Decision Dace

DACE Loop Automation (with Insight Transformation)

Core Principles

1. Define is direction, Analyze is evidence: Analysis without clear objectives and decisions without evidence are equally dangerous
2. Conclude authority belongs to humans, Execute tracking belongs to the system: AI provides options and boundaries, humans make final decisions, the system tracks execution effectiveness
3. A closed loop is complete: DACE is inseparable; Conclude without Execute is empty talk, Conclude without Analyze is gambling
4. Data is the starting point, insight is the endpoint, action is the purpose: Insights without actionable direction are just data displays
5. Narrative over jargon: Translate "p=0.001" into "99.9% confidence", so decision-makers can understand and act
6. Boundary annotation is more important than recommendations: Clearly marking what can be auto-executed vs. what needs human confirmation is more valuable than simple recommendations

Interaction Mode

🤖→👤 AI suggests, human approves

Input

| Input Item | Type | Required | Source | Description | |--------|------|------|------|------| | OKR Data | object | Yes | User provided | Objectives and key results, baseline values and target values | | KR Progress | object | Yes | output/pm-metrics-ops/analysis-anomaly/anomalyreport.json | Current progress and deviation analysis for each KR | | Experiment Results | object | Yes | output/pm-metrics-ops/experiment-execution/experimentresult.json | A/B test results, anomaly detection data | | Analysis Results | object | Yes | output/pm-metrics-ops/analysis-anomaly/anomalyreport.json | anomaly/funnel/retention reports | | Business Context | object | ○ | User provided | Product stage, team objectives | | Historical Insight Library | object[] | ○ | output/pm-metrics-ops/decision-dace/insightlibrary.json | Avoid duplication |

Execution Steps

DACE Four Phases

┌────────────────────────────────────────────────────────┐
│                     DACE Loop                           │
├────────────────────────────────────────────────────────┤
│                                                        │
│   ┌─────────┐                                          │
│   │  Define │  Define objectives and success metrics    │
│   └────┬────┘                                          │
│        │                                                │
│        ▼                                                │
│   ┌─────────┐                                          │
│   │ Analyze │  Insight generation: Data→Story→Decision  │
│   └────┬────┘                                          │
│        │                                                │
│        ▼                                                │
│   ┌─────────┐                                          │
│   │Conclude │  Draw conclusions and decision recs  ◀──┐ │
│   └────┬────┘                               │         │
│        │                                    │         │
│        ▼                                    │         │
│   ┌─────────┐                               │         │
│   │ Execute │  Execute strategy and track ──┘         │
│   └─────────┘       │                             │    │
│        │            │                             │    │
│        ▼            │                             │    │
│   Return to Analyze ◀─────┘                       │    │
│                                                        │
└────────────────────────────────────────────────────────┘

Step 1: Define 🤖 [Core]

Automatically establish OKR tracking system:

define:
  status: "automated"
  trigger: "OKR update or quarter start"

  output:
    current_cycle: "2024_Q1"
    cycle_id: "dace_2024_Q1"

    objectives:
      - id: "obj_1"
        text: "Increase user engagement"
        owner: "product_team"

        key_results:
          - id: "kr_1_1"
            text: "DAU reaches 12 million"
            metric: "dau"
            baseline: 10500000
            target: 12000000
            current: 10800000
            progress: 30

          - id: "kr_1_2"
            text: "D7 retention rate reaches 30%"
            metric: "d7_retention"
            baseline: 0.25
            target: 0.30
            current: 0.285
            progress: 70

      - id: "obj_2"
        text: "Increase monetization revenue"
        owner: "biz_team"

        key_results:
          - id: "kr_2_1"
            text: "Monthly revenue reaches 50 million"
            metric: "monthly_revenue"
            baseline: 42000000
            target: 50000000
            current: 45500000
            progress: 43.75

    success_metrics:
      primary: ["dau", "d7_retention", "monthly_revenue"]
      supporting: ["dau_conversion", "arpu", "paying_users"]
      guardrail: ["user_satisfaction", "app_crash_rate"]

Step 2: Analyze (Insight Generation) 🤖 [Core]

Narrative insight transformation, decision recommendations, decision boundaries, confidence assessment

Integrates the insight transformation capabilities from the original decision-insight, converting analysis results into narrative insights.

