Skip to content

borghei/Claude-Skills / project-management/execution/activation-funnel

activation-funnel

Verified

Designs and analyzes activation funnels: stage conversion, drop-off, bottlenecks and the aha event that predicts retention. Use when a funnel leaks, signups grow but activation is flat, or you must pick an activation event.

references/funnel-analysis-playbook.md

View on GitHub
# Funnel Analysis Playbook: AARRR, Aha Events, Math, Workflow

Read this when you need the full operating detail of funnel work — the AARRR/AAARRR stage tables, how to define the activation "aha" event, the conversion/drop-off math, counter-metric pairing, leading-vs-lagging indicators, the step-by-step workflow, the `funnel_analyzer.py` reference, troubleshooting, and success criteria.

## Pirate Metrics framework (AARRR)

Dave McClure proposed Pirate Metrics in 2007 as a five-stage growth funnel:

| Stage | What it measures | Typical events |
|---|---|---|
| **Acquisition** | Users find you | Landing-page visit, ad click, organic search visit |
| **Activation** | First good experience | Signup, profile complete, "aha" event |
| **Retention** | They come back | D1, D7, D30 active; weekly active |
| **Revenue** | They pay | Trial-to-paid, upgrade, expansion |
| **Referral** | They tell others | Invite sent, invite accepted, share rate |

McClure's insight was that most teams obsess over Acquisition (the top) and Revenue (the bottom), while the leverage is usually in Activation and Retention. A 5-point improvement in activation flows through to retention, revenue, and referral; a 5-point improvement in acquisition often gets eaten by churn.

### AAARRR (Awareness-prefixed)

Some teams add **Awareness** at the front: how many people know your product exists at all? AAARRR is useful when the brand is new or the category is undefined; AARRR is fine when product-market fit is established.

| Stage | Question |
|---|---|
| **Awareness** | Do they know you exist? (impressions, mentions, branded search) |
| **Acquisition** | Do they visit you? (visits, signups-started) |
| **Activation** | Do they get value? (the aha event) |
| **Retention** | Do they come back? (D1, D7, D30) |
| **Revenue** | Do they pay? (trial-to-paid) |
| **Referral** | Do they bring friends? (invites, shares) |

## Defining activation: the "aha" event

Sean Ellis introduced the **activation-rate framework**: the activation event is the action that, statistically, separates users who retain from users who churn. Find it by:

1. Identify the cohort of users who are still active at D30 (or your retention horizon).
2. Look at what they did in their first session / first week that the churned cohort did not do.
3. The behavior with the highest correlation to D30 retention is the activation event candidate.

### Real activation events (publicly discussed)

| Product | Activation event |
|---|---|
| Slack | A team sends 2,000 messages |
| Facebook | A user adds 7 friends in 10 days |
| Dropbox | A user puts 1 file in 1 folder on 1 device |
| Airbnb (host) | A host books their first guest |
| Airbnb (guest) | A guest books their first trip |
| Notion | A user creates 3+ pages and invites 1+ collaborator |
| Spotify | A user listens to 25+ minutes in their first week |
| Duolingo | A user completes 1 lesson in 3 of their first 7 days |
| Twitter | A user follows 30+ accounts |
| HubSpot | A user adds 5 contacts and 2 deals |
| Pinterest | A user follows 5 boards and saves 1 pin |
| LinkedIn | A user adds 5 connections in their first week |
| Figma | A user creates a file and adds 2+ collaborators |
| Canva | A user creates a design and downloads or shares it |

The pattern: a **specific count + a specific time window + a specific action**. Vague "completed onboarding" is not an activation event -- it is an output. A good activation event is observable, instrumented, and predictive of long-term value.

## Funnel math

### Conversion rate

For each stage transition: `conversion[i] = users[i+1] / users[i]`

### Drop-off (absolute)

`drop[i] = users[i] - users[i+1]`

### Drop-off (rate)

`drop_rate[i] = 1 - conversion[i]`

### Cumulative conversion

`cumulative[i] = users[i] / users[0]`

The largest absolute drop is where the funnel leaks the most users. The largest *relative* drop (worst conversion rate) is where the stage is most broken. Both signals matter; this skill's tool surfaces both.

