I.
Vanity Metrics
The numbers dashboards reach for when they have nothing real to say — pageviews, follower counts, “engagement” without a denominator. We catalog them, define them, and recommend the replacement metric.
- Pageviews vs. sessions vs. users
- Why “engagement rate” is not a metric
- The five numbers to delete from Monday reports
- The Vanity Metrics Awards (annual)
84 dispatches
II.
Attribution
Multi-touch attribution, media-mix modeling, incrementality testing — the methodology wars of marketing analytics, written for practitioners who actually have to defend the model on Monday.
- First-touch / last-touch / MTA field tests
- MMM done badly, done well, done by us
- Incrementality is not a test
- Paid social attribution, a 14-month teardown
76 dispatches
III.
Retention & Cohorts
Retention is a shape, not a number. Decay curves, N-day retention, cohort overlays, and the survival math behind the charts your CS team is afraid to show the board.
- Reading decay curves across SaaS verticals
- COVID cohort retention, six-month field notes
- Retention chart debugging post-mortems
- N-day, weekly, and rolling-window pitfalls
71 dispatches
IV.
Forecasting
Time-series forecasting on messy weekly revenue, Prophet vs. SARIMAX vs. the things vendors will not put in the pitch deck. Every forecast is back-tested against the original dataset.
- Prophet vs. SARIMAX, practitioner’s guide
- Forecasting with insufficient data
- Confidence intervals you can actually defend
- Forecast drift: when to retire a model
58 dispatches
V.
Dashboard Design
Operational dashboards, executive dashboards, the kind nobody reads — redesigned from scratch with the principle that a chart earns its place by changing a decision. Audit-driven, no decoration.
- The one-page operational dashboard
- Why “executive dashboards” fail
- Common patterns across 800+ audited dashboards
- Chart selection by decision, not by aesthetics
112 dispatches
VI.
Growth
PLG, referral loops, funnel math, and the parts of growth nobody puts in the investor update. Field studies with 4M-user consumer apps and Series-C SaaS, fully dataset-attached.
- Referral loops, properly measured
- PLG is a measurement problem first
- Funnel math nobody shows the board
- When acquisition stops working
86 dispatches