Become a Product Data Scientist
A step-by-step path from working analyst to product data scientist: SQL, statistics, experiments, product sense and interview prep. Plan on about 12 months at 6.5 hours a week, or go at your own pace.
✓ No account needed. Your progress saves in this browser.
Your Roadmap
Nine phases, in order. Pass a short quiz to unlock the next one.
One Course, One Path
Built for working analysts moving into product data science. It's free. A few weeks assign books, which you can often borrow from a library.
Same Rhythm Every Week
About 6.5 hours a week, split into the same four blocks. You never have to decide what to do next.
Recall
Write down what you remember from earlier weeks before opening your notes.
Learn
Guess first, work through the material, then explain it in your own words.
Practice
Solve problems on your own, then test yourself.
Ship
Save your work to GitHub and write a short summary of what you learned.
Backed by Learning Science
Built around study methods with strong research behind them in cognitive psychology.
Test yourself
You recall ideas from memory instead of rereading. In studies, this beat rereading on later tests.
Review over time
Topics come back after a week, a month and three months, so you keep what you learn.
Mix old with new
Old topics return alongside new ones, just like a real interview.
Pass before moving on
A short quiz unlocks each phase, so gaps get fixed early.
See the research
These studies tested the methods with other subjects and learners. No study has tested this exact course.
- Self-testing: Roediger & Karpicke, 2006; Dunlosky et al., 2013
- Spaced review: Cepeda et al., 2006, mostly memory for words and facts
- Mixing topics: Rohrer et al., 2015, tested in school math
- Trying before learning: Kornell et al., 2009, lab memory studies
- If-then study plans: Gollwitzer & Sheeran, 2006
- Passing before moving on: Kulik et al., 1990
- Short pauses: Buch et al., 2021, tested on a finger-tapping skill
- Quiet rest after learning: Dewar et al., 2012, tested in older adults
- Sleep: Diekelmann & Born, 2010
- Phones out of reach: Böttger et al., 2023, though the evidence is mixed
Want to go deeper? Try Andrew Huberman's episode Optimal Protocols for Studying & Learning. Data Craft isn't affiliated with him or the show.
Your Study Routine
The same five steps every time. The lessons remind you.
Before
Start at your planned time, with your phone out of reach.
During
Try each problem before looking anything up. Pause briefly between problems.
Right after
Explain the idea out loud, then take 10 quiet minutes before your phone.
Next session
Write what you remember from last time before starting anything new.
Tonight
Protect your sleep. It helps new memories last.
Four Portfolio Projects
Each one answers a real business question and ends in a recommendation, not just charts.
1 · Product health report
"How healthy is this business, and where should the team focus?"
2 · Experiment analysis
"Should we launch this change?"
3 · Metrics framework
"How should success be measured, and what if it drops 10%?"
4 · Capstone decision
One real question, from framing to recommendation.
What you'll deliver and how it's scored
- Project 1 (Weeks 9–10): SQL repo, funnel and cohort charts, 1-page memo.
- Project 2 (Weeks 29–30): design doc, sanity checks, analysis notebook, launch recommendation.
- Project 3 (Weeks 35–36): HEART table, metric tree, diagnosis memo.
- Project 4 (Weeks 39–42): AI-use and verification log, memo and 5-minute video.
Rubric (score each 1–4; pass = 3+ on all)
- Question: clear, linked to a decision, scoped
- Correctness: checks shown, edge cases handled
- Recommendation: specific, with risks and confidence stated
- Communication: answer first, readable in 2 minutes
Make AI Test You, Not Answer for You
Copy a prompt into any AI chat and get quizzed.
See the prompts
Concept quiz
Quiz me on [topic] with 5 questions that get harder, at product data scientist interview level. Ask one at a time. Don't reveal answers until I reply. After all 5, tell me my weakest area and give me 1 Anki card for it.Product case interviewer
Act as a product data scientist interviewer at a top tech company. Give me one product case on [metric drop / feature success / launch trade-off]. Ask follow-ups like a real interviewer. Don't help me. At the end, score me 1–4 on: problem structure, metric choice, statistical rigor, recommendation. Be strict.Experiment design drill
Give me a one-line feature idea. I'll write a full A/B test design. Then critique it on hypothesis, metrics, randomization unit, sample size, duration, and risks. Point out anything an interviewer would catch.What's Changing in the Field
Questions or feedback?
We'd love to hear from you.