Data Scientist in the US

Quick answer: what is a Data Scientist, and how do you become one?

The 5-stage data scientist roadmap at a glance

The first four stages take you to entry-level data scientist skills and total roughly 39 to 63 weeks of part-time study from zero, about 9 to 14.5 months at 10 hours a week. This page rounds that to a suggested planning range of 9 to 15 months to leave room for portfolio work. The ranges assume you start with no coding and no statistics beyond high school, study about 10 hours a week, and build the project named at each stage. The fifth stage, production machine learning and leadership, comes after your first role. If you already work as an analyst or hold a quantitative degree or PhD, you start further along, and the estimates by background run from 4 to 11 months. Both figures are study time and neither includes the job search: the stage sum measures the full path from zero, while the background estimates count only the stages each starting point still needs.

Every range is a suggested planning range built from official course durations for closely matching content, compiled October 5, 2026, not a measured outcome. Each stage is expanded with what you learn, what you build and how you know you’re done in the full roadmap, where the evidence behind each estimate is set out.

What is a Data Scientist?

A data scientist builds, validates and explains statistical and machine learning models, experiments and forecasts that answer business questions, using programming, statistics and domain knowledge. Where a data analyst mainly reports on what happened, a data scientist predicts outcomes and tests what drives them. The deliverable is a validated statistical or machine learning model, experiment or forecast that informs a business decision.

Two neighboring roles pin the definition down. A Data Analyst turns data into answers. A Data Scientist builds statistical and machine learning models that predict outcomes and test what drives them. A Data Scientist builds and validates the model that answers a business question. A Machine Learning Engineer typically owns the model in production. In practice the boundaries overlap, and employers divide the work differently. The full comparison, including data scientist vs data analyst, is under Data Scientist vs Data Analyst vs Machine Learning Engineer.

Is data scientist a real job title?

Yes. Data Scientist has its own federal occupation code, SOC 15-2051, and O*NET code 15-2051.00, and six US salary sources, Glassdoor, Indeed, Payscale, Salary.com, Levels.fyi and Robert Half, publish pay for the exact title. On October 5, 2026, LinkedIn showed 11,000+ and Indeed about 8,000 US listings matching 'data scientist'. Seniority variants of the same title appear on published salary ladders, such as Junior, Lead and Principal Data Scientist on Indeed and Data Scientist I to VI on Salary.com.

What the government occupation code covers

Government figures on this page describe the BLS Data Scientists occupation, not only people with the job title. O*NET lists two other detailed occupations under the same code: Business Intelligence Analysts (15-2051.01) and Clinical Data Managers (15-2051.02). Statisticians have a separate code, 15-2041, and are not included. Because the code is shared, every government figure on this page carries that caveat.

Is an AI and data scientist a different role?

Not as a separate occupation. The federal occupation taxonomy has no separate title for an AI and data scientist; the work sits within the Data Scientist occupation, and employers use the titles differently. Machine learning is already part of the role: in 365 Data Science's analysis of 1,121 US Glassdoor postings, published April 2026, 77% asked for machine learning skills. Building applications on large language models is a typical emphasis of AI engineer roles, though employers divide that work differently; the tracks are set out in the roadmap.

What does a Data Scientist do day to day?

A data scientist's work runs from a business question to a decision. Most of the responsibilities below appear in the federal description of the occupation. This guide emphasizes experiment design and causal inference, a typical emphasis of the role rather than skills unique to it.

Compiled from O*NET 15-2051.00 tasks (clean and manipulate raw data; apply feature selection; test, validate and reformulate models; compare models using statistical performance metrics; present results), the BLS description of the occupation, GitLab's published data science ladder and the US postings reviewed for this page, accessed October 5, 2026. This page makes no claim about how the working week divides between these responsibilities.

How the work differs by industry

The core methods carry across industries; the domain knowledge you add changes. The list below is descriptive, and pay by industry is set out under salary.

• Finance and insurance: credit, fraud and pricing problems, which add knowledge of risk and of how lending and insurance decisions are regulated.
• Healthcare and life sciences: clinical and biological data, which add study design, survival analysis and patient-privacy rules.
• Technology and e-commerce: product experiments, recommendations and demand forecasting at large scale.
• Government: federal data scientist positions in occupational series 1560, such as the U.S. Department of State and Department of the Air Force postings reviewed for this page.
• Publishing and media: the industry with the highest median wage among the five that BLS lists as top employers of data scientists.

Industry list from the BLS top employing industries (May 2025) and federal postings retrieved October 5, 2026. The domain descriptions are editorial guidance written for this page.

Data Scientist salary in the United States

The official figure is a $120,230 median annual wage for the BLS Data Scientists occupation in May 2025; the lowest 10% earned less than $67,240 and the highest 10% more than $199,130. That occupation code also covers business intelligence analysts and clinical data managers, so it describes a broader group than the job title alone. Salary platforms publish figures for the exact title from $103,820 (Payscale average base salary) to $181,800 (Levels.fyi median total compensation). The gap reflects what each one measures and who it samples, not two different markets.

Pay figures on this page are for the Data Scientist role in the United States, in US dollars per year. The official figure is the US Bureau of Labor Statistics wage for its Data Scientists occupation (May 2025). Platform figures come from Glassdoor, Indeed, Payscale, Salary.com, Levels.fyi and Robert Half; each measures pay differently (base salary, total pay, advertised pay or projected benchmarks), so they are shown separately and never averaged.

The four ways data scientist pay is measured

Families as grouped in the salary research behind this page, from source pages accessed October 5, 2026. Read each figure only against the family it belongs to.

