KPI and SLO Catalog — Operating Quantum Computers

This appendix provides a starting catalog of operational metrics for quantum platforms. The exact thresholds are system-specific; the metric families are broadly useful.

H.1 Metric hierarchy

Diagram loads as you read

H.1 Metric hierarchy · Figure 1

View diagram source

flowchart TB
    Metrics[Quantum platform metrics] --> Health[Hardware health]
    Metrics --> Quality[Computational quality]
    Metrics --> Flow[Workload flow]
    Metrics --> Cost[Cost and capacity]
    Metrics --> Evidence[Evidence quality]
    Metrics --> Security[Security and governance]

Metrics should be linked to decisions. A metric that does not change routing, release, admission, investment, or remediation is probably dashboard decoration.

H.2 Hardware health metrics

Metric	Decision it supports
calibration convergence rate	whether recipes are stable
median calibration age at execution	whether results risk stale context
active usable qubits/zones/modes	whether target model is valid
readout error trend	whether measurement mitigation is needed
two-qubit or entangling-operation quality trend	whether circuit class remains admissible
crosstalk sentinel score	whether parallelism should be restricted
facility alarm count	whether hardware regressions may be environmental

H.3 Workload-flow metrics

Diagram loads as you read

H.3 Workload-flow metrics · Figure 2

View diagram source

flowchart LR
    Submit[Submitted] --> Admit[Admitted]
    Admit --> Queue[Queued]
    Queue --> Execute[Executed]
    Execute --> Complete[Completed]
    Complete --> Accepted[Evidence accepted]

Metric	Decision it supports
admission rejection rate by cause	improve preflight or policy
queue wait median and tail	adjust capacity and reservations
cancellation rate	detect poor scheduling promises
reservation utilization	reduce expensive idle windows
simulator preflight pass rate	catch avoidable QPU failures
retry rate	identify platform or workload instability

H.4 Quality metrics

Metric	Decision it supports
benchmark score with confidence interval	release and vendor acceptance
workload success likelihood by class	routing and admission
baseline comparison delta	claim acceptance
mitigation sensitivity	result qualification
drift-adjusted quality score	calibration and maintenance timing
reproducibility score	evidence package completeness

H.5 Evidence metrics

Diagram loads as you read

H.5 Evidence metrics · Figure 3

View diagram source

flowchart TB
    Evidence[Evidence quality] --> Provenance[Provenance completeness]
    Evidence --> Statistics[Statistical adequacy]
    Evidence --> Review[Review status]
    Evidence --> Retention[Retention compliance]
    Evidence --> Trace[Trace correlation]

Metric	Decision it supports
evidence package completeness	publication and customer readiness
percentage of results with confidence intervals	statistical governance
percentage of jobs with tags	searchability and incident analysis
trace correlation coverage	operational debugging
raw artifact retention compliance	audit readiness
rejected claims by reason	training and process improvement

OpenTelemetry semantic conventions can help normalize traces and metrics across platform services. NIST CSF 2.0 can help align security metrics with broader risk governance. [R73] [R89]

H.6 Cost and capacity metrics

Metric	Decision it supports
cost per accepted evidence package	investment governance
QPU seconds per accepted result	workload efficiency
shots wasted by failed preflight	engineering remediation
utilization by workload class	capacity planning
reserved-window idle percentage	reservation policy
cost per learning milestone	portfolio review

H.7 Governance metrics

Metric	Decision it supports
overdue access reviews	security risk management
stale exceptions	reliability and compliance debt
unreviewed external claims	claims governance
unclosed incident actions	operational maturity
vendor benchmark recency	procurement confidence
PQC migration completion by system	quantum-safe program management

H.8 SLO template

Illustrative listing · yaml

quantum_slo:
  name: string
  workload_class: string
  objective: string
  measurement_window: string
  target_threshold: string
  exclusions:
    - planned maintenance
    - provider outage
  alert_policy: string
  owner: string
  review_cadence: string

H.9 Dashboard layout

Diagram loads as you read

H.9 Dashboard layout · Figure 4

View diagram source

flowchart TB
    Dashboard[Operations dashboard] --> Now[Now: health and incidents]
    Dashboard --> Flow[Flow: queue and throughput]
    Dashboard --> Quality[Quality: benchmark and workload fit]
    Dashboard --> Evidence[Evidence: provenance and review]
    Dashboard --> Cost[Cost: utilization and spend]
    Dashboard --> Risk[Risk: security and compliance]

A good dashboard shows whether the platform should keep accepting work. A bad dashboard shows impressive numbers that nobody uses.