Nw Outcome Kpi FrameworkSAFE

AI agents that guide you from idea to working code, with you in control at every step.

The instruction file, verbatim from the audited commit — this is the text the model reads, and the surface the audit's instruction layer examines. Quoted here so you can judge it without cloning anything.

---
name: nw-outcome-kpi-framework
description: Outcome KPI definition methodology - synthesizes Who Does What By How Much (Gothelf/Seiden), Running Lean (Maurya), and Measure What Matters (Doerr) into a practical framework for measurable outcome KPIs
user-invocable: false
disable-model-invocation: true
---

# Outcome KPI Framework

"Doing stuff isn't the point. Achieving stuff is." -- Jeff Gothelf

Defines measurable outcome KPIs for user stories and features. Loaded during Phase 4 (Requirements Crafting) to produce `outcome-kpis.md`. Synthesizes three frameworks: customer-centric OKRs, lean metrics, and OKR methodology.

## The Outcome KPI Formula

Primary template from Gothelf/Seiden. Every KPI answers five questions:

| Component | Question | Example |
|-----------|----------|---------|
| **Who** | Which user segment? | Returning customers with 2+ orders |
| **Does What** | What observable behavior changes? | Complete checkout without contacting support |
| **By How Much** | What is the measurable target? | 40% reduction in support tickets |
| **Measured By** | How do we collect the data? | Support ticket system + checkout analytics |
| **Timeframe** | When do we measure? | 30 days post-release, then weekly |

Formula: **[Who] [Does what] [By how much]**

Apply as litmus test: if a KPI cannot answer all five components, it measures an output (feature delivery), not an outcome (behavior change).

### Good vs Bad KPIs

| Bad (Output) | Good (Outcome) |
|-------------|----------------|
| Launch mobile app v2 | Mobile users complete purchases 40% more often |
| Build recommendation engine | Users purchase from recommendations, increasing from 10% to 25% |
| Deploy onboarding redesign | New users complete onboarding within 24 hours 30% more often |
| Ship CSV export | Analysts resolve data questions without engineering support 60% of the time |

## Leading vs Lagging Indicators

From Gothelf/Seiden: business results are lagging -- teams cannot directly influence them. Target leading indicators instead.

| Type | Definition | Examples | Actionable? |
|------|-----------|----------|-------------|
| **Lagging** (Impact) | Business results already happened | Revenue, NPS, market share, churn rate | No -- too slow, too many variables |
| **Leading** (Outcome) | Behavior changes predicting business results | Purchase completion rate, feature adoption, retention | Yes -- teams can run experiments |
| **Leading** (Secondary) | Behaviors predicting primary leading indicators | Page visits, trial starts, onboarding steps completed | Yes -- most granular, fastest signal |

### Outcome Mapping Chain

Map every KPI through this chain to ensure traceability:

```
Business KPI (Lagging/Impact)
    Example: "Increase quarterly revenue by 15%"
        |
        v
    Customer Behavior (Leading/Outcome)
        +-- Users complete purchases from recommendations (+25%)
        +-- Users return within 7 days (+20%)
        |
        v
    Secondary Behavior (Leading/Secondary)
        +-- Users browse recommendation pages (+30%)
        +-- Users enable push notifications (+15%)
```

Each layer decomposes into more granular behavioral metrics. Teams target the highest-leverage behavior.

## Actionable vs Vanity Metrics

From Maurya (Running Lean): actionable metrics "tie specific and repeatable actions to observed results."

| Dimension | Vanity | Actionable |
|-----------|--------|------------|
| Measures | Business size (totals) | Individual behavior (rates) |
| Data type | Gross aggregates | Ratios and unit economics |
| Cause/effect | No insight into why | Directly signal product-market fit |
| Examples | Total users, page views, downloads | Activation rate, retention cohort, churn rate |
| Decision value | Cannot inform action | Drives specific experiments |

### The OMTM (One Metric That Matters)

Pick ONE metric per product stage. Optimizing one metric reveals the next.

| Stage | Focus | Example OMTM |
|-------|-------|--------------|
| Empathy | Problem validation | Interview pain intensity (qualitative) |
| Stickiness | Retention | Churn rate, DAU/MAU ratio |
| Virality | Organic growth | Viral coefficient, referral rate |
| Revenue | Monetization | Customer Lifetime Value, MRR |
| Scale | Growth efficiency | CAC/LTV ratio, payback period |

**Good metric characteristics**: rate or ratio (not absolute number) | comparable across time | simple enough to remember | predictive | behavior-changing.

### Customer Factory (AARRR) Constraint Mapping

From Maurya: model the business as a production line. Identify the bottleneck, then focus KPIs there.

| Stage | Key Question | Example Metric |
|-------|-------------|----------------|
| **Acquisition** | Are we reaching the right people? | Visitor-to-signup conversion rate |
| **Activation** | Do users get the "aha moment"? | % completing core action in first session |
| **Retention** | Do users come back? | Week-1 return rate, DAU/MAU |
| **Revenue** | Do users pay? | Trial-to-paid conversion rate |
| **Referral** | Do users tell others? | Referral rate, viral coefficient |

Activation is causal -- it drives retention, revenue, and referral. Prioritize activation KPIs when uncertain.

## OKR Integration

From Doerr (Measure What Matters): connect KPIs to strategic objectives.

### Writing Key Results

Every Key Result uses the outcome formula. Quality criteria:

1. **Measurable**: "It's not a Key Result unless it has a number" (Marissa Mayer)
2. **Outcome-focused**: "Increase email subscribers by 20%" not "Launch newsletter"
3. **Time-bound**: deadline (typically end of quarter)
4. **Verifiable**: no ambiguity about whether met
5. **Aggressive yet realistic**: stretch without demoralizing

### Committed vs Aspirational

| Type | Expected Score | Resource Allocation | Failure Response |
|------|---------------|--------------------|-----------------|
| **Committed** | 1.0 (must deliver) | Consume most available resources | Requires explanation, replanning |
|

SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.

What the source does

Filesystem

none-observed

Network

none-observed

Shell

none-observed

Dependencies

pinned

Secrets in source

none-found

Findings (0)

No findings outside the package's declared scope.

Gates applied: no_behavioural_pass.