What each salary source reports for data scientists

Each row is a separate source with its own method, sample and pay definition, so the rows are not comparable to one another and are not averaged. Glassdoor's $125,347 and $203,643 are its 25th and 75th percentiles, not the lowest and highest salaries. Glassdoor's page was read through a country redirect and is flagged for recapture from a US session. Robert Half's figures are projections for 2027 from its own placements, not observed pay.

Data scientist salary by level

Every source that publishes a ladder shows pay rising with level, but the ladders are built differently and do not line up: Salary.com uses job levels, Indeed uses job titles, Payscale uses years of experience and total compensation, and Robert Half uses qualitative tiers projected to 2027. Each ladder is shown as its source published it.

Source: Salary.com, average base salary, data updated October 1, 2026. Salary.com does not state the years of experience behind each level in the page captured for this research.

Source: Indeed, average base salary from job postings, updated September 28, 2026. Seniority is implied by the job title, not by years. Indeed also lists Chief Data Scientist, an executive title, and an hourly internship; both are different jobs and are left out.

Source: Payscale, average total compensation including bonus and overtime, updated July 14, 2026. Payscale publishes its mid-career, experienced and late-career bands only as relative changes with no dollar figure, so no figure is given for them.

Source: Robert Half, 2027 Salary Guide, read October 5, 2026. These are projected benchmarks from Robert Half's placements and posting data, and the guide does not say whether they are base or total pay.

Entry-level figures conflict, and the conflict is reported rather than averaged. Indeed's Junior Data Scientist average of $85,488 (advertised base) and Payscale's under-1-year average of $88,797 (self-reported total compensation) sit well below Robert Half's low tier of $123,750. The likely causes are base versus total pay, and a staffing-firm sample made up of candidates it placed, which the salary research behind this page reads as skewed toward larger employers and experienced hires, projected forward to 2027. A higher entry-level figure circulating in a secondary re-report of Glassdoor data is not published, because its pay type and filtering are not stated.

Data scientist salary by US metro and state

Location is the clearest driver in the government data. The San Jose metro median is $185,080 and the Houston median $106,750, against a national median of $120,230.

Source: US Bureau of Labor Statistics OEWS, May 2025 wage data, read through O*NET OnLine's local-wage tables for California, Washington, New York, Massachusetts, Texas, accessed October 5, 2026. Medians with 10th to 90th percentile ranges, for the whole Data Scientists occupation code. Metro figures for Washington DC, Chicago, Atlanta and Denver were not retrieved and are not estimated.

Source: US Bureau of Labor Statistics OEWS, May 2025, through O*NET OnLine state tables (California, Washington, New York, Massachusetts, Texas), accessed October 5, 2026.

Indeed publishes city averages from job postings. They measure advertised base pay, not wages, and are not comparable with the government rows above.

Source: Indeed, average base salary from job postings, updated September 28, 2026. Per-city sample sizes are not disclosed. Cities are listed as Indeed published them.

Data scientist salary by industry

Source: US Bureau of Labor Statistics, Occupational Outlook Handbook, Data Scientists, May 2025. These are medians in the five industries that employ the most data scientists, not a ranking of every industry. Industry figures from a salary platform are not published, because the per-industry samples were undisclosed and the top of its list was dominated by a few employers.

What moves a data scientist's salary

Drivers grouped from the source rows above; no new figures. Skill, certification and degree premiums are not listed as drivers; see the next question.

Do any skills carry a verified salary premium?

None was found in the sources reviewed for this page (listed under methodology, accessed October 5, 2026). Employers ask most often for Python (85% of 1,121 sampled postings), machine learning (77%) and SQL (59%), according to 365 Data Science's 2026 analysis of US Glassdoor postings, but none of those sources puts a dollar or percentage premium on these skills, or on any certification or degree level, for data scientists. How often postings ask for each skill is set out under skills.

Data scientist and data analyst pay: where each role's figures are

This page does not compare data scientist pay with data analyst pay. Each role's figures come from their own sources and dates, and placing them side by side would mix measurements that were never taken the same way. For data scientists, the government median is $120,230 (May 2025), with the platform figures above. Data analyst pay is set out on the Data Analyst career roadmap.

The official wage baseline

Data Scientist has its own federal occupation code, so the government figure is the role's own baseline, not a proxy. Its one caveat is the code's scope.

Source: US Bureau of Labor Statistics, Occupational Outlook Handbook, Data Scientists, May 2025 wages, page last modified August 27, 2026, accessed October 5, 2026. The BLS occupation page with the mean wage and the 25th and 75th percentiles could not be retrieved, so those figures are not given.

Data Scientist jobs and hiring demand in the US

Data scientist employment is projected to grow fast over the decade, while the most recent sector posting data in this research, from late 2025, was weak. The Bureau of Labor Statistics projects 35% growth for its Data Scientists occupation from 2025 to 2035, from a base of 275,600 jobs to 371,000, with about 24,800 openings a year, including openings from workers who leave the occupation. In late October 2025, Indeed Hiring Lab's job postings index for the data and analytics sector was 60.4, the lowest of any sector it tracked, while applications per posting were rising. That is dated history; it does not establish current hiring conditions.

How fast is data scientist employment projected to grow?

Source: US Bureau of Labor Statistics, Occupational Outlook Handbook, Data Scientists, 2025-2035 projections, page last modified August 27, 2026, accessed October 5, 2026. Older projection vintages still shown elsewhere, including on O*NET, are not mixed in.

How many data scientist jobs are listed?

Single-day snapshots. Keyword counts include senior and staff roles and duplicates across boards, so the two are never added together. A Glassdoor figure that counts companies with salary data, and an Indeed figure that counts salaries from postings over 36 months, are not job counts and are not used.

What education do data scientist postings ask for?

