Skip to content

rajann44/pm-skills / product-management

product-management

Verified

Act as an expert product manager whenever writing product requirements documents (PRDs), product one-pagers, specs, defining product strategy, prioritizing backlogs or feature requests, planning roadmaps, conducting customer interviews or user discovery, defining metrics, OKRs, and North Stars, analyzing product growth or retention, running sprint planning, or writing stakeholder updates. Use this skill even if the user does not explicitly say "product management."

references/metrics-growth.md

View on GitHub
# Product Metrics & Growth Reference

This document provides detailed guidelines for selecting metrics, designing growth loops, analyzing retention cohorts, and establishing leading proxy metrics.

---

## TABLE OF CONTENTS
- [1. Diagnostic Step-0: Instrumentation Check](#1-diagnostic-step-0-instrumentation-check)
- [2. Diagnostic Step-1: Shape Analysis](#2-diagnostic-step-1-shape-analysis)
- [3. Funnel Separation: Pirate Metrics (AARRR)](#3-funnel-separation-pirate-metrics-aarrr)
- [4. North Star Metric Selection](#4-north-star-metric-selection)
- [5. Growth Loops vs. Funnels](#5-growth-loops-vs-funnels)
- [6. Retention Curves & Cohort Analysis](#6-retention-curves--cohort-analysis)
- [7. Proxy Metrics](#7-proxy-metrics)

---

## 1. Diagnostic Step-0: Instrumentation Check & Re-baselining

Before formulating hypotheses or taking action on any product metric change (e.g., "activation dropped 15%"), you **must** perform a sanity check on your telemetry tracking and re-baseline if a bug is identified.
*   **Rule 1:** Never trust a metric change without checking the data pipeline.
*   **Rule 2 (Re-baselining):** If an instrumentation bug is found, you **must** calculate the measurement error:
    1.  *Quantify the Error:* Identify what percentage of the drop is due to the logging failure (e.g., if iOS users represent 66% of the traffic, and iOS logging was completely broken, the logging bug explains a `15% * 0.66 = 10%` reported drop).
    2.  *Calculate the Residual:* Subtract the quantified tracking error from the reported change to calculate the **residual real change**: `Reported Drop (15%) - Measurement Error (10%) = Residual Real Drop (5%)`.
    3.  *Pivot Decision:* If the residual is zero, halt the investigation and resolve the tracking bug. If the residual is non-zero, proceed to Step-1 using *only* the residual change (5%) for shape analysis and causal hypotheses.
*   **Rule 3 (Decomposition & Uncertainty):** Metric decompositions must explicitly state their quantification method (e.g., comparing affected cohort counts against pre-release baselines or stable control groups) and acknowledge uncertainty (error bands, confidence intervals, or behavioral interaction effects). If a diagnostic attribution changes between analyses (e.g., a technical outage is ruled out in favor of an acquisition mix shift), the PM must document the evidence for that transition rather than silently overwriting the history.
*   **Checklist:**
    1.  *Schema Changes:* Has the event definition or trigger point changed in the code repository?
    2.  *Logging Outage:* Are events failing to log on specific platforms (e.g., did an iOS app update break Segment/Amplitude SDK initialization)?
    3.  *Telemetry Latency:* Is there a delay in the database replication or processing pipeline?
    4.  *Double Logging:* Are events being sent twice due to a frontend loop?

---

## 2. Diagnostic Step-1: Shape Analysis

Once instrumentation is verified, plot the metric over time to analyze the **shape** of the change. The shape directs you toward distinct cause families:
*   **Step-Change (Sudden Drop/Spike):** The metric drops precipitously overnight.
    *   *Cause Families:* Technical deployment outage, third-party API failure, sudden marketing campaign launch/stop, seasonal holiday (e.g., Christmas Day), or tracking bug.
*   **Gradual Trend (Slow Decay/Growth):** The metric degrades slowly over weeks or months.
    *   *Cause Families:* Shifting user cohorts, change in acquisition traffic mix (e.g., higher volume of lower-intent paid ads traffic), slow competitive erosion, or onboarding drift.

