Purpose
Engineered interdependence (1.2.5) named Goodhart's Law as a design hazard and identified proxy capture as the mechanism by which well-designed incentive structures degrade. It forward-referenced this sidebar for the full diagnostic tool. The Goodhart map is that tool: a structured process for detecting whether an institution's metrics have decoupled from the purposes they were designed to track.
The Goodhart map does not replace the hazard scan constructed in the article. It operationalizes the scan's second diagnostic question — Which metrics have become targets? — with a step-by-step process that any institution can apply to its own performance measurement system.
The Four-Stage Diagnostic
Stage 1 — Map the Metric to Its Purpose
Every metric was designed to measure something. The first stage recovers that design intent: what reality is this metric supposed to track? What institutional purpose does it serve?
The mapping requires specificity. "Measures performance" is not sufficient. The question is: What would have to be true about the world for this metric to be high? If the answer is "the institution is fulfilling its purpose," the metric is properly grounded. If the answer is "the participants are producing the number," the metric has already begun to decouple.
Three diagnostic questions:
- What institutional outcome was this metric designed to track?
- If the metric improved by 20% tomorrow, would the institutional outcome necessarily improve as well?
- Can the metric improve without the underlying outcome improving — and if so, through what mechanism?
Stage 2 — Identify the Gaming Pathways
Every metric with consequences creates an incentive to produce the number. The second stage maps the pathways through which the metric can be satisfied without satisfying the purpose it was designed to measure.
Gaming pathways fall into three structural categories.
Definitional gaming. Participants exploit ambiguity in how the metric is defined or measured. If the metric counts "cases resolved" but does not define resolution rigor, participants can resolve cases superficially to increase the count. The structural response: tighten the definition, specify what counts, audit the classification process.
Boundary gaming. Participants shift activity across measurement boundaries — between reporting periods, between organizational units, between metric categories — to improve the number without changing the underlying reality. The structural response: measure across boundaries, not within them; use rolling averages rather than period-end snapshots; cross-reference metrics across units.
Substitution gaming. Participants shift effort from unmeasured activities to measured ones — even when the unmeasured activities are more important. The structural response: ensure that the metric ecosystem covers the full range of institutional purposes, not merely the purposes that are easiest to quantify.
Stage 3 — Test for Decoupling
The third stage applies a direct test: has the metric already decoupled from the purpose it was designed to track?
Four decoupling signals:
Metric-reality divergence. The metric is improving but the stakeholders closest to the institutional output report declining quality or relevance. When the numbers look good and the people look concerned, the metric is tracking something other than reality.
Gaming normalization. Participants discuss gaming strategies openly and without embarrassment. When "meeting the target" has become a distinct activity from "doing the work," decoupling is advanced.
Audit avoidance. Institutional leadership resists proposals to audit the metric's relationship to the underlying purpose. Resistance to audit is a structural signal: the parties who benefit from the current metric have an interest in not discovering that it has decoupled.
Metric proliferation. The institution responds to concerns about metric gaming by adding more metrics — without retiring the gamed ones. Metric proliferation without metric retirement produces a measurement environment in which participants optimize across a growing portfolio of targets, none of which tracks reality.
Stage 4 — Remediate or Replace
The fourth stage addresses the finding. If the metric has decoupled, two structural responses are available.
Remediation. The metric can be redesigned to close the gaming pathways identified in Stage 2. This works when the underlying purpose is well-defined and the gaming was an exploitation of design gaps rather than a fundamental mismatch between what is measurable and what matters. Remediation must include a transition period during which the old and new metrics run in parallel — to verify that the new metric tracks reality better than the old one.
Replacement. The metric is abandoned and replaced with a different measurement approach. This is necessary when the metric's design is fundamentally incompatible with the purpose it was supposed to track — when the act of measuring distorts the thing being measured. Replacement is structurally more disruptive and requires explicit institutional commitment, because participants who have optimized their behavior around the old metric will experience the transition as destabilizing.
Both responses must include a metric audit commitment — a scheduled reassessment (per the review ritual cadence from sidebar 1.2.7-3) to verify that the new or redesigned metric has not itself begun to decouple.