Direction-Sensitive Metric Dashboard¶
Direction-aware monitor — instantiates Directed Asymmetry Mapping and Calibration
Tracks a matched pair of metrics — one per side of the relation — and watches the gap between them, so a drift toward one side is caught while it is still small.
Once a relation is known to be oriented, the danger is not that it is unequal today but that the inequality creeps. The Direction-Sensitive Metric Dashboard is the standing instrument for that: for each relation it carries a matched pair of metrics — the same outcome measured on both sides — and tracks the difference between them over time as a first-class number. Its defining move is refusing to report a single blended figure: a combined "health" metric can look flat while one side quietly erodes, so the dashboard always keeps the two sides disaggregated and side-labelled, and raises a flag when the gap drifts past a set band. It is the archetype's early-warning organ, distinct from the tools that map, judge, or govern the asymmetry.
Example¶
A food-delivery platform runs a paired dashboard for its two-sided market. On the courier side it watches earnings per active hour, wait-between-orders, and cancellation exposure; on the customer side, delivery time, price paid, and refund rate. Each pair sits on one row, with the gap plotted, not just the two lines. A pricing tweak aimed at speeding up deliveries ships in Q2. The blended "marketplace health" score barely moves — but the dashboard shows the courier earnings-per-hour line bending down while the customer delivery-time line improves, and the gap crossing its amber band within weeks.
Because the drift is caught early, the platform can respond while it is a few percent, not after couriers have started leaving. The dashboard does not tell them the change was unfair — only that the relation is sliding one way faster than intended, which is exactly the signal the rest of the appraisal needs.
How it works¶
- Pair the metric, don't blend it. Each measure is defined once and captured separately for each side, so the two are always comparable and never silently averaged into one figure.
- Make the gap the headline. The tracked quantity is the difference (or ratio) between the sides, so the object being monitored is the asymmetry itself, not either side's level.
- Band and alert on drift. Thresholds sit on the gap's movement, so a slow, one-directional slide trips a flag before any single reading looks alarming.
Tuning parameters¶
- Metric pairing — which outcomes get a matched pair. Pairs must be genuinely comparable across sides; a mismatched pair (cost for one side, satisfaction for the other) produces a gap that means nothing.
- Drift band width — how far the gap may move before alerting. Tight bands catch creep early but cry wolf on noise; wide bands stay quiet until the asymmetry is entrenched.
- Aggregation window — the time base over which drift is measured. Short windows react fast and jitter; long windows are stable but let slow erosion run.
- Segmentation depth — whether each side is one number or split by subgroup, since a stable side-average can hide a reversing subgroup underneath.
When it helps, and when it misleads¶
Its strength is catching direction and drift that a blended metric hides — the aggregate that stays healthy while one side erodes is a textbook Simpson's-paradox trap, and side-paired metrics are the standard antidote.[1] It turns "something feels off between us and them" into a dated, banded signal a team can act on before the gap becomes a crisis.
Its failure modes are the monitor's. A dashboard measures the gap but says nothing about whether the gap is warranted — a clinician-patient relation should show an information asymmetry — so a drift alert is a prompt to look, not a verdict. It is easily gamed by metric choice: pick the paired measure that flatters the side you favour and the gap stays reassuringly small, the monitoring equivalent of running the analysis backwards. And a pair that is not truly comparable manufactures alarms or false calm. The discipline is to fix the metric pairs before the numbers arrive, hand every persistent drift to a relevance judgement rather than acting on the gap alone, and keep subgroups visible under each side.
How it implements the components¶
The dashboard fills the archetype's measurement-over-time slots — what a monitor, not a map or a policy, can fill:
side_specific_metric_pair— its core structure: one metric defined once and carried separately for each side of the relation.asymmetry_drift_monitor— it makes the side-to-side gap a tracked, banded quantity and alerts when it slides one way.
It does not build the underlying map of who-leans-on-whom (oriented_relation_map → Directed Relation Matrix), judge whether a measured gap is legitimate (relevant_difference_warrant → Relevant Asymmetry Test), or set what either side must do about it (role_specific_obligation_map → Role-Specific Policy Table).
Related¶
- Instantiates: Directed Asymmetry Mapping and Calibration — it is the appraisal's running early-warning gauge on the mapped asymmetry.
- Consumes: Directed Relation Matrix tells it which sides to pair and which direction "worse" points in.
- Sibling mechanisms: Directed Relation Matrix · Relevant Asymmetry Test · Asymmetry Dimension Scorecard · Asymmetry Sunset Review · Role-Specific Policy Table
Editorial Notes¶
Form Classification¶
Form family: Monitoring, Sensing & Alerting
Rationale: Direction-Sensitive Metric Dashboard operates as an ongoing sensing arrangement that repeatedly observes actual state and surfaces changes or alerts because it tracks a matched pair of metrics — one per side of the relation — and watches the gap between them, so a drift toward one side is caught while it is still small.
Independent corroboration: The frozen evidence defines Direction-Sensitive Metric Dashboard as 'Tracks a matched pair of metrics — one per side of the relation — and watches the gap between them, so a drift toward one side is caught while it is still small', so its operative form is Monitoring, Sensing & Alerting.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Data Science & Analytics
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Operational analytics cohered continuously refreshed disaggregated measures, paired gaps, tolerances, and alerts into an inspectable decision surface.
Related originating lineages:
- Organizational & Management Science — Management-control practice supplied owned dashboards, tolerance bands, and escalation.
- Experimental Design & Statistics — Statistical disaggregation established why pooled measures can conceal or reverse directional change.
Review resolution: The mechanism is primarily a live analytic display, making data science the best home; statistical disaggregation and management control are independently formative.
Attribution caveat: The anti-aggregation principle is statistical, while the standing paired-dashboard artifact is a data-science and management implementation.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
- NIST AI RMF Playbook
- Simpson: The Interpretation of Interaction in Contingency Tables
- GAO Green Book: Standards for Internal Control
References¶
[1] Simpson, E. H. “The Interpretation of Interaction in Contingency Tables”. Journal of the Royal Statistical Society: Series B (Methodological) 13(2), 238–241 (1951). Shows how aggregating strata can hide or alter relationships visible in the disaggregated data. registry ↩ Show verification details
Supported in partVerified against the publisher's abstract
The 1951 paper is the origin of the interaction-in-contingency-tables result behind Simpson's paradox, but its abstract does not state the aggregation trap or the side-paired antidote.
“The definition of second order interaction in a (2 × 2 × 2) table given by Bartlett is accepted, but it is shown by an example that the vanishing of this second order interaction does not necessarily justify the mechanical procedure of forming the three component 2 × 2 tables and testing each of these for significance by standard methods.”