2.1 Data Collection and Analysis [Core]

Automatically collect and analyze data:

analyze:
  status: "automated"
  execution_mode: "continuous"

  data_sources:
    - type: "metrics"
      name: "daily_metrics"
      frequency: "hourly"

    - type: "experiments"
      name: "ab_test_results"
      frequency: "on_completion"

    - type: "events"
      name: "product_events"
      frequency: "realtime"

  analyses_performed:
    - type: "anomaly_detection"
      findings:
        - metric: "dau_conversion"
          status: "warning"
          change: -3.2%
          reason: "Registration flow change impact"

    - type: "experiment_summary"
      findings:
        - experiment: "Simplified registration experiment"
          result: "positive"
          lift: +8.2%

    - type: "funnel_analysis"
      findings:
        - funnel: "Purchase conversion"
          conversion: 7.2%
          critical_drop: "step_1_to_2"

2.2 From Numbers to Stories [Core]

Data Analysis → Business Narrative

Transformation Principles:

| Data Language | Business Language | |---------|---------| | Conversion rate dropped 3.2% | "Out of every 100 visitors, 3 fewer complete registration" | | p-value=0.001 | "This conclusion has 99.9% confidence" | | Confidence interval [2%,5%] | "We are confident the improvement is between 2% and 5%" | | D7 retention 28.5% | "After one week, about 3 in 10 users are still using the product" |

Narrative Template:

narrative_template: |
  ## Insight Title

  ### Background
  How did [product/feature] perform during [time range]?

  ### Findings
  We found [core data change], which means [business impact].

  ### Impact
  Without intervention, [impact degree] is expected after [time].
  If intervention succeeds, [benefit] is expected.

  ### Recommendation
  Based on the data, we recommend [specific action].

2.3 Decision Recommendation Generation [Conditional]

Generate multiple actionable decision options:

action_options:
  - option_id: "opt_001"
    option_name: "Full rollout of new feature"
    description: "Roll out the new registration flow from the experiment group to all users"

    expected_effect:
      primary_metric: "+8.2% registration conversion rate"
      secondary_metrics:
        - "Registered users +12%"
        - "Overall DAU +2%"

    risk:
      level: "low"
      factors:
        - "All guardrail metrics are safe"
        - "Effect is stable with no novelty effect"
        - "Can be quickly rolled back"

    confidence:
      level: "high"
      basis:
        - "Statistically significant (p=0.001)"
        - "Complete experiment period (14 days)"
        - "Sufficient sample size (24830)"

    resource_requirements:
      engineering: "2 person-days (release deployment)"
      qa: "1 person-day (regression testing)"

    timeline:
      ready_for_release: "In 2 days"

    prerequisites:
      - "Technical review approved"
      - "Monitoring alerts configured"

  - option_id: "opt_002"
    option_name: "Phased rollout"
    description: "Release iOS first, then Android after stabilization"

    expected_effect:
      primary_metric: "+5.2% iOS registration conversion rate"
      secondary_metrics:
        - "Android effect pending verification"

    risk:
      level: "medium"
      factors:
        - "Android effect uncertain"
        - "Maintaining two logic branches"

    confidence:
      level: "medium"
      basis:
        - "iOS statistically significant"
        - "Android effect not significant"

2.4 Decision Boundary Annotation [Deep]

Distinguish between different types of decisions:

decision_boundary:
  type: "data_decision"
  criteria:
    - "Statistically significant (p  20%"
      severity: "high"
      action: "Trigger Conclude"

    - condition: "KR cannot be completed"
      severity: "critical"
      action: "Escalate + OKR adjustment"

Insight Type Processing

Anomaly Insight

anomaly_insight:
  type: "anomaly"

  narrative: |
    ## Anomaly Detection Insight

    Today's finding: Registration conversion rate dropped from 35% to 32%.
    Anomaly start time: Today at 9:00.
    Affected users: Approximately 15,000.

    Most likely cause: Registration flow change in v2.5.0.
    Confidence: 85%.

    Recommendation: Immediately check the new version implementation, prepare rollback plan.

  action_options:
    - "Immediately roll back to previous version"
    - "Emergency fix and hotfix release"
    - "Continue monitoring for 24 hours"

  decision_boundary:
    type: "data_decision"
    auto_execute_eligible: true
    condition: "Conversion rate continues to drop more than 5%"

Funnel Insight

funnel_insight:
  type: "funnel_analysis"

  narrative: |
    ## Purchase Conversion Funnel Insight

    Overall funnel conversion rate is 7.2%, down 0.5 percentage points from last week.

    Largest drop-off point: From browsing to add-to-cart, 84% of users lost.
    Drop-off concentrated in: Price-sensitive users, Android users.

