Product & Analytics Data Science
A study guide for Data Scientist loops at modern AI companies where the work is product analytics, experimentation, and analytics leadership — calibrated against the HeyGen founding DS role and the Cohere Lead Data Scientist role (Analytics & Data Insights). 17 interactive HTML chapters with sticky navigation, syntax-highlighted code, copy buttons, drill modes, and dark mode.
If you only have an hour
Open 00-START-HERE first. Then drill 15-interview-questions with drill mode on. Reread 16-day-of the morning of.
Drill: Interview Questions
Hide-show answers · per-Q tracker
Drill: Coding & SQL Problems
Python + SQL · multiple approaches
Core: Experimentation
A/B, power, sample size, pitfalls
Section A · Orient
00
Start Here
Master index, study schedule, the single reframe that makes the rest of the guide click.
01
The Roles Decoded
HeyGen founding DS and Cohere Lead DS, side by side. What each team actually does, what the stack means, what the loop is really testing.
02
Positioning From Scratch
Mindset before content — how to interview honestly when you don't have a direct analog on your résumé, and where to lean on adjacent experience.
Section B · Core Data Science
03
SQL for Product Analytics
Window functions, cohort and funnel queries, retention curves, attribution, the gotchas interviewers love (NULLs, gaps-and-islands, dedupe).
04
Experimentation Foundations
A/B testing math, hypothesis tests, power and sample size, peeking, p-hacking, common pitfalls, and how to design an experiment from a fuzzy product question.
05
Advanced Experimentation
Multi-armed bandits, sequential testing, CUPED variance reduction, switchback tests, network effects and SUTVA violations, ratio metrics and the delta method.
06
Causal Inference
When you can't randomize — difference-in-differences, instrumental variables, propensity scoring, regression discontinuity, synthetic controls. The Cohere JD calls this out by name.
07
Product Metrics
North star, input vs output, leading vs lagging, guardrails, funnels, retention curves, sensitivity. How to design a metric and defend it under pressure.
08
Predictive Modeling for Business
Forecasting, segmentation, propensity scoring, opportunity sizing. Models that drive sales, marketing, and finance decisions — not ML for its own sake.
09
Analytics Leadership
Setting the technical bar, prioritizing across workstreams, building narrative from ambiguous questions, mentoring, and shipping insights leadership can act on. The Cohere Lead DS bar.
Section C · Coding
10
Coding Fundamentals
Python patterns DS interviewers reach for — dict-based aggregations, generators, pandas vs SQL, vectorization, sorting tricks, sliding window, Big-O cheat sheet.
11
Coding & SQL Problems
Mixed drill set — Python data-manipulation problems and SQL puzzles (funnel, cohort retention, sessionization, gaps-and-islands). Multiple approaches per problem, drill mode hides solutions.
Section D · Production / Cloud
12
Defining the Data Stack
Greenfield problem — how to define the experimentation, warehouse, and event-tracking stack at an early-stage company. The HeyGen "lay the foundation" prompt, answered.
13
Dashboards & BI
Self-serve principles, Looker / Mode / Hex patterns, metric layers, dashboard hygiene, the "vanity vs decision" distinction stakeholders never name but always feel.
Section E · Reference & Execution
14
LLM & AI-SaaS Domain Context
Vocabulary you need to be credible at an AI-native company — tokens, fine-tuning, evals, latency budgets, the economics of inference, and how product analytics for AI products differs from classical SaaS.
15
Practice Interview Questions
~30 Q&A across SQL, experimentation, causal, product sense, ML, and leadership. Drill mode hides answers; per-question practice tracker; progress saved locally.
16
Day-Of Tactics
Structural moves, traps to watch, recovery patterns, the right questions to ask them, closing statement. Reread morning of.
Study Paths
If you have 7+ days
- Day 1: 01, 02 (orient) → 03 (SQL — the universal screen)
- Day 2: 04, 05 (experimentation)
- Day 3: 06 (causal — critical for Cohere), 07 (metrics)
- Day 4: 08 (predictive modeling), 09 (leadership — Cohere only, skim if HeyGen-only)
- Day 5: 10, 11 (coding + SQL drills on a timer)
- Day 6: 12, 13 (data stack, BI — skim if Cohere-only), 14 (LLM context)
- Day 7: Drill 15 in drill mode. Reread 16. Sleep.
If you have 2–3 days
01, 02, 03 (SQL), 04 (experimentation foundations), 06 (causal — at least pass through propensity + DiD), 07 (metrics), 11 (drill 4–5 problems), 15 (drill), 16. Skim everything else.
If you have < 24 hours
01, 02, 03 SQL gotchas section only, 04 (just the design-an-experiment recipe), 07, 15 (drill all questions), 16. Skim 06 headings only.