
Paper Details
CategoryObservability & Drift
AudienceEngineering teams, platform architects, and compliance leads governing telemetry and audit-trail design
MethodProduction case study analysis + human-factors literature + schema-first telemetry design patterns + compliance audit-trail standards (NIST SP 800-92)
Length~960 · 4 min
Sections6
DateMay 2026
AuthorsKellerAI
Read the full paper
Related
Placeholder — pending analytics
Section 01

The field that is always empty

In one production compliance platform, an event fires on every run. It carries a field meant to record which compliance obligations that run actually verified. The field has a clear, meaningful name. It satisfies the schema. It shows up on dashboards. And it is permanently, structurally empty — the code that assigns it writes a hard-coded empty list, never computed from anything. The docstring beside it is honest about the gap: the value is a placeholder, and the wiring needed to fill it “does not yet exist.” The dashboard downstream displays which obligations a run checked. It always displays zero.

Section 02

Empty is not neutral

The intuitive assumption is that an empty field costs nothing: it carries no data, so it does no harm. The human-factors research says otherwise. When an automated system reports the same stable output on every event, operators stop scrutinizing it. The output stops registering as a signal and starts functioning as background. A field that always reads empty trains the people watching it to treat empty as the expected, correct state. After enough exposures, nobody questions it — the field is visually present but cognitively dismissed. This is not a failure of attention. It is the predictable way human attention responds to a stimulus that never changes.

Section 03

The alarm-fatigue parallel

Hospitals know this dynamic. Studies of intensive-care units find that the overwhelming majority of alarms are clinically insignificant, and that nuisance alarms erode nurses' confidence in the monitoring system as a whole. Security operations centers see the same: high false-positive rates produce desensitization, and desensitized analysts miss real incidents. An empty telemetry field is the silent version of the same failure. A nuisance alarm is loud and wrong; an empty field is quiet and wrong. There is no alert to dismiss — the field simply reads zero, on every event, indefinitely. Operators have nothing to react to and therefore nothing to flag. The controlling variable is not loudness; it is invariance.

Section 04

Why this is compliance theater

In a compliance-automation product, governance telemetry is not optional instrumentation — it is the audit trail. Logging standards require that records carry sufficient detail for after-the-fact investigation. An event that permanently emits an empty obligations field fails that requirement on every event, silently. The audit trail is present; the audit evidence is absent. Worse, the empty value cannot tell two very different facts apart: “no obligations were evaluated” and “the wiring to record obligations does not exist” produce the identical zero. An internal engineer may know it is a placeholder — but the day an external auditor queries the event stream, a comment buried in the source code is not a documented control. The placeholder presents as evidence.

Section 05

The fix is a liveness contract

The remedy is not a smarter dashboard. It is a field-level liveness contract: a schema commitment that asserts not only the type of a field but the conditions under which it must carry content. The obligations field MUST contain at least one obligation identifier when a governance pack was applied and the run completed evaluation — emitting empty in that case is a contract violation, not valid data. Until the real wiring lands, a single instrument helps immediately: a counter that increments every time the field is emitted empty. That converts a silent failure into a visible debt counter. Once leadership can see how many audit records per day ship hollow, the cost of inaction becomes legible — and the business case for the fix becomes tractable.

Section 06

The point

The core line is worth saying flatly: a named field is a promise. When it promises specific content and emits none on every event, it has not merely failed to inform — it has produced false confidence in monitoring coverage that does not exist. Fixing it is not only a code change. Operators who have internalized empty as normal will read the first real value as an anomaly to be investigated rather than a migration that finally worked. A fix that repairs the wiring and skips the human handoff just replaces one form of theater with another. The honest goal is telemetry whose fields carry information — and that can be checked, by a machine, on every event.

For the full argument — with the code-level evidence, the human-factors literature, and the detection, remediation, and migration playbooks behind it — read the companion technical whitepaper, Observability Theater: In Depth .