Free Email Course
AI-Powered Analytical Thinking
30 emails over 15 weeks teaching the thinking and the tools product data scientists actually use. Built for people who make decisions with data and want to stop waiting on someone else to run the numbers.
Learn to build with Claude Code · Hex · Querio
You don't need to become a data scientist
The thinking, not the title. 15 weeks.
Watch me solve this in 10 minutes
Support tickets spiked 40%. Here's the whole analysis.
The four moves behind every good analysis
Frame, explore, root cause, deliverable.
The file that turns Claude Code into a data scientist
CLAUDE.md, skills, agents, slash commands.
Four rungs between a vague ask and a useful answer
Goal, Decision, Metric, Hypothesis.
Why only 1 in 3 experiments wins (that's good news)
Kohavi, Bing's $100M headline, and what win rate tells you.
How correlation tricked your team into a $500K mistake
The hidden confounder behind "power users retain better."
Why 'look into retention' never produces an answer
Topics produce exploration. Questions produce answers.
Five rows that save three days of wasted work
The Analysis Design Spec, before you touch data.
Impact vs Curiosity. The 2x2 that kills your backlog.
Interesting questions and important questions are not the same thing.
Why your company has 3 different 'active user' numbers
Marketing says 25K. Product says 18K. Finance says 8K.
Seven fields that kill metric ambiguity forever
The Metric Anatomy Template.
Why revenue dropped. Here's the 5-minute diagnosis.
Driver trees and the 'one branch' principle.
The metric that tells you your 'win' broke something else
Guardrails before you ship, not after.
Never look at the aggregate first
Segment-first thinking and why aggregates lie.
Five steps to find why any metric dropped
Segment, narrow, compare, hypothesize, test.
Where in your funnel the problem actually lives
One transition owns the drop. Find it in 10 minutes.
The 5-minute check that prevents embarrassing mistakes
Nulls, duplicates, date ranges, join logic.
The sentence structure that saves your next experiment
If/Then/Because plus Amount and Mechanism.
The five ways experiments lie to you
Peeking, underpowered, SRM, multiple comparisons, novelty.
The one diagram that reveals hidden confounders
Causal DAGs and how to read them.
Your sample is too small. Now what?
The observational causal toolbox for when you can't randomize.
The chart title that stops your VP from asking 'so what?'
Action titles vs description titles.
Three slides. Decisions made.
The 3-Slide Executive Readout format.
Stop reporting what the data says
Having an opinion, backed by data.
The same analysis, five different formats
Stakeholder Communication Matrix.
Stop answering questions. Start finding them.
Proactive analytics and the unasked question.
The math that says which bet is worth making
Opportunity sizing in ten minutes.
Impact times effort. Pick three, kill the rest.
Prioritization plus the close-the-loop habit.
The shape of the data job in two years
The convergence and where this goes.
From
Shane, Sravya, and Hai.
Shane Butler
Principal DS at Ontra
Previously at Stripe · Nextdoor · AppFolio
Sravya Madipalli
Sr Manager DS at Superhuman
Previously at Microsoft · eBay · Nextdoor
Hai Guan
Head of Data at Ontra
Previously at LinkedIn · Nextdoor · Pinterest
Start thinking like a product data scientist
Free. No spam. Day 1 lands in your inbox right after you confirm your email.