Skip to content

Receiver Response Dashboard

Response monitor — instantiates Signal Value Preservation

Tracks what receivers actually do when the signal fires — acknowledge, act, dismiss, or ignore — and how that response rate drifts as the channel keeps firing.

Receiver Response Dashboard is a live monitor of receiver behavior: for each firing, does the receiver open it, act on it, dismiss it, or ignore it — and how are those rates trending as the signal fires more often? Its distinguishing idea is that it measures the observed, behavioral response, the ground-level evidence of whether a signal still moves anyone, and it reads a falling response rate as information — often rational receiver learning — rather than as laziness. Where the fatigue survey asks receivers how they feel, this dashboard watches what they do; where a base-rate dashboard tracks whether the signal is accurate, this tracks whether it still lands.

Example

A bank's anti-money-laundering system flags "suspicious" transactions for investigator review. The dashboard tracks, per alert type, the share investigators escalate versus clear-without-action, over time. One rule's escalation rate had fallen from ≈30% to under 3% — investigators had learned it fired mostly on false positives and were now clearing it in seconds, effectively ignoring it. That behavioral collapse, visible only in the response data and not in whether the rule was "correct" on paper, told stewards the signal had devalued into background noise and needed retuning or retirement. The falling response was not negligence; it was rational inattention to a signal that had stopped separating cases.[1]

How it works

The distinctive move is to instrument the receiver's action on each firing — acknowledged, acted, dismissed, ignored — and trend those rates by signal type and receiver segment, treating a declining response as a measured symptom of devaluation. It observes behavior in the running system; it neither asks receivers how they feel nor judges whether the signal was accurate.

Tuning parameters

  • Response taxonomy — how finely actions are classified (opened vs. acted vs. dismissed vs. auto-cleared). Finer distinguishes "ignored" from "handled fast" but needs richer instrumentation.
  • Trend window — how much history frames "normal" response. Long windows reveal slow erosion; short ones catch sudden collapse.
  • Segmentation — by signal type, receiver, or time of day. Fine slices locate where response is dying but dilute sample size.
  • Alert-on-drift — the drop in response rate that trips a flag. Sensitive catches erosion early but fires on noise.
  • Attribution guardrail — how carefully a low response is read as devaluation versus a genuinely quiet period, so "nothing happened" isn't misread as "ignored."

When it helps, and when it misleads

Its strength is that it is the objective, behavioral ground truth of whether a signal still recruits action, and it reframes a falling response as rational learning worth heeding rather than a discipline problem. Its failure mode is that response rate conflates "ignored because worthless" with "correctly not acted on because nothing was wrong" — a low action rate can be the right outcome. The classic misuse is optimizing to raise the response rate — making the signal louder or scarier — when the honest fix is to issue it less often. The discipline that guards against this is to always read response alongside signal accuracy (the Base-Rate and Precision Dashboard) and receiver sentiment (the Receiver Fatigue Survey): a low response on an accurate signal means something very different from a low response on a noisy one.

How it implements the components

  • receiver_interpretation_model — it infers, from behavior, what receivers take the signal to mean and what action they take on it.
  • habituation_and_adaptation_monitor — trending the response rate is precisely the measure of receivers habituating and adapting as the channel fires more.

It does not ask receivers how they feel (receiver_feedback_panel, receiver_attention_budget) — that is the Receiver Fatigue Survey — and it does not establish the signal's causal effect versus no signal (false_alarm_and_miss_ledger), which needs a Signal A/B Test or Holdout.

References

[1] Rational inattention — the idea from information economics (Christopher Sims) that attention is a scarce resource receivers allocate deliberately, rationally ignoring signals whose expected information does not repay the cost of attending. It is why a falling response rate is often a correct receiver judgment, not a failure of diligence. ↩