Data Science Study Group: How to Find or Start One

A data science study group works on the same skills each week (SQL, statistics, a portfolio piece) and reviews each other's work at a weekly meeting. Online, the longest-running ones are book clubs, a weekly dataset challenge and free course cohorts, and you can start your own around one course or one public dataset.

Book clubs, challenges and cohorts to join

We opened each of these on November 14, 2026:

The Data Science Learning Community has run its weekly book clubs since early 2020 and grown to thousands of members. You join its free Slack, then pick a club from its calendar or propose one around any free online book. It also organizes TidyTuesday, which welcomes R, Python, Julia or any other language. The Zoomcamp suits a group that wants SQL at scale: its modules run through PostgreSQL, BigQuery and dbt, and only the live cohort gets scored homework and a leaderboard. A run with dates also answers one of the main reasons self-paced online courses stall: nothing sets the pace.

On StudyDuo, Data science is one of the interests you can search in Find Partners, with a country or city if you'd like to meet locally. In any Slack, Discord or Telegram group, keep your phone number and logins to yourself.

What a data science study group should do together

Statistics has the clearest evidence on learning in groups. Sema Kalaian and Rafa Kasim pooled nine studies of undergraduate statistics courses, 926 students in all, that compared small-group learning with lectures (Journal of Statistics Education, 2014). Working through problems in small groups raised exam scores; exploring a problem before learning its terms didn't:

By the authors' estimate, an effect of 0.60 lifts a student from the 50th to the 73rd percentile on a statistics exam. The evidence is thin, though. Studies published from 2001 on averaged 0.20, and the three experimental or quasi-experimental ones averaged 0.16.

Structured and unstructured group work did about equally well, so the order matters more than the format: read the chapter first, then solve together. To keep everyone involved, borrow the cooperative format and rotate roles each week, such as a solver, a checker and someone who explains the answer in plain words. Split the other skills the same way:

A week built on TidyTuesday

A data science study group needs a shared assignment that nobody has to prepare, and a public dataset that changes every week is one. TidyTuesday posts its data on Monday mornings. Week 45 came as three CSV files and a list of suggested questions, such as which fictional franchises undo their characters' deaths most often. Its notes also warn that the rankings are a fan site's editorial judgments, not measurements, which is a statistics lesson in itself. Another suggested question, how often a lower-ranked character kills a higher-ranked one, needs the deaths file joined to the rankings file: a ready-made exercise for the swapped-query slot.

1. Monday: the new dataset appears, and each person writes one question to answer with it.
2. Tuesday to Thursday: silent work sessions on camera. Load the files into a database, answer your question with SQL, and make one chart.
3. Weekend meeting, 60 minutes: 10 minutes of swapped queries, 20 minutes on a statistics problem in roles, 25 minutes of portfolio review, 5 minutes to set next week's targets.
4. After the meeting: post your output and code with the #TidyTuesday hashtag, as the project asks. A few months in, each member has a public record of finished analyses.

Job hunting as a group

In the US, the Bureau of Labor Statistics counted 275,600 data scientist jobs in 2025 (Occupational Outlook Handbook, updated August 27, 2026). The median wage was $120,230 in May 2025. It projects 35% growth from 2025 to 2035, about 24,800 openings a year. Most employers want at least a bachelor's degree in math, statistics, computer science or a related field.

Studies of job search programs suggest what a group can add. Songqi Liu, Jason Huang and Mo Wang pooled 47 programs tested in experiments or quasi-experiments (Psychological Bulletin, 2014). Participants' odds of getting a job were 2.67 times those of people without a program. The stronger programs taught job search skills, improved self-presentation, built confidence, encouraged initiative, set goals and enlisted social support. They worked only when they built skills and motivation together.

Those were organized programs, often for unemployed people, not peer groups. A group can still cover several of the parts:

• Goals: each member names a weekly number, such as five applications or two messages to people in the field, and reports it at the meeting.
• Self-presentation: one portfolio README or one resume section gets reviewed each week.
• Skills: practice explaining one project in two minutes: the question, the data, what you found. The BLS lists communication among the qualities data scientists need, alongside analytical, math and problem-solving skills.
• Support: keep a shared tracker of applications and replies, so a slow month is visible and the group can help.

For coding interviews, our LeetCode study group guide covers problem lists and mock interviews. If your group would rather read machine learning papers, see Machine Learning Study Group.

Where StudyDuo fits

We make StudyDuo, and it fits the days between meetings. Focus Room never closes: put this week's question in your task, point the camera at your face or your desk, and work in a silent room. For a data science study group spread across countries, Study Hour is a single daily hour that each member's app converts to local time. Tasks in the app can carry reminders for the weekly application target. A 1:1, which has sound, works for walking someone through your portfolio; both of you need a verified photo. More on running a group this way is on the online study group page.