The government and the postings point in different directions, and both are reported. The Bureau of Labor Statistics lists a bachelor's degree as the typical entry-level education, with no related work experience and no on-the-job training. In 365 Data Science's posting sample, graduate degrees were named more often than a bachelor's.

Sources: 365 Data Science, Data Scientist Job Outlook 2026, published April 23, 2026 (collection window not stated); O*NET OnLine, Data Scientists 15-2051.00. The two measure different things: the first counts the education level a posting names, the second what survey respondents say the job requires. Shares are of each source only and are not summed.

How much experience do data scientist postings ask for?

In the individual postings reviewed for this page, entry roles accept new graduates or 0-2 years, the Data Scientist rung asks for about 2-5 years and Principal roles 8+ years; the full ladder is under career progression. The one posting analysis in this research gives its 2026 experience shares only as a chart, so no market-wide share is published here. As dated context, Indeed Hiring Lab reported in November 2025 that entry-level roles had held their share of postings, or grown it, but because total postings shrank, that still left fewer openings for new graduates.

Which employers are advertising data scientist roles?

Employer names are plain text and are not linked. A name means the employer had a Data Scientist-titled listing or posting on the date shown, not that it hires in volume; listings change daily, and no ranking is implied. Whether any listed name is a staffing intermediary rather than the end employer was not checked. Company logos shown on Simplilearn program pages are not evidence of hiring and are not used.

Which industries employ the most data scientists?

The Bureau of Labor Statistics lists five top employing industries for the occupation: publishing, broadcasting and content providers; computer systems design and related services; credit intermediation and related activities; management of companies and enterprises; and insurance carriers and related activities. Median wages in each are under salary.

Which job titles do postings use?

Seniority variants of the title appear on salary ladders and in postings: Junior, Lead and Principal Data Scientist on Indeed, Data Scientist I to VI on Salary.com, and Associate, Senior Associate and Principal Associate at Capital One. How often each title appears was not measured, so no distribution is given. Executive titles and internships are treated as different jobs.

What tempers the demand picture

• In late October 2025, Indeed Hiring Lab's data and analytics sector index was 60.4, the lowest of any sector it tracked. That is dated history, not current conditions; the sector covers data and analytics broadly, not the Data Scientist title alone, and no 2026 sector figure was found.
• In November 2025, Indeed reported US tech job postings almost a third below their early-2020 level.
• The 35% projection covers a code that includes business intelligence analysts and clinical data managers, and it is a ten-year projection, not current hiring.
• Listing counts are single-day keyword snapshots, not unique open positions.

Sources: Indeed Hiring Lab, 2026 US Jobs and Hiring Trends Report, November 20, 2025; US Bureau of Labor Statistics, Occupational Outlook Handbook, Data Scientists; LinkedIn and Indeed job searches, October 5, 2026.

Data Scientist career progression and future scope

The data scientist ladder runs from an entry or associate rung through Data Scientist, Senior, Staff and Principal on the individual-contributor track, with a separate management track. Simplilearn's editorial role definition describes the seniority signal as owning the modeling approach and its validation for a product or business area, and mentoring others on method. Titles vary by employer: Capital One uses Associate, Senior Associate and Principal Associate, and DoorDash uses numbered levels.

Rungs, experience and scope from postings and ladders retrieved October 5, 2026: GitLab's data science job family handbook, Waymo, DoorDash, Capital One, Toast and Walmart postings, and USAJOBS federal postings. Pay: Indeed, average base salary from postings, updated September 28, 2026, matched to a rung by title only. Indeed's Lead Data Scientist ($162,775) has no matching rung here. Individual employers' posted ranges are not shown, because one employer's range is not a market rate. Promotion criteria differ by employer.

Individual contributor or manager?

Both tracks exist. GitLab's published data science family is individual-contributor only, a public career ladder framework on GitHub splits the individual and management tracks at its Data Scientist V level, and the federal government advertises supervisory data scientist positions at a senior grade. Total years of experience for the management rung are not stated in any source reviewed.

Where the path leads after your first data scientist role

One lateral move stays close to the work: toward production machine learning. A Data Scientist builds and validates the model that answers a business question. A Machine Learning Engineer typically owns the model in production. Moving that way adds feature engineering at depth, training pipelines and serving; see the Machine Learning Engineer career roadmap. The tracks you can choose after the foundations are set out in the roadmap.

Is data science a good long-term career?

The official outlook is growth: the Bureau of Labor Statistics projects 35% employment growth for Data Scientists from 2025 to 2035 and names the integration of AI into business workflows, the growing volume of data and demand for data-driven decisions as drivers. That is a ten-year projection, not a measure of current hiring, and the dated counterweights are under jobs and hiring demand.

Data Scientist vs Data Analyst vs Machine Learning Engineer

These three roles share data, Python and statistics, and differ mainly in what they typically deliver. A Data Analyst turns data into answers. A Data Scientist builds statistical and machine learning models that predict outcomes and test what drives them. A Data Scientist builds and validates the model that answers a business question. A Machine Learning Engineer typically owns the model in production. The rows below show typical emphasis; the roles overlap, and employers divide the work differently.

Source: Simplilearn's editorial role definitions for data and business analytics and for AI roles, October 5, 2026; the Data Scientist row is reasoned from O*NET and BLS task descriptions and the skill progression behind this page. Salary is deliberately not compared: each role's pay comes from its own sources and dates, so this page states only data scientist figures. Full treatments: Data Analyst and Machine Learning Engineer career roadmaps.

Data scientist vs data analyst: the full comparison

A Data Analyst turns data into answers. A Data Scientist builds statistical and machine learning models that predict outcomes and test what drives them. The two roles overlap most at the foundation: both query data with SQL, clean it and chart it. They split at statistics and modeling. An analyst's work is complete when the question is answered and the answer is communicated; a data scientist's work continues into predicting what comes next and testing whether a change caused an effect.

