Other Metrics¶

This section documents the supporting readiness diagnostics used within the Electric Barometer framework, excluding Cost-Weighted Service Loss (CWSL), which is documented separately.

These metrics do not compete with CWSL. Instead, they decompose readiness behavior along dimensions that CWSL alone cannot isolate, such as reliability, severity, tolerance responsiveness, and structural failure modes.

Role of supporting metrics¶

Electric Barometer does not rely on a single scalar metric to characterize forecast readiness.

Operational failure modes are multidimensional: - some systems fail often but shallowly, - others fail rarely but catastrophically, - some respond to adjustment, - others are structurally insensitive.

The metrics in this section exist to diagnose these patterns explicitly, rather than forcing them into a single objective.

Relationship to CWSL¶

Cost-Weighted Service Loss (CWSL) measures aggregate, asymmetric cost exposure. It answers:

How much effective throughput is lost under a declared cost structure?

The metrics documented here answer complementary questions, such as:

How often does the forecast fail to cover demand?
How severe are failures when they occur?
How does behavior change under tolerance or adjustment?
Are failures structural or tunable?

They are intended to be interpreted alongside CWSL, not instead of it.

Metric categories¶

The metrics in this section fall into four broad categories.

Reliability metrics¶

These metrics describe how often forecasts avoid failure, independent of magnitude.

Examples include: - No–Shortfall Level (NSL) — frequency of full coverage, - tolerance-based hit rates (e.g., HR@τ).

Reliability metrics are sensitive to occurrence, not severity.

Severity metrics¶

These metrics describe how bad failures are when they occur.

Examples include: - Underbuild Depth (UD) — conditional shortfall magnitude.

Severity metrics isolate tail behavior that is obscured by frequency-based measures.

Tolerance and responsiveness metrics¶

These metrics describe how forecasts behave under admissible perturbation, such as: - changes in tolerance, - bounded readiness adjustment, - or cost asymmetry sweeps.

They are used to assess responsiveness, not to optimize outcomes.

Structural diagnostics¶

These diagnostics do not measure performance at all.

Instead, they determine whether: - evaluation semantics are admissible, - readiness interventions are structurally meaningful, - or policy application is valid.

Examples include: - Demand Quantization Compatibility (DQC), - Forecast Primitive Compatibility (FPC).

These diagnostics gate interpretation and policy; they are not scored or ranked.

Interpretive boundaries¶

All metrics documented here share the following properties:

they are evaluative, not prescriptive,
they rely only on observable quantities,
they do not imply operational action,
and they must be interpreted under governed unit semantics.

No metric in this section should be optimized directly.

How these metrics are used¶

In practice, these metrics are used to:

explain why two forecasts with similar CWSL behave differently,
identify structural incompatibility before adjustment or deployment,
support governance decisions with explicit evidence,
and communicate readiness risk in operational reviews.

They provide resolution, not decisions.

Use the pages in this section to understand individual metrics in detail.

For composite readiness evaluation and cost-sensitive analysis, refer to: - Cost-Weighted Service Loss (CWSL), - and Forecast Readiness Score (FRS).

Summary¶

CWSL captures cost exposure. The metrics in this section explain why that exposure arises.

Together, they form a diagnostic stack that supports defensible readiness evaluation without collapsing interpretation into a single number.