Objectuve Docs

Appearance

PRD: Optional Enneagram Assessment for Coach Personalization ​

Product: Objectuve (Enkidu) Feature: Enneagram personality assessment (opt-in) Status: ✅ Shipped (v1.18, 2026-05-14) Date: 2026-05-04 Owner: Josh Lockhart

Context ​

Objectuve's Coach already personalizes feedback using UserDetail jsonb fields (coach_persona, coach_tone_*, coach_depth, coach_focus_*, coach_rhythm) and AiEmployeeMemory injected into the system prompt at agent_runner/src/agent/promptBuilder.ts. These capture how a user wants to be coached (tone, frequency, focus) but not the underlying motivational structure — why a particular user resists, plateaus, or thrives.

The Enneagram is one of the few popular typologies that explicitly maps motivation and core fear/desire patterns, which translate cleanly into coaching framings (e.g., a Type 1 needs permission to be imperfect; a Type 7 needs help finishing rather than starting; a Type 4 needs continuity across mood swings). Adding an optional assessment gives Coach a richer, durable trait layer without requiring users to articulate it themselves.

The reference UX (Multifactor Enneagram) is a 40-question forced-choice format that resolves to scores across all 9 types plus wing/tritype/instincts. We'll mirror this structure but slim the result model and adapt the visual language to Objectuve's brand.

This is strictly optional and must not interfere with core flow. The product bet: users who self-select into a 5-minute assessment are signaling appetite for deeper coaching, and Coach output for those users should feel measurably more apt.

Goals & Success Metrics ​

Primary goals

  1. Give Coach a durable trait input that improves perceived feedback quality for opted-in users.
  2. Offer the assessment as a self-knowledge moment, not a gate or default.

Metrics

MetricTargetNotes
Assessment completion rate (of users who start)≥ 65%40 questions; abandonment is the main risk
Eligible-user start rate (dashboard card → start)≥ 18%Card shown after 7 days of activity
Coach feedback "this felt accurate" rating, opted-in vs. not+10pp liftMeasured via existing thumbs/feedback signal on Coach replies
Retake rate within 90 days< 15%High retakes signal poor confidence in result

Guardrails

  • Zero impact on signup/onboarding completion rate.
  • No blocking modals, no notifications nudging the assessment.
  • Coach quality for non-opted-in users must not regress (no shared prompt changes that assume Enneagram presence).

Non-Goals ​

  • Therapeutic claims or clinical framing. The assessment is self-discovery, not diagnosis.
  • Sharing types socially (no leaderboards, no "find your Type 4 friends," no community filtering).
  • Replacing the existing coach_persona/tone preferences. Enneagram is additive context, not a substitute.
  • Mandatory step in any flow.

User Stories ​

  • As a 2-week-old user who's curious, I see a dismissible card on my dashboard offering a 5-minute personality assessment that will help my Coach understand me better.
  • As a privacy-conscious user, I can permanently dismiss the dashboard card and still find the assessment in Settings → Personalization if I change my mind.
  • As an opted-in user, my Coach references my type subtly (framing, not labels — e.g., "I know you'd rather start something new, but…" for a Type 7) without making me feel reduced to a category.
  • As a user whose self-perception has shifted, I can retake the assessment from Settings, and prior results are preserved as history.

Scope ​

Assessment format ​

  • 40 forced-choice question pairs, mirroring the reference test's structure.
  • Single screen per question, two statement cards, tap to choose, auto-advance.
  • Progress indicator (e.g., "12 / 40").
  • Back button allowed; no required fields.
  • Resumable: partial state persisted client-side; if abandoned, "Continue assessment" replaces "Start" on the dashboard card and settings entry.
  • Question content authored fresh in Objectuve's voice rather than copied verbatim from the reference. 40-pair set with each pair scoring +1 to one of two competing types (each type appearing in ~9 pairs to balance exposure).

Result model ​

  • Full 9-type scores (0–100 normalized), plus computed:
    • Dominant type (1–9)
    • Wing (adjacent type with higher score)
    • Tritype (top type from each center: body 8/9/1, heart 2/3/4, head 5/6/7)
  • Stored as a new enneagram_assessments table (history retained per user).
  • Latest assessment is what Coach uses; prior results visible in Settings as a timeline.

