SOP-096 — Learner health review & intervention

SOP · owner: Customer Success (Customer Success Director). Layer: SOP/Runbook (docs/canon.md). The learning model lives in docs/concept.md; the requirements↔schema boundary in docs/units/data-analytics/charter.md — not restated here. Pre-launch note: no health-score telemetry exists yet (ADR-0006); this authors the review stub-first against the planned health score and activates at the Operate/Grow gate when real engagement data flows. Unit boundaries: ../units/customer-success/charter.md.

Interface Contract

Inputs: learner health-score summary / engagement telemetry ← from Data & Analytics Director · trigger: briefing posted
Inputs: community retention signal ← from Community Manager · trigger: community signal raised
Outputs: refined health-score requirements (new signal / threshold / play) → to Data & Analytics Director · trigger: health-score requirements issued
Outputs: field-feedback brief (at-risk patterns worth a product change) → to Innovation Director · trigger: field-feedback routed
Outputs: research-grade signal (a disengagement pattern needing a formal study) → to Design Research Director · trigger: field-feedback routed
Outputs: approved health-review report → to Archivist · trigger: artifact approved
Outputs: health-review SOP / report → to Quality & Risk Director · trigger: review requested
Handoff trigger (in): the review cadence is due, or a churn / at-risk health signal fires
Handoff trigger (out): at-risk learners are triaged, an intervention is assigned and executed, and 14-day re-engagement is tracked and recorded
Definition of Done before handoff:
- [ ] Health-score summary pulled from Data & Analytics output (not invented)
- [ ] At-risk learners flagged and grouped by disengagement pattern (never-activated / early-dropout / plateau / lapsed)
- [ ] The right intervention play selected per group — or a new play drafted if the pattern is novel
- [ ] Intervention assigned and executed (direct outreach / content recommendation / Community Manager engagement)
- [ ] Outcome tracked — did the learner re-engage within 14 days?
- [ ] A silent learner is treated as at-risk, never as satisfied
- [ ] No learner contact data in the report or any edge (PA-only)
- [ ] Pre-launch status stated; no fabricated health scores or re-engagement figures
- [ ] Learnings captured to docs/knowledge/ if any
- [ ] Report written to reports/cs-health-<YYYY-MM>-<segment>.md; handed to the Archivist (0 drift)

Purpose

On a regular cadence (or when a churn signal fires), turn engagement/progress data into action: flag at-risk learners, group them by disengagement pattern, run the right intervention play, and measure whether it worked — before disengagement compounds into churn. A silent learner is a risk to act on, not a neutral signal.

When to use

The health-review cadence is due, or a churn / at-risk health signal fires from Data & Analytics output or a community retention signal.

Flow

flowchart LR
  A["Pull health-score summary<br/>(Data &amp; Analytics output)"] --> B["Flag at-risk learners"]
  B --> C["Group by pattern<br/>(never-activated / dropout / plateau / lapsed)"]
  C --> D{"Pattern in the<br/>playbook?"}
  D -->|yes| E["Select intervention play"]
  D -->|no| F["Draft a new play"]
  E --> G["Assign + execute<br/>(outreach / content / Community Mgr)"]
  F --> G
  G --> H["Track 14-day<br/>re-engagement"]
  H --> I["Report → Archivist;<br/>refine requirements → Data &amp; Analytics"]
  classDef s fill:#1d4ed8,color:#fff,stroke:#1e3a8a
  classDef d fill:#0f766e,color:#fff,stroke:#134e4a
  class A,B,C,E,F,G,H,I s
  class D d

Text fallback: pull the health-score summary from Data & Analytics output → flag at-risk learners → group them by disengagement pattern (never-activated / early-dropout / plateau / lapsed) → is the pattern in the playbook? (yes → select the intervention play; no → draft a new play) → assign and execute (direct outreach, content recommendation, or Community Manager engagement) → track 14-day re-engagement → write the report to the Archivist and refine the health-score requirements for Data & Analytics if a new signal/threshold/play emerged.

Steps

Pull the health summary. Take the latest health-score summary from Data & Analytics output / reports (briefing posted). Do not invent scores — pre-launch, this is stub-first against the planned score.
Flag at-risk learners. Everyone below the at-risk threshold defined in the health-score requirements.
Group by disengagement pattern. never-activated · early-dropout · plateau · lapsed. The pattern determines the play.
Select or draft the intervention play. Pull the matching play from the success playbook; if the pattern is novel, draft a new play (it becomes a playbook candidate).
Assign and execute. Direct outreach, a content recommendation, or a Community Manager engagement (commissioned, reviewed by you). Relationship messaging is CS’s; if a community response is needed, the Community Manager drafts, never posts.
Track the outcome. Did the learner re-engage within 14 days? Record yes/no per intervention — this is how a play earns its place or gets retired.
Refine and report. If a new signal, threshold, or play emerged, hand refined requirements to Data & Analytics (health-score requirements issued); route any product-worthy pattern to Innovation and any study-worthy pattern to Design Research (field-feedback routed); write the report to reports/cs-health-<YYYY-MM>-<segment>.md and hand to the Archivist.

Roles / approvers

Run by the Customer Success Director. SOP approval is the Director’s own (owner’s Director per canon RACI — for a Lead, the cluster head). Data & Analytics supplies the health summary and instruments refined requirements; the Community Manager executes commissioned community interventions; relationship-critical delivery risk is flagged to the Project Manager (who does not own the plan).

Guardrails

A silent learner is at-risk — intervention before assumption; silence is never read as satisfaction.
Requirements, not schemas — refine what to measure and the thresholds; DA owns the scoring mechanism.
Field feedback, not formal research — route study-worthy patterns to Design Research; do not run the study yourself (SOP-053 is theirs).
No personal data — learner contact data is PA-only; no CS edge carries it.
State the pre-launch status — stub-first; no fabricated health scores or figures.
Do not restate docs/concept.md or docs/units/data-analytics/charter.md — link to them.

Job aid: JA-096-at-risk-triage.md. See also: SOP-095 (learner onboarding) · docs/units/customer-success/charter.md · docs/sop/SOP-053-telemetry-feedback-synthesis.md (the Design Research synthesis seam).