## Counter-metric pairing

Funnel optimization is famously easy to game. If you measure only conversion, you can boost it by lowering the bar at each stage:

- Drop email verification -> signup conversion goes up; fake-account problem goes up too.
- Auto-import data -> "activation" rate climbs; users feel nothing happened.
- Add a "skip" button -> setup completion climbs; D7 retention drops.

For every funnel stage, declare a **counter-metric** that must not degrade. The pair (conversion, counter-metric) is what gets reviewed -- not conversion alone.

| Funnel stage | Primary metric | Counter-metric |
|---|---|---|
| Acquisition | Signups / visitors | Quality signup rate (e.g., real-email %) |
| Activation | Activation rate | D7 retention of activated users |
| Retention | D30 retention | Engagement depth (events / active user) |
| Revenue | Trial-to-paid | First-month churn |
| Referral | Invite acceptance | Quality of referred user (do they activate?) |

## Leading vs lagging activation indicators

| Indicator | Timing | Use |
|---|---|---|
| **Leading** | Hours / days into use | Daily dashboard, on-call signal (e.g., "step 2 completion in first session") |
| **Activation** | First N days | The headline activation rate |
| **Lagging** | D30 retention of activated users | Quarterly truth check (is the activation event still predictive?) |

Leading indicators move first, so they are what the team optimizes. The activation rate is the headline. The lagging indicator audits whether the activation definition is still correct.

## Workflow

1. **Pick the framework.** AARRR for established product, AAARRR if the category is undefined.
2. **Define each stage as an event.** Each transition is `event A -> event B`. No vague stages.
3. **Pin the activation event.** Use Ellis's framework: find the behavior that separates retained from churned users.
4. **Instrument.** Telemetry must capture each stage event with a tenant / user / session id. Without instrumentation, the funnel is a guess.
5. **Pull stage counts** over a fixed window (week, month, or cohort). Note: a funnel for a cohort is more honest than a snapshot across cohorts.
6. **Run `funnel_analyzer.py`** to get conversion, drop-off, bottleneck, and the Mermaid diagram.
7. **Pair every stage with a counter-metric.** Document and review.
8. **Identify the bottleneck.** The largest absolute drop is usually where to invest. Use `prioritization-frameworks/` to rank fix-the-funnel projects.
9. **Wire the funnel into reporting.** Funnel deltas appear in weekly status (`status-update-generator/`) and quarterly review.

## Tools

| Tool | Purpose | Command |
|---|---|---|
| `funnel_analyzer.py` | Compute funnel conversions + drop-off + bottleneck; render in 6 formats | `python scripts/funnel_analyzer.py --input funnel.json --format markdown` |
| `funnel_analyzer.py --demo` | Inspect a worked SaaS activation funnel | `python scripts/funnel_analyzer.py --demo --format mermaid` |

## Troubleshooting

| Symptom | Likely Cause | Resolution |
|---|---|---|
| Funnel conversion looks great but D30 retention is flat | Activation event is too easy / not predictive of retention | Re-run Ellis's framework on the latest cohort; pick a harder + more predictive event |
| Stage counts do not add up; later stages exceed earlier ones | Multi-session or multi-tenant double-counting; events are not properly scoped | Constrain analysis to a single-cohort, single-user funnel; deduplicate by user_id; check event timestamps |
| Team optimizes acquisition; nothing improves downstream | The bottleneck is downstream, not at the top | Run the tool, look at absolute drop sizes, redirect investment to the largest drop stage |
| Activation rate jumps after a UX change but engagement drops | Counter-metric not paired; the funnel is being gamed | Add a counter-metric per stage; review the pair, not the conversion alone |
| Funnel looks different every week | Snapshot funnel across mixed cohorts | Switch to cohort-based funnel: pick a signup cohort, follow it through stages over the same N days |
| Funnel only covers the happy path | Edge cases and re-entries not modeled | List the 3-5 most common alternative paths; instrument them as parallel funnels or recovery loops |
| Stakeholders ignore the funnel chart | Chart has 12 stages and reads as a wall of numbers | Compress to 4-6 stages; the deeper drill-down lives in a dashboard, not the headline funnel |