Built from Simplilearn's editorial role definitions, O*NET's separation of business intelligence analysts (15-2051.01) within the Data Scientists code, and the 365 Data Science sample of 1,121 US Glassdoor postings, April 2026. The Data Analyst column summarizes that role; its skills, roadmap and pay are set out on its own page.

Pay is not compared here. Data scientist figures are under salary; data analyst figures are on the Data Analyst roadmap, from their own sources and dates.

Should you become a data analyst first?

It is one route, not a requirement. An analyst role already covers SQL, visualization and metrics, so moving up means adding math for machine learning, experiment design and machine learning itself, an estimated 6 to 10 months of part-time study. The Bureau of Labor Statistics lists no related work experience as typical for entry into the occupation, so a quantitative graduate who can code can apply for data scientist roles directly. The transition is set out under career-switch paths.

Data scientist vs machine learning engineer

A Data Scientist builds and validates the model that answers a business question. A Machine Learning Engineer typically owns the model in production. In practice production ownership varies by team: some data scientists develop and operate models, and GitLab's published ladder expects automated models in production at its Intermediate data scientist level, while other teams hand models to machine learning or platform engineers. Feature pipelines, training and tuning at production scale, and MLOps are the typical emphasis of machine learning engineering. This guide focuses on choosing, building and validating models to answer a business question, and on experiment design; see the Machine Learning Engineer career roadmap for production systems in depth.

Data science vs traditional statistics and business analytics

For a statistician, economist or business analyst, data science keeps the reasoning about evidence and changes the data, the tooling and the output. The question is more often what will happen than what happened, the data is larger and messier, and the result is checked on data the model has not seen. The table shows typical emphasis: the fields overlap in inference, prediction and experimental design.

Comparison written for this page. The seniority row quotes Simplilearn's editorial role definitions for data and business analytics and for AI roles, October 5, 2026.

How is AI changing the data scientist roadmap?

The Bureau of Labor Statistics names the integration of AI into business workflows as one driver of demand for data scientists. As editorial guidance, this roadmap treats large language model literacy as an add-on after core machine learning. Building applications on large language models, with retrieval and agents, is a typical emphasis of AI engineering roles, and it is set out on the AI Engineer career roadmap.

Is data science the same as statistics?

They overlap in inference, prediction and experimental design. Statisticians have their own federal occupation code, separate from Data Scientists, and data science roles often combine statistical methods with programming, data systems and product delivery. For statisticians, economists and actuaries this page gives an editorial estimate of 5 to 9 months of part-time study, set out under career-switch paths.

Data Scientist skills required, and skill levels

A data scientist needs fluent statistics, Python and SQL, the ability to design and analyze experiments, and enough machine learning to choose, build and validate a model that answers a business question. The skill progression for the role runs from L1 to L5, and L3 and L4 are where the role is defined.

Skill progression from the researched roadmap behind this page, levelled from O*NET tasks, the BLS description of how to become one, GitLab's data science ladder and US postings, October 5, 2026. The depth column shows how far this guide goes; where a linked guide goes deeper, the row says so.

Skill levels, from beginner to expert

Levels written for this page and used across it, including the program router.

Which skills do data scientist postings ask for?

Python, machine learning and SQL lead. In 365 Data Science's analysis of 1,121 US Glassdoor postings, published April 2026, 85% named Python, 77% machine learning and 59% SQL, with R at 46% and AWS at 27%. Statistical analysis was named as a skill in 18%.

Source: 365 Data Science, Data Scientist Job Outlook 2026, 1,121 US Glassdoor postings, published April 23, 2026; collection dates not stated. Percentages are shares of this one sample as published, not market-wide rates. Salary premiums are covered under salary.

The math a data scientist actually uses

Probability and statistics are the math you use every week; linear algebra and calculus explain how models work. The Bureau of Labor Statistics names all four as preparation for the occupation, and the depth needed at entry level differs by topic.

Topics from the BLS description of how to become a data scientist and the roadmap's first and third stages. Depth guidance written for this page.

Experimentation and A/B testing for data scientists

Experiment design, with causal inference, is a typical emphasis of data scientist roles, and it is a named requirement in the ladders and postings behind this page: GitLab's Intermediate data scientist designs and evaluates experiments through hypothesis testing, and the DoorDash, Figma and Toast postings ask for A/B testing, experimentation or causal inference. A good test is decided before it runs.

Steps written for this page from standard experimental practice, as editorial guidance. A significance test alone does not show that an effect matters: report the estimated effect, its uncertainty and the assumptions checked, and judge whether it is large enough to act on.

Causal inference when you cannot run an experiment

Many business questions cannot be randomized: a price change hits every customer, or a policy starts on a fixed date. Causal inference methods estimate an effect from observational data, and each rests on an assumption you have to defend. In the roadmap they sit in stage 3, after regression.

Methods written for this page. Causal inference appears in the Figma and Toast postings reviewed for this page.

How much machine learning does a data scientist need before applying?

Enough to choose, build and validate a model honestly, with depth in one specialty. GitLab's ladder expects machine learning with feature engineering and validation at its Intermediate level and adds deep learning, NLP and recommendation systems at Senior. In the 1,121-posting sample, 77% asked for machine learning and 20% for deep learning.

• This page's entry-level checklist: linear and logistic regression, decision trees and ensembles, clustering, cross-validation, evaluation metrics such as precision, recall and ROC-AUC, and comparison against a simple baseline.
• Later, or in one chosen specialty: deep learning in TensorFlow or PyTorch, NLP, time-series forecasting at depth, computer vision and recommendation systems.