Coach integration ​

  1. System-prompt injection — Add a "User personality profile" block to the agent system prompt builder when the user has a completed assessment. Format: dominant type + wing, top-3 type scores, and a curated 1–2 sentence framing per dominant type from a static lookup (authored once, brand-voice approved). Cap at ~250 tokens.
  2. Feedback templating — Where Coach generates streak-repair, milestone, and missed-day messages, expose the dominant type to template selection so type-aware framings can be authored over time. Initial release ships with system-prompt injection only; templating lands in a follow-up.
  3. User-visible self-knowledge — Settings → Personalization shows the user their result with a short, non-pathologizing description.

Discoverability ​

  1. Dashboard card — Appears once user has 7+ days of activity AND has not dismissed/completed the assessment. Dismissible (permanently). Lives alongside existing cards in Dashboard.vue with a v-if="shouldShowEnneagramCard" gate, mirroring StreakRepairCard / GettingStartedCard patterns.
  2. Settings → Personalization — New tab in Settings.vue (added to the existing tabs array). Always available. Shows current result, retake button, and assessment history.

Out of scope (this version) ​

  • Type-aware feedback templating (queued as v2).
  • Explicit type filters in community/discovery surfaces.
  • Coach acknowledging the assessment unprompted in conversation.
  • Exporting results.

Design Notes ​

  • Tone: the assessment is framed as a "5-minute self-portrait for your Coach," not a personality test. No quiz-app energy.
  • Visual: glassmorphism question cards consistent with docs/brand/brand.md. Single primary CTA per screen. Progress bar uses the primary gradient.
  • Result reveal: one screen, dominant type front and center with a 2–3 sentence description, expandable "see full breakdown" for the 9-type bars. No pop-ups, no confetti — quiet and respectful.
  • Copy guardrails: never use the word "AI" in user-facing copy. Use "Coach." Avoid clinical/diagnostic language.

Implementation Outline (high-level — not a phase plan) ​

Backend

  • Migration: enneagram_assessments (user_id, scores jsonb, dominant_type, wing, tritype, completed_at, soft-delete via acts_as_paranoid).
  • Model inheriting from PublicRecord.
  • SubmitEnneagramAssessment interaction (validates 40 answers, computes scores, persists row).
  • GraphQL: enneagramAssessment query (latest), enneagramAssessmentHistory query, submitEnneagramAssessment mutation.
  • Extend agent_runner/src/agent/promptBuilder.ts system prompt with a "User personality profile" section sourced from the latest assessment (passed in via the Rails-side run payload).

Frontend

  • New view: EnneagramAssessment.vue (question runner, resumable state via localStorage).
  • New components: EnneagramQuestionCard.vue, EnneagramResultCard.vue, EnneagramHistoryList.vue.
  • EnneagramPromptCard.vue on dashboard (gated by 7-day activity + opt-out flag).
  • New Settings tab: panel-personalization.
  • Question bank: src/constants/enneagramQuestions.ts (40 pairs, scoring map).

Telemetry

  • Events: enneagram_card_shown, enneagram_started, enneagram_completed, enneagram_abandoned (with question index), enneagram_retake_started, enneagram_dismissed.
  • Coach feedback signal already exists; segment by has_enneagram_assessment in dashboards.

Rollout ​

  1. Internal QA — admin/demo accounts complete the flow end-to-end; verify Coach prompt receives the profile block.
  2. Staging soak — at least one full real run with opted-in user, confirm Coach output reads naturally.
  3. Production — ship behind no flag (it's strictly opt-in by surface). Monitor completion rate and Coach feedback delta for 2 weeks before deciding whether to invest in v2 templating.

Open Questions ​

  1. Question bank authorship: adapt the reference 40 pairs (with significant rewriting for tone/IP separation) or commission an original set? Recommend original — cheaper than the IP/tone risk.
  2. Tritype value: surface tritype in the user-facing result, or keep it internal as Coach context only? Lean: internal only at launch (less to explain).
  3. Where exactly does the prompt block sit in buildSystemPrompt? Above or below AiEmployeeMemory? Decide during planning — likely above, since it's durable trait context.

Verification (post-implementation) ​

  • Submit assessment as test user → confirm enneagram_assessments row created with valid 9-type scores summing to expected total.
  • Trigger a Coach run for that user → fetch the resolved system prompt → confirm "User personality profile" block is present and well-formed.
  • Dashboard card visibility: seed a user with 6 days of activity (card hidden), 7 days (card shown), dismissed (card hidden), completed (card hidden).
  • Settings → Personalization tab: shows latest result, history list, retake CTA. Retake produces a new row, latest is what Coach reads.
  • Smoke test: full assessment completion path on staging.

Critical Files (anchors for future planning) ​

Last updated: 2026-07-21 — Phase 100 audit. PRD for shipped feature (v1.18, Phases 77–81) archived to completed/.

Loading…