    Recommended optimization directions: Price display strategy, add-to-cart guidance messaging.

  action_options:
    - "Optimize price display (show discounts, comparisons)"
    - "Enhance add-to-cart guidance (overlay, prompts)"
    - "Survey dropped-off users"

Output

Storage Path: output/pm-metrics-ops/decision-dace/

Output Depth Levels

| Depth Level | Output Scope | Description | |----------|----------|------| | quick | Decision recommendations + key evidence | Core conclusions + minimum viable deliverables, only output Define conclusions and Conclude recommended options | | standard | Complete decision analysis (current default) | Complete deliverables, including all four DACE phases output | | deep | Complete analysis + extended analysis | Complete deliverables + decision tree + sensitivity analysis + counterfactual reasoning + decision records + risk assessment |

Output Files: dacestatus.json, okrtracking.json, actionlog.json, dacecyclereport.md, decisioninsight.json, insight_library.json

Output Schema:

{
  "type": "object",
  "required": ["dace_status", "okr_tracking", "insight_id", "source", "narrative", "action_options"],
  "properties": {
    "dace_status": {"type": "object", "description": "DACE loop status, including current phase and progress"},
    "okr_tracking": {"type": "object", "description": "OKR tracking data, including objectives, key results and achievement rates"},
    "action_log": {"type": "array", "description": "Action log, including executed decisions and pending items"},
    "cycle_report": {"type": "object", "description": "Cycle report, including analysis conclusions and execution recommendations"},
    "insight_id": {"type": "string", "description": "Unique insight identifier"},
    "created_at": {"type": "string", "description": "Creation time"},
    "source": {"type": "object", "description": "Insight source, including type and confidence"},
    "narrative": {"type": "string", "description": "Narrative description, including background, findings, impact and recommendations"},
    "action_options": {"type": "array", "description": "Decision options list, including expected effects, risks and confidence"},
    "decision_boundary": {"type": "object", "description": "Decision boundary, including type and auto-execute eligibility"},
    "decision_maker": {"type": "string", "description": "Decision maker role"},
    "deadline": {"type": "string", "description": "Decision deadline"}
  }
}

Insight Output Example

data_insight:
  insight_id: "insight_20240115_001"
  created_at: "2024-01-15T14:30:00Z"

  source:
    type: "experiment_result"
    experiment_id: "exp_20240115_simplified_register"
    confidence: "high"

  narrative: |
    ## Simplified Registration Flow Experiment Insight

    ### Background
    The product team launched a simplified registration flow experiment on January 15, 2024,
    shortening the 5-step registration process to 3 steps.
    The experiment ran for 14 days with 24,830 users participating.

    ### Findings
    The experiment group (simplified flow) achieved a registration conversion rate of 38.1%,
    compared to 35.2% for the control group (standard flow), an improvement of 8.2 percentage points.
    This conclusion has 99.9% confidence (p=0.001).

    More importantly, this improvement is stable —
    from day 1 to day 14 of the experiment, the effect did not diminish,
    indicating this is not a novelty effect but a genuine experience improvement.

    ### Impact
    If we fully roll out this feature:
    - Expected new registered users per month **+12%** (approximately 36,000 users/month)
    - Based on the current conversion funnel, expected **+8%** DAU growth

    ### Risks
    We checked all guardrail metrics:
    - User 7-day retention: 42.0% → 41.8% (down 0.2%, acceptable)
    - DAU: Stable
    - Crash rate: No change

    All guardrail metrics are within safe range.

    ### Recommendation
    **Recommend full rollout of the simplified registration flow.**
    This is a low-risk, high-reward change, and the data supports immediate execution.

  action_options:
    - option: "Full rollout of simplified registration flow"
      option_id: "opt_001"
      expected_effect:
        primary: "Registration conversion rate +8.2%"
        secondary: ["DAU +2%", "New users +12%"]
      risk: "low"
      confidence: "high"

    - option: "Platform-by-platform rollout (iOS first)"
      option_id: "opt_002"
      expected_effect:
        primary: "iOS conversion +5.2%"
        secondary: ["Android pending verification"]
      risk: "medium"
      confidence: "medium"

    - option: "Continue experiment for 2 more weeks"
      option_id: "opt_003"
      expected_effect:
        primary: "More data for verification"
        secondary: ["Reduce uncertainty"]
      risk: "low"
      confidence: "low"

  decision_boundary:
    type: "data_decision"
    description: |
      The data clearly supports the "full rollout" option:
      - Statistically significant (p=0.001)
      - Practically significant (+8.2%)
      - All guardrail metrics safe
      - No novelty effect

    auto_execute_eligible: true

    automation_conditions:
      - condition: "Engineering team confirms ready for release"
        required: true
      - condition: "Monitoring alerts configured"
        required: true
      - condition: "Rollback plan prepared"
        required: true

    override_conditions:
      - condition: "Business strategy change"
        action: "Pause auto-execution, wait for human confirmation"

  recommended_action:
    action: "Full rollout of simplified registration flow"
    priority: "high"
    reason: "Strong data support, low risk, s

…

## Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

- **Author:** [LuckyOneTwoThree](https://github.com/LuckyOneTwoThree)
- **Source:** [LuckyOneTwoThree/vibe-skill](https://github.com/LuckyOneTwoThree/vibe-skill)
- **License:** MIT
- **Homepage:** https://luckyonetwothree.github.io/all-skill-html/

Install and usage instructions live in the source repository linked above.