Feature engineering: what data scientists need to know

Feature engineering is creating and selecting the inputs a model learns from, and a data scientist uses it in every predictive project. At this role's depth that means sensible encodings, ratios and aggregates built from domain knowledge, feature selection that is validated rather than assumed, and strict care that nothing from the future leaks into training.

Feature pipelines and feature stores at production scale are a typical emphasis of machine learning engineering, though in some teams data scientists build and maintain them too. This guide stays with feature work for a single predictive project; see the Machine Learning Engineer career roadmap for production feature systems in depth.

Should you learn Python, R or SQL first?

Learn Python and SQL first, together if you can. SQL gets you the data and Python does the analysis and modeling. Add R later if the roles you want name it.

Shares from 365 Data Science, Data Scientist Job Outlook 2026, 1,121 US Glassdoor postings, April 2026. Order recommendation written for this page.

Libraries and tools compared

This page recommends scikit-learn as the first modeling library, because it teaches validation and model comparison on the classical models stage 4 of the roadmap starts with. That is a recommendation, not a finding: the posting sample measured TensorFlow (23%) and PyTorch (21%) by name but did not measure scikit-learn, pandas or NumPy.

Best-at descriptions written for this page. "Not measured" means the sample did not report the tool, not that postings never name it.

Do employers expect data scientists to know Tableau or Power BI?

Some do. In the 1,121-posting sample, 19% named Tableau and 14% Power BI, and 22% named data visualization in general. This page treats Python charting as the core skill for a data scientist and places BI tools mainly on the analytics and business track. BI tools in depth are covered on the Data Analyst career roadmap.

Which soft skills does a data scientist need?

• Problem framing: turning a vague request into a question that data can answer, and saying when it cannot.
• Data storytelling: presenting a result and its uncertainty so a non-specialist can act on it, which O*NET lists among the occupation's tasks.
• Stakeholder communication: advising decision makers, which the federal GS-14 data scientist posting reviewed for this page names as part of the job.
• Domain knowledge: knowing which effects are plausible and which data is trustworthy in your industry.
• Working with engineers: writing code others can run, and handing models over with documentation.

Portfolio artifacts that prove each level

Artifacts written for this page. Posting shares from 365 Data Science, Data Scientist Job Outlook 2026, 1,121 US Glassdoor postings, April 2026.

The data science tool stack

The data science stack runs from SQL and a programming language at the bottom to cloud and distributed data at the top. You need one working choice per layer and an understanding of what the layer is for, not every tool listed.

Tools as named in the skill progression behind this page and the 365 Data Science posting sample; the grouping by layer is editorial guidance. How often postings name each tool is shown only where measured, under skills. Tools are listed as text so that every one is readable.

How to become a Data Scientist: the step-by-step data scientist roadmap

To become a data scientist, work through five stages in order: math, statistics and code foundations; data wrangling and visualization; statistical modeling and experiments; machine learning; then production and leadership skills after your first role. Each stage ends with something you build. From zero, the first four stages total 39 to 63 weeks of part-time study, about 390 to 630 hours, which this page rounds to a suggested planning range of 9 to 15 months for a beginner. That assumes no prior coding or statistics beyond high school, about 10 hours of study a week, and the project at each stage.

Stage sequence from the researched data scientist skill progression behind this page. Durations assume about 10 hours of study a week and are suggested planning ranges, not measured outcomes. Stage 5 is a post-hire stage: its production and cloud skills carry a study range, while mentoring and data science strategy are built on the job and are given none. The before-you-start and done checks are editorial guidance built from the stage order and each stage's project, not published benchmarks.

Choose your track after the foundations

Once the foundations and core machine learning are in place, this page groups the next step into three common directions. The grouping is editorial guidance, not a measured taxonomy, and the tracks overlap. Choose by the deliverable you want: decisions from experiments, models that predict, or applications built on language models. This guide covers the analytics and business track's experiments and causal inference, and model building for the applied machine learning track; the related guides linked below go deeper on the rest.

Tracks grouped by this page as editorial guidance. Search answers captured October 1, 2026 show that readers ask about these directions; the evidence column cites the postings, GitLab's ladder, the 365 Data Science sample and BLS, accessed October 5, 2026.

The evidence behind each duration

Sources: provider course pages as listed, accessed October 5, 2026. "Reasoned from adjacent data" means the range combines two or more verifiable figures by inference. Stage 4 is the widest because deep learning depth varies; its low end counts two of the five deep learning courses. No fetched course states a duration for Power BI as a second tool, metrics and cohort analysis, or stakeholder presentation on its own, so those sit in the top of each range rather than in the arithmetic.

How long it takes from your starting point

The stage sum and the background estimates measure different things. The stage sum is the whole path from zero. The background estimates count only the stages each starting point still needs, so they are shorter for analysts, quantitative graduates and PhDs. All of them are suggested planning ranges for study time at about 10 hours a week, and none includes the job search. If you work full time, these are the ranges that apply to you, because they already assume part-time study of about 10 hours a week.

Estimates from the roadmap timeline research behind this page, built from official course durations, BLS and O*NET context, PhD fellowship and part-time bootcamp durations (Cornell Tech, Course Report) and one US training provider's hour totals (Noble Desktop), accessed October 5, 2026. The statistician, economist and actuary range is an editorial estimate and is labeled as one. The career-switcher row is an addition beyond the roadmap's four target backgrounds.

How many hours a week, and is one year enough?

This page's estimates assume about 10 hours of study a week, the pace most course pages state. The first four stages add up to about 390 to 630 hours, so you can scale the time to your own pace: more hours a week shortens the calendar time, and at 10 hours a week the planning range from zero is 9 to 15 months. One year of part-time study is therefore enough for many beginners and not for all. It covers the foundations, statistics and experiments, core machine learning and a first portfolio; specialization depth and production skills come later.

