
Metrics & Diagnostics Decisions · Module 4 of 4
The Confounded Read
You're Maya, Senior PM at NorthPeak, a B2B SaaS platform. Five correlations are on the table and every one is real — integration users retain +30pp, onboarding step 3 doubles activation, support-contacting customers churn 3× more, the new AI feature spiked usage, annual-plan teams expand more — each with an obvious action already stapled to it, and a growth team eager to ship all five. Diagnose each claim's causal structure (a genuine cause, a lurking variable driving both, a reversed arrow, a decaying novelty spike, or a real uncertainty), then deliver a verdict: roll it out, drop it, or spend a controlled test. Discover why the biggest number on the board is the emptiest (committed teams both integrate and retain, so the effect won't transfer), why reducing support contact is the most harmful move available (churning customers contact support — the arrow runs the other way), why a launch spike is the least informative moment to judge a feature, and why the disciplined read (net 200) beats both the naive believer who ships every strong number (−60, the worst on the board) and the reflexive skeptic who ships none (0, and the whole 200 left on the table). The mdd capstone: a correlation is a question, not a cause.
0 of 22 items completed0%~78 min remaining
Material · 22 items