Week 05 · Days 21–25
Metrics & Analytics
Get comfortable with the metrics PMs live in: funnels, retention curves, and basic SQL.
Portfolio deliverable · A metrics dashboard mockup with defined KPIs
North star metrics & KPI trees
Lesson: North star metrics & KPI trees
A north star metric (NSM) is the single number that best captures the value your product delivers to users — and that correlates with long-term business success. Good examples: Spotify's 'time spent listening,' Airbnb's 'nights booked.' A bad NSM is a vanity metric (total signups) that doesn't reflect ongoing value. Once you have an NSM, build a KPI tree below it — 3-5 metrics that, if improved, would move the NSM. Each input metric should be ownable by a specific team (e.g. 'onboarding completion rate' is ownable by the activation team). This is how growth orgs assign accountability without everyone chasing the same vague goal.
Task: Build a KPI tree for your practice app
Propose a north star metric for your practice app and break it into 3-5 input metrics in a simple tree diagram.
Retention curves & cohorts
Lesson: Retention curves & cohort analysis
A retention curve plots % of users still active on Day 1, 7, 14, 30, etc. after signup. Most curves decline sharply then flatten — the 'smile curve' (or more accurately, an asymptote) is when the curve stabilizes above zero, meaning you've found a core group of users who stick around long-term. If the curve keeps declining toward zero, you have a leaky bucket no amount of acquisition will fix. A cohort is a group of users who started on the same date — comparing cohorts over time (Jan signups vs Feb signups) shows whether retention is improving as you ship changes. Averages hide this: a flat 'average DAU' can mask a shrinking core user base offset by new signups.
Task: Sketch a retention curve hypothesis
Sketch (by hand or in a spreadsheet) what you think your practice app's Day 1/7/30 retention curve looks like, and what would need to be true for it to look 'good.'
SQL basics for PMs
Lesson: SQL fundamentals (SELECT, WHERE, GROUP BY, JOIN)
PMs who can self-serve data move faster and ask sharper questions. Four commands cover most day-to-day needs: SELECT (which columns you want), WHERE (filter rows — e.g. WHERE country = 'US'), GROUP BY (aggregate — e.g. count of signups per day), and JOIN (combine data from two tables, e.g. users + their events). A realistic PM query: 'SELECT date, COUNT(user_id) FROM events WHERE event_name = \'completed_onboarding\' GROUP BY date' — this answers 'how many users completed onboarding each day?' without needing an analyst. Practice on a free sandbox; the syntax sticks fastest by writing it yourself.
Task: Write 3 practice SQL queries
Using a free SQL practice tool (e.g. SQLZoo, Mode SQL tutorial), write 3 queries: total count with a filter, group by with count, and a simple join.
Designing a metrics dashboard
Task: Design a dashboard mockup
In Figma or a spreadsheet, design a mockup of a metrics dashboard for your practice app - include north star, 3-4 input metrics, and retention chart placeholder.
Synthesize: Metrics framework doc
Deliverable: Metrics Framework + Dashboard
Combine your KPI tree, retention hypothesis, and dashboard mockup into a 'Metrics Framework' doc explaining how you'd measure success for your practice app.
Advanced Challenge: Design an AI feature scorecard
Add a section to your metrics framework specifically for AI features: define a scorecard with (1) quality metrics — e.g. task completion rate, response relevance (could be human-rated 1-5 on a sample), hallucination/error rate on a golden test set; (2) trust metrics — e.g. % of users who edit/reject AI suggestions, fallback-to-manual rate; (3) business metrics — does the AI feature move your north star (e.g. conversion, time-to-value)? Most teams only track #3 — having a framework for all three is a differentiator. Apply the scorecard to a real AI product as an example.