## Success Criteria

- The funnel has 4-7 stages (5 is the sweet spot). Each stage is a specific event.
- The activation event is specific (count + window + action) and statistically predictive of retention.
- Every stage has a paired counter-metric documented.
- The analysis is cohort-based, not snapshot-based.
- The largest absolute drop is identified and assigned an owner.
- The funnel is reviewed weekly during build, monthly during steady-state.
- A new hire can name the activation event and the current funnel's biggest leak within their first week.

## Tool Reference

### funnel_analyzer.py

Analyzes a funnel from a JSON definition. Computes per-stage conversion + absolute and relative drop-off + bottleneck. Renders in 6 formats. Mermaid output uses `flowchart LR` with conversion % on each arrow.

| Flag | Type | Default | Description |
|---|---|---|---|
| `--input` | string | (required unless `--demo`) | Path to JSON funnel definition |
| `--demo` | flag | false | Use a built-in SaaS activation funnel |
| `--format` | choice | markdown | Output: json, markdown, mermaid, confluence, notion, linear |
| `--output` | string | stdout | Output file path |
| `--name` | string | (from input) | Override funnel name |
| `--cohort` | string | (from input) | Override cohort label |

### Input JSON shape

```json
{
  "name": "SaaS Free Trial Activation",
  "cohort": "Cohort: signups in week of 2026-05-04",
  "framework": "AARRR",
  "stages": [
    {"name": "Landing visit", "stage": "acquisition", "count": 10000, "counter_metric": "Bounce rate < 60%"},
    {"name": "Signup started", "stage": "acquisition", "count": 2400, "counter_metric": "Signup-to-real-email > 90%"},
    {"name": "Email verified", "stage": "acquisition", "count": 1800, "counter_metric": null},
    {"name": "First workspace created", "stage": "activation", "count": 1100, "counter_metric": null},
    {"name": "Aha event: 3 docs created + 1 collaborator", "stage": "activation", "count": 410, "counter_metric": "D7 retention of activated > 55%"},
    {"name": "D30 retention", "stage": "retention", "count": 290, "counter_metric": "Engagement events/day > 4"},
    {"name": "Paid", "stage": "revenue", "count": 180, "counter_metric": "First-month churn < 8%"}
  ]
}
```

### Mermaid output sample

```mermaid
flowchart LR
    S0["Landing visit<br/>10000"] -->|24.0%| S1["Signup started<br/>2400"]
    S1 -->|75.0%| S2["Email verified<br/>1800"]
    S2 -->|61.1%| S3["First workspace created<br/>1100"]
    S3 -->|37.3%| S4["Aha event<br/>410"]
    S4 -->|70.7%| S5["D30 retention<br/>290"]
    S5 -->|62.1%| S6["Paid<br/>180"]
```

The tool flags the largest absolute drop with a callout in the rendered output. In the demo above, the biggest leak in absolute users is at Landing -> Signup (7600 users), but the biggest *rate* drop is at First workspace -> Aha (37.3% conversion). Both get surfaced.
brainstormingobra/superpowersYou MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).Stars: 322kUpdated Claude CodeCodexVerifiedInstall
brand-guidelinesanthropics/skillsApplies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
frontend-designanthropics/skillsGuidance for distinctive, intentional visual design when building new UI or reshaping an existing one. Helps with aesthetic direction, typography, and making choices that don't read as templated defaults.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
mcp-builderanthropics/skillsGuide for creating high-quality MCP (Model Context Protocol) servers that enable LLMs to interact with external services through well-designed tools. Use when building MCP servers to integrate external APIs or services, whether in Python (FastMCP) or Node/TypeScript (MCP SDK).Stars: 198kUpdated Claude CodeCodexVerifiedInstall