Starting in high school

The Bureau of Labor Statistics lists a bachelor's degree as the typical entry-level education for the occupation, and requirements vary by employer, so check postings in the field you want. A high school student can start the foundations early. Take the most advanced math your school offers, especially statistics and algebra, and computer science if it is available. Learn basic Python and finish one small project with a public dataset. A degree in mathematics, statistics, computer science or a related field, the fields the Bureau of Labor Statistics names for the occupation, is the typical next step. This page gives no duration for this path.

The three stages between learning and hiring

Written for this page as editorial guidance; no duration is given for these stages.

Qualifications and prerequisites for a Data Scientist

The hard prerequisites are statistics and a programming language, plus SQL. A bachelor's degree is what the Bureau of Labor Statistics lists as typical for entry, and postings name graduate degrees often.

Necessity ratings are this page's guidance. Sources: US Bureau of Labor Statistics, Occupational Outlook Handbook, Data Scientists, accessed October 5, 2026; 365 Data Science, Data Scientist Job Outlook 2026, 1,121 US Glassdoor postings, April 2026.

Do you need a degree to become a data scientist?

A degree is typical, and requirements vary by employer. The Bureau of Labor Statistics lists a bachelor's degree as the typical entry-level education, in mathematics, statistics, computer science or a related field, and notes that some employers require or prefer a master's or doctoral degree. Check individual postings for degree, equivalent-experience and specialist requirements. In 365 Data Science's sample of 1,121 US postings, 17.7% did not mention an education level. Those shares describe one sample, not the whole market.

Do you need a master's degree or PhD?

Not always, but postings name them often. In the same sample, 31.4% named a master's and 34.7% a PhD, while 16.2% named a bachelor's. Among O*NET survey respondents, 44% said a master's is required and 48% a bachelor's. The shares come from different sources measuring different things, so they are reported separately and never added together.

Which degree is best for becoming a data scientist?

The Bureau of Labor Statistics names mathematics, statistics and computer science, or a related field. This page reads related fields to include economics, physics and other quantitative degrees with statistics and programming. For comparing degree and certificate programs, see data science courses and programs.

Do you need coding to become a data scientist?

Yes. At entry level that means Python scripts with pandas and scikit-learn, SQL queries with joins and aggregations, and Git for version control. If you prefer analysis with little code, the Data Analyst career roadmap relies more on SQL, spreadsheets and BI tools.

Are there hard gates such as licenses or security clearance?

None were found. Data science requires no license. Security clearance appears on some defense-contractor listings; how often was not measured, so no share is given.

Before you start: a quick self-check

• Comfortable with high school algebra? If not, start with the math in stage 1.
• Know what a mean, median and standard deviation tell you? If not, start with stage 1 statistics.
• Comfortable working in spreadsheets? If not, practice before stage 2.
• Willing to write code most days? If not, consider the data analyst route first.

How to switch careers into data science

The shortest estimated paths into data science start from roles that already hold part of the roadmap: PhDs and researchers, statisticians, analysts and quantitative graduates. The longest starts from no technical background, and that path also runs into the degree expectation.

Timelines from the roadmap timeline research behind this page, at about 10 hours a week, accessed October 5, 2026. The statistician, economist and actuary range is an editorial estimate. Where a transition is not measured, no range is given.

From data analyst to data scientist

An analyst already holds the data handling and much of the visualization the role needs. The work is math for machine learning, experiment design and machine learning itself, an estimated 6 to 10 months. Analyst skills in depth are on the Data Analyst career roadmap, and how the two roles differ is under role comparison.

From software engineer to data scientist

Programming, Git and deployment transfer directly. What the roadmap adds is statistical reasoning: experiment design, inference and honest model evaluation, which are stages 3 and 4. This page gives no estimate for the move. If you would rather build and run models in production than design experiments, compare the Machine Learning Engineer career roadmap.

From a math, finance or economics background

This page assumes graduates and professionals from these fields already hold the statistics; the gap is Python, SQL and machine learning. A finance or economics background is also domain knowledge for the analytics and business track, where causal questions about pricing, policy and risk arise.

With no technical background, no degree or no experience

Plan for the longest path: a suggested planning range of 9 to 15 months of part-time study to reach entry-level skills, not including the job search. That time buys the skills, not the credential. The Bureau of Labor Statistics lists a bachelor's degree as the typical entry-level education and 44% of O*NET respondents say a master's is required, so this page's guidance without a degree is to start in an analyst role, build a portfolio of complete projects and move toward modeling work from there. Start with stage 1 math and Python, and a first exploratory project on a public dataset.

Portfolio projects that show you can work as a Data Scientist

A data scientist portfolio should show decisions supported by evidence, not only models that train. Each project should start from a business question and end with a recommendation, with the method and its validation in between.

Archetypes written for this page. Posting skills from 365 Data Science, Data Scientist Job Outlook 2026, 1,121 US Glassdoor postings, April 2026.

How many projects do you need?

There is no measured employer requirement for a project count. This page's editorial guidance is depth over count: two or three complete projects that use different methods, each with a question, a validated method, documented code and a recommendation. That is a practice plan, not a hiring statistic.

Does Kaggle experience help?

Kaggle is a good source of datasets and practice. A competition leaderboard shows modeling skill on a problem someone else framed; a portfolio project also shows that you can frame the question and explain the result.

What every project write-up should state

• The business question and the decision it informs
• Where the data came from and how it was cleaned
• The method, and why it fits the question
• How the result was validated, including the baseline
• What the result means, its uncertainty and its limits
• How to run the code

Suggested portfolio project: customer retention analysis and intervention

One worked example of a project that covers stages 2 to 4. It is an editorial suggestion, not a measured hiring requirement, and it has no expected result: what you find depends on the data.

