Sentinel Outcome Dashboard¶
Monitoring artifact — instantiates Proxy–Target Divergence Detection and Recalibration
A standing, owner-facing display that lines up the proxy against downstream outcome and harm signals so silent decoupling becomes visible at a glance.
Sentinel Outcome Dashboard is the loop's standing human interface — the always-visible surface that places the proxy side by side with the real-world outcomes it is supposed to be improving, so a human can see divergence emerging rather than reading about it in a quarterly report. Its defining move is juxtaposition of the leading proxy against lagging outcome and harm signals: not the proxy's own statistics in isolation, but the proxy next to complaints, incidents, satisfaction, and downstream damage, arranged so a growing gap between "the number is up" and "the world got worse" is immediately legible. It performs no statistical test and takes no action; its whole job is to make silent decoupling observable to the person responsible, gathering the outcome-side sentinels into one place where they cannot be ignored.
Example¶
A video platform's recommender is tuned on watch time, the proxy for user satisfaction. A sentinel outcome dashboard is built for the recommendations product owner. On the left it shows watch time, climbing steadily. But it deliberately refuses to show that alone. Beside it, on the same screen and the same timeline, sit the outcome and harm signals: unsubscribe rate, "not interested" and "don't recommend this channel" clicks, survey-reported regret, and support complaints mentioning "couldn't stop" or "wasted my evening." For months the panels move together and the picture is reassuring. Then they split — watch time keeps rising while regret surveys and unsubscribes climb with it. The dashboard proves nothing about why, but the visible divergence between the leading proxy and the lagging outcomes is what pulls the owner's attention and triggers a real investigation, long before the annual satisfaction study would have. The design principle it embodies is pairing leading indicators with lagging ones so the leading number can never be celebrated alone.[n1]
How it works¶
- Co-locate proxy and outcomes. Put the proxy and its downstream outcome/harm signals on one surface, aligned to a shared timeline, so the eye compares them directly.
- Foreground the gap, not the level. Design the display to make divergence between proxy and outcomes salient — a widening spread, a crossing — rather than celebrating the proxy's absolute level.
- Surface harm signals explicitly. Give complaints, incidents, regret, and adverse events dedicated, unmissable panels rather than burying them.
- Route to the owner. Attach the display to a named person whose job is to watch it and escalate, so a visible divergence becomes an action rather than a screenshot.
Tuning parameters¶
- Signal selection — which outcome and harm measures sit beside the proxy. Well-chosen lagging signals catch real decoupling; a thin or flattering selection produces a dashboard that only ever reassures.
- Salience of divergence — how aggressively the layout highlights gaps (color, alerts, spread bands). More salience catches attention faster but risks alarm fatigue.
- Refresh latency — how current the outcome signals are. Fresher signals shorten the gap between decoupling and notice but cost pipeline effort, and true outcomes are inherently laggy.
- Aggregation level — one org-wide view versus per-segment panels. Segmented views expose decoupling hidden in a healthy average but crowd the surface.
- Escalation binding — whether a visible divergence merely informs or is wired to a required review. Binding escalation prevents "seen and ignored."
When it helps, and when it misleads¶
Its strength is continuous, low-friction visibility for a human owner. It is the mechanism that keeps the target in view next to the proxy every day, so that a rising number can never be quietly mistaken for success while the outcomes it stands for erode. Placed in front of the accountable owner, it is what converts a slow decoupling into a noticed one.
Its failure mode is dashboard theater: a screen full of green that creates the feeling of oversight while measuring nothing that would actually contradict the proxy. If the outcome signals shown are themselves proxies, or are chosen because they tend to agree, the dashboard manufactures false confidence — and because it only displays, a divergence that everyone can see but no one is required to act on changes nothing. The classic misuse is treating "we have a dashboard" as evidence the proxy is safe. The guarding discipline is to curate the outcome panels adversarially (include the signals most likely to embarrass the proxy), refresh them honestly, and wire visible divergence to a required escalation rather than optional attention — and to send confirmed gaps to a real target-side check, since the dashboard can show a split but cannot adjudicate it.
How it implements the components¶
divergence_sentinel— the display is a sentinel: a visible, standing gap between the proxy and its outcomes serves as the human-legible trigger that something has decoupled.harm_and_consequence_monitor— it gives downstream harms, complaints, and adverse outcomes dedicated panels beside the proxy, keeping real-world consequences continuously in view.
This dashboard displays and juxtaposes; it does not run the statistical break-detection or set the trigger threshold — the decoupling_threshold_rule and regime_and_context_marker machinery is Drift and Change-Point Detection — and it does not execute a fix, the recalibration_or_replacement_path owned by Proxy Retirement Decision Record.
Related¶
- Instantiates: Proxy–Target Divergence Detection and Recalibration — it is the loop's standing, owner-facing window onto proxy-versus-outcome divergence.
- Consumes: Drift and Change-Point Detection — statistical flags feed the dashboard as one of the sentinels it surfaces to the owner.
- Sibling mechanisms: Drift and Change-Point Detection · Holdout Ground-Truth Audit · Incentive Impact Review · Proxy Retirement Decision Record · Proxy–Target Correlation Refresh · Reference-Standard Recalibration Review · Shadow Target Measurement · Triangulated Proxy Panel
Editorial Notes¶
Form Classification¶
Form family: Monitoring, Sensing & Alerting
Rationale: Sentinel Outcome Dashboard operates as ongoing observation, sensing, or alerting that detects and surfaces state without itself executing the response because it a standing, owner-facing display that lines up the proxy against downstream outcome and harm signals so silent decoupling becomes visible at a glance.
Independent corroboration: The frozen evidence defines Sentinel Outcome Dashboard as 'A standing, owner-facing display that lines up the proxy against downstream outcome and harm signals so silent decoupling becomes visible at a glance', 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: Universal
Rationale: Displaying proxy, downstream outcome, and harm series together is analytical monitoring designed to detect metric-target decoupling.
Related originating lineages:
- Medicine & Healthcare — Outcome and balancing-measure dashboards guard against improving a process metric while harming patients.
- Organizational & Management Science — Owner-facing scorecards make silent goal displacement accountable.
- Statistics & Experimental Design — Statistics, experimental design, and measurement theory supplies a parallel or contributing lineage for the mechanism's defining operation: a standing, owner-facing display that lines up the proxy against downstream outcome and harm signals so silent decoupling becomes visible at a glance.
- Ethics of Technology & AI Governance — Responsible-system governance requires monitoring harms and proxy drift beyond headline performance.
Review resolution: The blind reviewers agree that data_science is the primary origin and differ only on alternate origin disagreement, origin mode disagreement, domain reach disagreement. I preserve every independently explained alternate from both records rather than imposing a numeric cap. I retain cross_disciplinary_synthesis because the combined record shows material contributions from several lineages. The broader reach of universal records portability separately from historical provenance, and encyclopedia_synthesis=true preserves the affirmative synthesis judgment where either reviewer identified one.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; medium confidence.
Notes¶
[n1] Leading indicators move early but predict imperfectly; lagging indicators confirm outcomes but arrive late. Pairing them on one surface — so a rising leading proxy is always read against the lagging outcomes it is meant to cause — is a standard management-measurement discipline and the core design principle of this dashboard. ↩