Skip to main content
Correspondia
Paper overview

Metric and Indicator Architecture

Metrics and indicators are often treated casually in organizations. This paper distinguishes them carefully. A measure is an execution result for a specific object, context, and time. A metric is a governed reusable specification for producing and representing measures of a defined property. An indicator is a theoretically justified, context-bounded interpretation-bearing signal derived from one or more measures or metrics. A profile is a structured collection of non-equivalent measures or indicators that preserves identity, level, time, evidence, uncertainty, and limitations. A score is a numerical or ordinal representation produced by a specified rule. An index is an aggregated representation formed under explicit weighting, normalization, compensability, and validity assumptions.

Figure 3 places these terms in sequence and shows the downstream assessment boundary.

Figure 3

Metric, Indicator, and Profile Architecture

FIG-0018 depicts Metric, Indicator, and Profile Architecture. It represents Construct, Property, Metric specification, Metric execution, Measure, Indicator, Profile, Index gate, Composite index, Assessment boundary, Assessment judgment, Recommendation. The intended relationships are: Metrics operationalize constructs; indicators signal interpretation; profiles preserve heterogeneity; index and assessment paths require gates. The figure should be read with this boundary: Indicators and profiles are not assessment judgments or recommendations.
Figure 3. Metric, Indicator, and Profile Architecture. Main Figure 3 distinguishes a measure from a metric specification, an indicator, a profile, an index, and an assessment judgment. The assessment boundary marks the point at which descriptive measurement can support but not replace evaluative judgment.

The architecture has three core principles. First, construct primacy: metrics must operationalize a defined construct, dimension, or property. Data availability alone cannot create a valid metric. Second, specification visibility: a metric must preserve object, level, evidence, procedure, temporal frame, validation status, uncertainty, and permitted interpretation. Third, profile preservation: multiple non-equivalent indicators should remain visible until aggregation is justified.

Consider a potential measure of decision-to-work correspondence. A weak design would count whether a work artifact links to a decision. A stronger design would specify the decision representation, the work representation, the relationship type, the evidence source, the temporal interval, the procedure used to determine correspondence, whether semantic or dependency preservation is observed or inferred, what evidence is missing, what confidence attaches to the result, and what interpretations are permitted. The result may then inform an indicator, such as "decision implementation trace state," but the indicator is still descriptive unless assessment criteria are separately applied.

This distinction prevents measurement from becoming hidden evaluation. Thresholds, benchmarks, and assessment criteria are not descriptive metrics unless the framework explicitly authorizes them as part of a governed use case. A measure can describe a correspondence state without declaring the state acceptable, risky, failed, or recommended for action. That judgment belongs to assessment, evaluation, diagnosis, or governance.