• Business question: which customers are likely to stop using a product, and which intervention is worth testing?
• Inputs: a public customer or activity dataset, or a synthetic one labeled as synthetic, with a clear time cutoff between the history you learn from and the period you predict.
• The work: define what counts as leaving, prevent leakage of information from after the cutoff, fit a simple baseline before any complex model, split the data by time where that matches how the model would be used, and inspect errors for customer subgroups.
• Deliverables: a reproducible notebook or package, a data dictionary, an evaluation report and a one-page decision memo.
• Failure cases to check: future information leaking into training, an imbalanced outcome that makes accuracy misleading, a decision threshold that shifts from sample to sample, and a correlation read as a cause.
• Evidence to show: the comparison with the baseline, the uncertainty around your estimates where it applies, examples of errors, and clear limits on any causal claim.
• Extension: propose an A/B test of one intervention as a separate plan. A model that predicts who leaves does not show that an intervention would make them stay.

What rigorous, production-grade data science means

A notebook result works once. Rigorous data science holds up on new data, can be reproduced and is used only for the decision it was built for. These competencies are hard to learn from tutorials, because a tutorial dataset is already clean and its answer is already known.

Written for this page.

Where MLOps fits

A data scientist needs to hand a model over in a state engineers can run: versioned code, documented inputs and outputs, and agreed checks for drift. Deploying, serving and monitoring models at production scale is MLOps, which this roadmap places in stage 5, after your first role. It is a typical emphasis of machine learning engineering, and in some teams data scientists own it too; see the Machine Learning Engineer career roadmap for MLOps in depth.

How to prepare for a Data Scientist interview

Prepare for the skills postings ask for, and for explaining your own projects in depth. The learning and interview suggestions on this page are an editorial preparation framework built from posting requirements, not a description of how any employer's hiring process runs.

Framework written for this page from the posting skills in the 365 Data Science, Data Scientist Job Outlook 2026 sample and the skills in the roadmap's stages 1 to 4.

Do certificates matter for becoming a Data Scientist?

No certification is required to become a data scientist, and a certificate does not show what a portfolio of validated work shows. A structured program gives you a sequence, deadlines and feedback; what you still have to build yourself is a public portfolio and practice explaining it.

Do certificates matter, or only projects?

This page makes no ranking, because how employers weigh certificates against projects was not measured. A certificate shows completed coursework; a portfolio shows work a reviewer can run and question, and many postings also name a degree. As editorial guidance, the practical order is to build first and certify alongside.

Degree, bootcamp or self-paced roadmap?

Sources: a US online master's program page and Course Report's bootcamp guide, as recorded in the roadmap timeline research, accessed October 5, 2026; self-paced estimate from this page's roadmap. Costs are not compared, because no cost figure in this research is current and verified.

Vendor certifications

Cloud vendor credentials attest to competence on one platform and fit after the roadmap, not at its start. Amazon Web Services, for example, recommends at least 1 year of hands-on experience with Amazon SageMaker and related services before its Certified Machine Learning Engineer - Associate exam.

What the Simplilearn programs on this page are, and are not

The three Simplilearn programs on this page are data and business analytics programs, not data science programs, and they cover the foundation stages of the roadmap only. This page makes no claim about placement rates, salary increases or job guarantees for any program. For programs that go further into machine learning, compare data science courses.

The biggest mistakes people make learning data science

Key data science terms explained

Data science

Data science is the practice of using statistics, programming and domain knowledge to turn raw data into models, forecasts and recommendations that inform decisions.

Hypothesis testing

Hypothesis testing is a statistical procedure for deciding whether an observed difference or relationship in data is larger than chance alone would plausibly produce.

P-value

A p-value is the probability of seeing a result at least as extreme as the one observed if there were no real effect. A small p-value is evidence against the no-effect assumption, not a measure of how large or important an effect is.

Confidence interval

A confidence interval is a range computed from data that, under repeated sampling, would contain the true value a stated share of the time. It reports the uncertainty around an estimate.

A/B test

An A/B test is a randomized controlled experiment in which users or units are randomly assigned to a control version and a variant, so that a difference in a chosen metric can be attributed to the change.

Statistical power

Statistical power is the probability that an experiment detects an effect of a given size when an effect of that size exists. It is set mostly by sample size and is calculated before a test starts.

Causal inference

Causal inference is the set of methods for estimating the effect of one variable on another, including from observational data where no randomized experiment was run.

Confounder

A confounder is a variable that influences both a cause and an outcome, creating an association between them that is not causal unless it is accounted for.

Regression

Regression is a family of models that estimate how an outcome changes with one or more inputs. Linear regression predicts a number; logistic regression predicts the probability of a category.

Classification

Classification is a supervised learning task in which a model assigns each record to a category, such as whether a customer will churn.

Cross-validation

Cross-validation is a way of estimating how a model will perform on new data by repeatedly training it on part of the data and testing it on the held-out remainder.

Overfitting

Overfitting is when a model learns noise in its training data and so performs well on that data but poorly on new data.

Data leakage

Data leakage is when information that would not be available at prediction time, such as a future value or the target itself, slips into model training and makes performance look better than it will be in use.

Precision and recall

Precision is the share of a model's positive predictions that are correct; recall is the share of actual positives the model finds. Raising one usually lowers the other.

Bias-variance tradeoff

The bias-variance tradeoff is the balance between a model too simple to capture a pattern and a model so flexible that it fits noise. Validation on held-out data is how the balance is judged.

Exploratory data analysis (EDA)

Exploratory data analysis is the first look at a dataset, using summaries and charts to understand its structure, quality and patterns before any model is built.

