That moment of scepticism, not a fancy algorithm, is the core of the job.
Pay, licensing and outlook figures below are mainly for the United States; "Around the world" shows how the job works elsewhere.
What the job actually is
A data scientist uses statistics, programming and knowledge of a field to turn messy data into answers, predictions and decisions that an organisation can act on.
The biggest misconceptions:
- "It's all machine learning and AI." A lot of it is finding, cleaning and understanding data. In Anaconda's annual surveys of data professionals, loading and cleaning data took about 38–45% of their time (self-reported, 2020–2022). The often-quoted "80% of the job is cleaning" overstates it, but data preparation is still one of the biggest parts of the job.
- "You need a PhD." The BLS lists a bachelor's degree as the typical entry level, though some employers prefer a master's or doctorate, especially for research roles.
- "It's a solitary coding job." The BLS lists presenting findings to technical and non-technical audiences and recommending business decisions as core duties. Communication can decide whether your analysis gets used at all.
The path in
United States. The typical route is a bachelor's degree in mathematics, statistics, computer science or a related field such as business, engineering or economics. Many working data scientists also hold a master's. The core toolkit is:
- Statistics and probability. The BLS recommends linear algebra, calculus and statistics as early as high school.
- Programming. Python or R, plus SQL for pulling data from databases.
- Communication. Charts, clear writing, and explaining uncertainty honestly.
There is no licence or required exam. Certificates exist, but portfolios and interviews (often with a take-home problem) carry more weight.
Alternative routes. Many people move in from science, economics, psychology, engineering or finance, adding programming as they go. Bootcamps and free online courses can help, but the market for junior roles is competitive, so a portfolio of real, well-explained projects matters. Many start as data analysts and grow into data science.
Elsewhere. In the UK, routes include degrees in maths, statistics, computer science or data science, postgraduate conversion master's for graduates of other subjects, and apprenticeships, including a Level 6 Data Scientist degree apprenticeship and Level 7 options. Across Europe and Latin America the pattern is similar: no licence, and employers look for statistics, programming and evidence of real projects. Job titles vary a lot between companies and countries.
A data scientist's most important sentence is often "I'm not sure the data can answer that," said early and politely.
A typical day
One plausible day at a mid-sized online retailer:
| Time | What's happening |
|---|---|
| 9:00 | Check overnight dashboards; one metric dropped sharply, so investigate |
| 9:30 | It's a broken data feed, not a real drop. Tell the team |
| 10:00 | Write SQL to pull six months of order data for a pricing question |
| 11:30 | Clean it: duplicates, missing values, mismatched time zones |
| 12:30 | Lunch |
| 1:30 | Explore patterns in Python; draft three charts |
| 3:00 | Meet product managers: what decision does this analysis feed? |
| 4:00 | Code review of a colleague's model |
| 5:00 | Write a one-page summary in plain English, with caveats |
How it varies. A product analytics role is all about experiments (A/B tests) and user behaviour. A machine-learning role builds and maintains prediction models, often working closely with software engineers. In healthcare, government or nonprofits, the questions are slower and the data may be sensitive. In research, the work looks more like academic science.
Where it can take you
| Role | What it involves |
|---|---|
| Data analyst → data scientist → senior / lead | The common ladder |
| Machine learning engineer | Getting models running reliably in real products |
| Data engineer | Building the pipelines that move and store data |
| Analytics / experimentation specialist | Designing and reading A/B tests |
| Statistician / biostatistician | Rigorous inference, often in health and science |
| Business intelligence analyst | Dashboards and reporting for decision-makers |
| Head of data / analytics manager | Leading teams and setting data strategy |
Adjacent careers the training opens: operations research, actuarial work, quantitative finance, policy analysis, epidemiology, and AI product roles.
Who thrives, who struggles
Thrives: people who are curious, sceptical of results that look too good, patient with messy data, comfortable with both maths and code, and able to explain a complicated result to a busy manager in two minutes. Knowing the field you work in (sales, medicine, sport) makes your analysis much better.
Struggles: people who want clean problems with one right answer, who dislike ambiguity, or who see communication as someone else's job. If tedious data cleaning drains you completely, a large part of the work will feel heavy.
Money & outlook
- US median pay: $120,230 a year (May 2025, BLS). By industry, the BLS reports computer systems design at $132,380 and insurance carriers at $108,650.
- Projected growth: 35% from 2025 to 2035, far above the 3% average for all occupations, with about 24,800 openings a year. The BLS credits rising demand for data-driven decisions, the growing volume of data, and companies building AI into their work.
- UK comparison: the National Careers Service gives £32,000 (starter) to £83,000 (experienced).
Pay varies widely by industry, city, seniority and the scope of the role. "Data scientist" covers everything from reporting-heavy analyst jobs to research-level modelling.