---

## 3. Funnel Separation: Pirate Metrics (AARRR)

*   **Definition:** A linear product funnel framework that maps the user journey.
*   **v2 Hardening Rule (Funnel Separation):** Keep Activation, Retention, and Referral behaviors in separate, non-overlapping funnel stages. Mixing them (e.g., treating "invited a colleague" as both activation and referral) makes diagnostic analysis impossible.
    1.  **Acquisition:** Visitor lands and signs up.
    2.  **Activation:** The user experiences the core value ("Aha! moment") for the first time within a fixed window (e.g., created 1st document in 24 hours).
    3.  **Retention:** The user continues to actively use the product over subsequent cycles (e.g., logs in 3+ days in week 2).
    4.  **Referral:** The user invites others, expanding the user base (e.g., sent team invitation).
    5.  **Revenue:** The user converts to a paid subscription.
*   **Worked Example:** A SaaS tool sets distinct telemetry hooks: `event: signup` (Acquisition) -> `event: first_project_created` (Activation) -> `event: session_start_week_2` (Retention) -> `event: invite_sent` (Referral).
*   **Source URL:** [mindtheproduct.com](https://www.mindtheproduct.com)

---

## 4. North Star Metric Selection

*   **Definition:** The single key metric that best captures the core value your product delivers to its customers, serving as the primary driver for long-term sustainable business growth.
*   **When to use:** Use this when defining a product's high-level strategy to align all product teams.
*   **When NOT to use:** Do not use revenue, stock price, or profit as a North Star metric (these are lagging financial indicators).
*   **Worked Example:** Airbnb uses **Nights Booked** rather than revenue. This captures value for guests and hosts, and directly drives Airbnb's business growth.
*   **Source URL:** [lennysnewsletter.com/p/choosing-your-north-star-metric](https://www.lennysnewsletter.com)

---

## 5. Growth Loops vs. Funnels

*   **Definition:** A growth loop is a closed-loop system where the output of one step is directly reinvested as an input to the next step, creating a self-reinforcing flywheel.
*   **When to use:** Use this when designing a product's sustainable, long-term acquisition and retention engine.
*   **Worked Example:** Pinterest's content loop. *Input:* User signs up and searches. *Action:* User saves a pin to their board. *Output:* Search engines index this board. *Reinvestment:* New users discover Pinterest via search, leading to new sign-ups.
*   **Source URL:** [reforge.com/blog/growth-loops](https://reforge.com/blog/)

---

## 6. Retention Curves & Cohort Analysis

*   **Definition:** A cohort analysis groups users by their signup date and tracks active usage over time. The resulting retention curve plots the percentage of users returning at Day/Week/Month X. A flattening curve parallel to the x-axis indicates a stable user base, serving as the primary metric signal of Product-Market Fit (PMF).
*   **When to use:** Use this to measure product-market fit, analyze onboarding health, and diagnose long-term user retention.
*   **Worked Example:** A B2B note-taking app measures weekly retention. If the retention curve drops from 100% to 20% by Week 4 but then remains flat at 20% through Week 12, it proves PMF for a specific 20% segment of users.
*   **Source URL:** [reforge.com/blog](https://reforge.com/blog/)

---

## 7. Proxy Metrics

*   **Definition:** A measurable, short-term leading indicator that product teams can directly influence, which is mathematically or behaviorally correlated to a lagging, long-term business outcome (like churn or customer lifetime value).
*   **When to use:** Use this to give product teams actionable, short-term targets for their cycles.
*   **Worked Example:** Netflix personalization proxy metric. *Lagging Metric:* Monthly subscriber retention. *Hypothesis:* Users who configure their preference queue early are less likely to churn. *Proxy Metric:* The percentage of members who add at least 3 titles to their Queue within their first 30 days.
*   **Source URL:** [gibsonbiddle.com/proxy-metrics](https://www.gibsonbiddle.com)
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 323kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 323kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 323kUpdated Claude CodeCodexVerifiedInstall