Feature engineering

Feature engineering is creating and selecting the input variables a model learns from, such as ratios, counts or encoded categories, to make useful patterns easier to learn.

Time-series forecasting

Time-series forecasting is predicting future values of a measure recorded over time, such as weekly demand, using its own history and related signals.

MLOps

MLOps is the set of practices for deploying, monitoring and updating machine learning models in production, so that a model keeps working as data changes.

Data Scientist: frequently asked questions

How we compiled this data

Every figure on this page comes from a named, dated source, and each salary figure is labeled with what it measures. Sources that disagree are shown side by side, not averaged, and cuts with no reliable source are left out rather than estimated.

What was collected, and how

How this page was produced

This page was researched and drafted with the help of AI tools. Every figure on it traces to a named source recorded in the page's research files, and each source is listed with the date it was checked in the table above. Where the evidence was missing or could not be read, the page says so rather than estimating. The reviewers named on this page checked it after it was built, including its title, description, image descriptions and structured data as well as the body text.

Limitations

• The government code is broader than the title. SOC 15-2051 also covers business intelligence analysts and clinical data managers, so every BLS figure describes the occupation, not only people titled Data Scientist.
• Salary sources measure different things. Wages, advertised base pay, self-reported base and total pay, and projected benchmarks are shown separately and never combined.
• Ladders do not line up. Salary.com, Indeed, Payscale and Robert Half band pay by job level, title, years and qualitative tier respectively, and each is shown as published.
• Some sources could not be read directly. ZipRecruiter and the BLS occupation-level wage page could not be retrieved, Glassdoor was read through a country redirect and needs recapture from a US session, and Payscale publishes no dollar figure for its later career bands.
• No skill premium was found. None of the sources reviewed prices any skill, certification or degree for data scientists.
• Unmeasured topics get editorial guidance, not figures. The sources reviewed do not measure interview structure, promotion criteria, how employers weigh certificates or Kaggle results, project counts, the job search, or the career transitions given no range. Guidance on those topics, including the specialization tracks and the suggested portfolio project, is labeled as editorial.
• The posting analysis is one sample. Skill and education shares come from 1,121 postings analyzed by one publisher, with no stated collection window, and are not market-wide rates.
• Listing counts are snapshots. LinkedIn and Indeed counts are single-day keyword matches, not unique open positions.
• No experience distribution, job-title distribution or remote-work share is published. The research found no usable 2026 measure for any of them.
• The demand counterweight is a year old. Indeed Hiring Lab's sector index is as of the end of October 2025 and covers data and analytics broadly; no 2026 sector figure was found.
• Timelines are suggested planning ranges. Durations are built from course lengths at about 10 hours a week, from no prior coding, with the project at each stage; the statistician, economist and actuary range is an editorial estimate, and no range is given for software engineers, high school students or the job search.
• No cross-role salary comparison is made. Each role page draws pay from its own sources and dates.
• The programs are analytics programs. None of the Simplilearn programs on this page is a data science program, and none covers the full roadmap.

Data left out, and why

• Secondary re-reports of Glassdoor figures, including figures by years of experience, because their pay type is not stated and they conflict with every primary entry-level figure.
• ZipRecruiter, whose page returned a bot challenge and was not retrieved.
• Levels.fyi figures by metro and by company, and a salary platform's figures by industry, because the per-group samples are undisclosed.
• Company pay from Indeed, where four of the five top-paying companies listed are staffing firms advertising for undisclosed clients.
• Pay by company type, which the sources reviewed do not group or disclose.
• Executive titles and internship pay, which are different jobs or pay periods.
• A table mapping proficiency levels to pay, because none of the sources reviewed links skill level to salary.
• Salary figures shown on Simplilearn's own course and category pages, which are unattributed or do not match current sources.

Reviewers

This page was reviewed by seven Simplilearn data science, analytics, machine learning and data platform authors and industry experts.

Simplilearn authors and industry experts who reviewed this page

Reviewer	Role	Relevant expertise	Profiles
Aditya Kumar	Senior Business Analyst	Data discovery and experimentation; writes on data scientist job descriptions, the data science syllabus and scikit-learn regression models	Simplilearn profile LinkedIn
Avijeet Biswal	Senior Research Analyst	Machine learning and deep learning with Python, Tableau, Hadoop and Spark; writes on the chi-square test, statistics for machine learning and linear regression	Simplilearn profile LinkedIn
Kshitij Choughule	Senior Product Analyst	SQL, dashboards and KPIs; writes on statistics for data science, data science interview questions and the CRISP-DM methodology	Simplilearn profile LinkedIn
Shruti M	Senior Research Analyst	Writes on probability distributions, random forests in R and the Hadoop and Hive data stack	Simplilearn profile LinkedIn
Tony Vigna	Data Science and Systems Engineering Leader	Data platforms, machine learning systems and data warehousing across trading, CRM and quantitative risk	Simplilearn profile LinkedIn
Pulkit Jain	Product Manager	Cloud and DevOps background; writes on AWS MLOps services, Azure Machine Learning, BigQuery and Redshift	Simplilearn profile LinkedIn
Priyadharshini	Educational Specialist	Writes on machine learning basics, R vs Python, data mining vs statistics and making a successful career change	Simplilearn profile LinkedIn

Full profiles at Simplilearn Authors and Industry Experts.

About the author

Jitendra Kumar is the Chief Technology Officer at Simplilearn, leading enterprise AI readiness and generative AI strategy. An IIT Kanpur alumnus and technology entrepreneur, he specializes in building complex AI systems integrated with scalable edtech solutions.

Full profile and published articles: Jitendra Kumar on Simplilearn · LinkedIn