Around the world
The toolkit is the same everywhere: statistics, code, SQL and clear explanation. No country in this table licenses data scientists. What changes is the job title, the pay, whether official statistics even count the job separately, and the data-protection laws you work under, which increasingly set rules for what models may do.
| Country | How you qualify | Who licenses / regulates | Pay in context | Worth knowing |
|---|---|---|---|---|
| United States | Bachelor's in maths, statistics, computer science or similar; many hold a master's | No licence | Median $120,230 a year (BLS, May 2025) | The BLS projects 35% growth, 2025–35 |
| United Kingdom | Degree, a conversion master's, or a Level 6 or Level 7 apprenticeship | No licence; personal data is covered by UK data-protection law | £32,000 starter to £83,000 experienced (National Careers Service) | Degree apprenticeships let you train while employed |
| Germany | Degree in maths, statistics, computer science or similar; a master's is common | No licence; the EU's GDPR and AI Act set rules for the data and the systems, not the person | — | The EU AI Act (Regulation 2024/1689), in force since August 2024, adds duties for "high-risk" systems, such as those used in hiring or credit scoring |
| Mexico | Licenciatura in maths, actuarial science, computer science or similar | No licence | Maths graduates are the best paid of all degrees in IMCO's ranking, averaging MX$29,887 a month; computer science graduates average MX$24,983 (IMCO Compara Carreras 2026) | IMCO has no separate "data science" degree; these are the nearest ones |
| Brazil | Degree in statistics, computer science, engineering or similar | No licence for data scientists; data protection under the LGPD (Law 13.709/2018) | — | "Statistician" is a regulated profession in Brazil (Law 4.739/1965), overseen by federal and regional councils; "data scientist" is not |
| India | Degree in computer science, statistics, maths or engineering (B.Tech, B.Sc, M.Sc) | No licence | — | The Digital Personal Data Protection Act 2023 sets national rules, including consent, for handling people's digital personal data |
| Canada | Degree in a quantitative field; a master's is common | No licence | Median C$46.15 an hour (Job Bank, 2023–2024) | Canada's occupation classification lists data scientists as their own occupation (NOC 21211) |
Moving abroad. With no licence to recognise, the hurdles are work permits and, for some jobs, language. Two things travel well: a degree in a quantitative field and a public portfolio of projects that anyone can read. In Europe, degrees are measured in ECTS credits, which makes them easier to compare between countries. The law travels less well: a model that is routine in one country can need documentation, risk checks and human oversight under the EU AI Act in another, so learn the data rules of the place you move to.
The honest downsides
- A crowded entry level. Fast projected growth hasn't made junior jobs easy to get. Many applicants have similar certificates, so standing out takes real projects.
- The tools keep changing. Libraries, platforms and AI tools turn over fast, so you'll keep learning on your own time.
- Your work can be ignored, or misused. Analyses get overruled by gut feeling, or quoted without their caveats.
- Ethical weight. Models can encode bias or invade privacy. You may have to push back on how data is used.
Try it before you commit
- Analyse something you care about. Download a free public dataset (sports stats, city open data, weather) and answer one question with a chart.
- Take a free course. Kaggle Learn's short courses or the free online book R for Data Science both work in a browser.
- Learn basic SQL. Free interactive tutorials cover the basics in a weekend, and almost every data job uses it.
- Publish one project on GitHub or a blog, explaining what you did, what went wrong and what you'd do next.
- Ask a data scientist what their last week actually looked like, and how much of it was cleaning.
The takeaway
Data scientists turn messy data into decisions using statistics, code and clear communication. Much of the time goes into preparing data and much of the value comes from honest scepticism, while the machine-learning glamour is a smaller share. US demand is projected to grow very fast and the pay is high, but entry-level competition is real, so a portfolio of well-explained projects is your best way in.
Try this: this weekend, pick a free public dataset about something you love, ask one specific question, and make one chart that answers it. Then write three sentences on what the data can't tell you.
Sources & Further Reading
Sources
- US Bureau of Labor Statistics, Occupational Outlook Handbook: Data Scientists (pay, May 2025; projections, 2025–35, and stated drivers; duties; education).
- ONET OnLine: Data Scientists, 15-2051.00 (interest code Investigative–Conventional).
- Anaconda, State of Data Science surveys, 2020–2022 (self-reported share of time on data preparation: about 45% in 2020, 39% in 2021, about 38% in 2022).
- National Careers Service (UK): Data scientist job profile.
- IMCO, Compara Carreras 2026*: Matemáticas (ranked first for income) and Ciencias computacionales (average monthly income in 2026 pesos; based on INEGI's ENOE survey).
- Regulation (EU) 2024/1689 (the AI Act; in force 1 August 2024; high-risk uses listed in Annex III include employment and creditworthiness). Regulation (EU) 2016/679 (GDPR).
- Brazil: Law 13.709/2018 (LGPD); Law 4.739/1965 (statistician profession and its councils).
- India: Digital Personal Data Protection Act, 2023.
- Job Bank (Government of Canada): Data scientist wages, NOC 21211 (updated November 2025).
Go deeper
- 🌐 BLS Occupational Outlook Handbook: Data Scientists
- 🌐 O*NET OnLine: Data Scientists
- 🌐 Kaggle Learn, free hands-on courses · 🌐 R for Data Science, a free online book
- 🌐 This Is Statistics, the American Statistical Association's careers site
- 📘 Naked Statistics, Charles Wheelan (statistics intuition without the pain)