Latest Insights
Exploring the frontiers of AI, software architecture, and engineering.
The Average Lies: Mean vs. Median on Skewed Metrics
Average revenue per user is up 40% — because one whale landed. On skewed metrics the mean is a hostage to the tail; here is when to trust the median instead.
Read Article →Simpson's Paradox: When Every Segment Improves and the Total Gets Worse
Conversion improved in every single segment — and fell overall. No bug, no fraud: a mix shift. How Simpson’s paradox ambushes aggregated metrics and how to see through it.
Read Article →Regression to the Mean: Why Your Best Week Is Followed by a Worse One
You praised the top performers and they got worse. You coached the worst and they improved. Before concluding anything about praise or coaching, meet regression to the mean.
Read Article →Base Rates Beat Gut Feel: Bayesian Thinking for Business Decisions
Your fraud detector is 99% accurate — and 9 out of 10 flags are innocent. The base rate eats accuracy for breakfast. A practical Bayesian habit for everyday decisions.
Read Article →A/B Testing on Small Traffic: Sample Size, Peeking, and When Not to Test
You ran an A/B test for a week, B won by 12%, you shipped it — and revenue did not move. Small-traffic testing fails in predictable ways. Here is how to survive it.
Read Article →Cohort Analysis: Reading Retention the Way It Actually Behaves
Overall retention is flat — but every new cohort retains better than the last. Blended metrics hide product progress; cohorts reveal it. How to read them properly.
Read Article →