operating-model / loops/renter-loop/diagnostics.md
id: renter-loop.diagnosticstype: domainstatus: activeversion: 2.0loop: renter-loop

Domain · Diagnostic feedback

What it covers

Producing the signals that drive improvement across the business — back into the Renter Loop's own activities, and outward to the Operator Loop and Platform Loop. This domain is what distinguishes a loop owner from a marketer.

Activities

  • Funnel analysis — drop-offs by stage, by channel, by unit type, by operator, by conversation pattern.
  • Channel diagnosis — which channels deliver converting renters, not just volume.
  • Concierge process-quality diagnosis — sampling conversations against the six process-quality dimensions (responsiveness, profile-updating, honest brokerage, specific reasoning, escalation judgement, feedback-loop capture) per ADR 0003. Owned by the conversation-health operation below.
  • Creative diagnosis — what messages are starting and progressing the right kinds of conversations.
  • Supply-quality diagnosis — which units convert when surfaced, which don't, why.
  • Surfacing findings, with proposed actions, to the relevant other loops.

Operations

Weekly funnel review operation

Signal: Week boundary. Steps: Pull funnel data → identify the largest drop-off → diagnose root cause → action this week (own the fix or surface to another loop). Closed state: Action taken or surfaced. Cadence: Weekly; the standing input to the cross-loop sync.

Cohort revenue analysis operation

Signal: Month boundary. Steps: Match channel spend to lease revenue at cohort level → produce true CAC by channel. Closed state: CAC by channel produced; input to next month's renter-loop.demand.weekly-spend-review. Cadence: Monthly.

Supply-quality signal operation

Signal: Pattern detected (e.g. units in segment X have zero conversion despite high match volume). Steps: Packaged as a signal with evidence → sent to Operator Loop with proposed action → tracked through to resolution. Closed state: Signal resolved in the Operator Loop (action taken, or declined with logged reason). Cadence: Continuous.

Conversation health operation

Signal: Sampling cadence fires. Steps: Sample of recent conversations reviewed → quality dimensions assessed (continuity across mediums, tone, accuracy, escalation correctness, time to response) → patterns surfaced → action taken on automation, escalation thresholds, or platform infrastructure. Closed state: Patterns surfaced; action taken. Cadence: Weekly.

Platform blocker operation

Signal: Recurring exception requires platform change. Steps: Spec written → submitted to Platform Loop queue → tracked through to ship and validation. Closed state: Platform change shipped and validated against original blocker. Cadence: Event-driven.

Notes

Because the concierge is the defining feature (ADR 0003), diagnostic work centres on whether the concierge process is operating correctly, not on an output-level "match quality" claim. An earlier draft of Domain D carried a match-quality-analysis operation that attempted to score matches retrospectively against post-signature outcomes; it has been retired because Rentiful cannot produce post-signature signal at meaningful sample yet, and a score computed without that signal is an unmeasurable output claim.

What replaces it is the conversation health operation, which samples live conversations against the six process-quality dimensions defined in ADR 0003 — responsiveness, profile-updating, honest brokerage, specific reasoning, escalation judgement, feedback-loop capture. Thresholds and tripwires live in renter-loop/metrics.md. Outcome-level measurement (retention, post-move satisfaction, referral) is held as a future instrumentation target; it is added to this domain when the signal exists.

Comment on GitHub →