Metrics Framework
How to define, structure, and analyze metrics for any product feature or situation.
When to Use This Framework
Use this framework when asked to: define success metrics, diagnose a metric drop, set up a measurement plan, or evaluate product health.
The Goal > Signal > Metric Hierarchy
Always start with the goal, not the metric. Metrics exist to measure progress toward goals.
Goal: What outcome are you trying to achieve? (qualitative) Signal: What observable behaviors indicate progress? (behavioral) Metric: How do you quantify that signal? (numeric)
Example:
The AARRR Funnel (Pirate Metrics)
For consumer products, the AARRR funnel maps the full user lifecycle. Think through each stage when defining a metrics plan.
Acquisition
How do users find you? Metrics: new user signups, cost per acquisition (CPA), traffic by channel, conversion rate from landing page.
Activation
Do new users have a great first experience? Metrics: time-to-first-value, onboarding completion rate, day-1 retention, feature adoption within the first session.
Retention
Do users come back? Metrics: D7/D30 retention, monthly active users (MAU), churn rate, subscription renewal rate.
Revenue
Does usage translate to money? Metrics: average revenue per user (ARPU), lifetime value (LTV), conversion from free to paid, monthly recurring revenue (MRR).
Referral
Do users bring others? Metrics: viral coefficient (K-factor), referral rate, Net Promoter Score (NPS), organic vs. paid new user split.
Use AARRR to diagnose where a product is healthy or broken. If acquisition is strong but activation is weak, fix onboarding before spending more on marketing.
The HEART Framework
Google's HEART framework provides five dimensions for measuring user experience quality.
H — Happiness
How satisfied are users? Measured through surveys (NPS, CSAT, App Store ratings).
E — Engagement
How frequently and deeply are users interacting? Examples: DAU/MAU ratio, session length, actions per session.
A — Adoption
Are new users finding value and activating? Examples: feature adoption rate, time-to-first-value.
R — Retention
Are users coming back? Examples: D1/D7/D30 retention, churn rate, renewal rate.
T — Task Success
Can users complete their goals? Examples: task completion rate, error rate, time-on-task.
Leading vs. Lagging Indicators
This distinction is critical for building a useful metrics plan.
Lagging indicators measure outcomes that have already happened. They tell you what happened, but not why or what to do next.
Examples: monthly revenue, quarterly churn, annual DAU growth.
Leading indicators predict future outcomes. They are actionable — you can move them before the lagging metric changes.
Examples: feature adoption rate (predicts retention), support ticket volume (predicts churn), onboarding completion (predicts activation).
Best practice: Always pair a lagging goal metric with leading indicators you can act on. "We want to grow 30-day retention (lagging) by improving onboarding completion (leading)."
Choosing the Right Metrics
North Star Metric
One primary metric that best captures the core value your product delivers. Examples:
Guardrail Metrics
Metrics that must not degrade while you optimize for your North Star. Examples: p99 latency, crash rate, customer support ticket volume.
Counter-Metrics
Metrics that prevent gaming. If your goal is engagement, add a counter-metric for spam or low-quality actions.
Diagnosing a Metric Drop
Use this structured approach when a metric suddenly drops:
- Is it real? Check for data pipeline issues, logging bugs, or timezone shifts.
- When did it start? Find the exact inflection point. Correlate with deployments or external events.
- Who is affected? Segment by platform, region, user cohort, and feature.
- What changed? Review recent releases, experiments, and external events.
- What is the hypothesis? Form a specific, falsifiable hypothesis.
- How do you validate? Define what data would confirm or disprove the hypothesis.
Common Mistakes to Avoid
- Picking vanity metrics (total signups) over actionable ones (activated users)
- Optimizing one metric without watching guardrails
- Confusing correlation with causation
- Setting metrics before defining the goal
- Ignoring qualitative signals alongside quantitative data
- Tracking only lagging indicators with no leading metrics to act on
- Skipping the AARRR funnel analysis and jumping straight to one metric