Last updated: November 2026. Written by Josh Hutcheson, OnlineCourseing editor. See our review methodology.
QUICK ANSWER
Short answer: most data scientists work daily in Python, SQL, Jupyter and Git, then add one dashboard tool. The rest of this list matters when a job or dataset calls for it.
- Learn first: SQL and Python, with pandas in a Jupyter notebook
- Machine learning: scikit-learn for tabular data; PyTorch for deep learning
- Dashboards: Tableau or Power BI, matched to your employer
- Big data: Spark, usually through Databricks or a cloud service
- Cost: the core stack is free; you pay for commercial platforms, sharing and compute
Data science tools are the software used to collect, clean, analyze, model and present data. No one needs all 20 below. We grouped them by the job they do, so you can see which ones belong in your first month and which you can ignore until a project demands them. Versions and prices were read from each project or vendor in November 2026; where a vendor blocked our read or sells by quote, we say so instead of guessing.
This page compares tools. To learn them with structure, see our best data science courses. For the language question on its own, read our guide to the best programming languages for data science.
Data Science Tools at a Glance
Before you sign up for another data science course, read this.
I've taken DataCamp, Dataquest, Coursera ML, and the Udacity nanodegrees. Get my Tuesday picks — plus reader-only codes when they drop.
No spam. Unsubscribe anytime.
| Tool | Job | Cost | Priority |
|---|---|---|---|
| Python, SQL, Jupyter, Git | Core daily work | Free (database hosting may cost) | Learn first |
| pandas, Excel | Cleaning and quick analysis | pandas free; Excel for the web free, desktop paid | Learn first |
| scikit-learn | Classical machine learning | Free | Learn second |
| Tableau, Power BI, Matplotlib | Charts and dashboards | Free tiers; Power BI Pro $14 a user a month, paid yearly | Learn second |
| R | Statistics and research | Free | If your field uses it |
| PyTorch, TensorFlow | Deep learning | Free; GPU compute may cost | When a project needs it |
| Spark, Databricks, Hadoop | Large-scale data | Open source or usage-based; Databricks Free Edition | When data outgrows one machine |
| MATLAB, SAS Viya, BigML | Specialized or commercial platforms | Paid; BigML from $30 a month | If your employer uses it |
Languages and Query Tools
1. Python
Python is the default language of data science, with most of the libraries below built around it. Python.org listed 3.15.2 as the current release on 5 November 2026. It is free and open source, and its syntax is approachable for people who have never programmed.
Use it when: you want one language for cleaning, analysis, modeling and automation. Library support can lag a brand-new Python release, so check compatibility before upgrading a project. Our Udemy Python course picks are one way to learn it.
2. R
R is a free language built for statistics and graphics, and version 4.6.2 was released on 29 October 2026. It remains strong in academia, biostatistics and research-heavy teams, where its statistical packages and publication-quality charts are hard to beat.
Use it when: your field already uses it, or your work is mainly statistical analysis rather than software. Otherwise start with Python, because it is more widely used for production machine learning.
3. SQL
SQL is the language for querying relational databases, and it is how most company data reaches an analyst. It is a standard rather than a single product, with PostgreSQL, MySQL, Snowflake and BigQuery each adding their own dialect. DuckDB (2.0.0 on PyPI) is a convenient way to practice, since it runs SQL on local files with no server.
Use it when: always. If you learn one tool besides Python, make it this one. See our best SQL courses.
Notebooks, Editors and Version Control
4. Jupyter
Jupyter notebooks mix code, charts and notes in one document, which suits exploration and explaining results. JupyterLab was at 4.6.5 on PyPI on 5 November 2026. It is free, and hosted versions come with many cloud platforms.
Use it when: you are exploring data or sharing an analysis. Move repeated work into scripts, because notebooks can hide errors when cells run out of order.
5. Visual Studio Code
Visual Studio Code is a widely used open-source code editor. Its extensions add Python support and Jupyter notebooks, so you can work on scripts and notebooks in one place.
Use it when: your projects grow beyond a few notebooks and you need debugging, file navigation and Git in one window.
6. Git
Git is free, open-source version control, at 2.56.0 on its site on 5 November 2026. It records every change to your code, so you can undo mistakes and work with a team. GitHub and GitLab host Git projects online, and a public GitHub profile doubles as a portfolio.
Use it when: you have any project worth keeping. Learn commit, branch and merge first; that covers most daily use.
Data Cleaning and Analysis
7. pandas
pandas is the standard Python library for working with tables of data: loading files, filtering, joining and summarizing. It reached 3.1.1 on PyPI on 4 November 2026 and is built on NumPy (2.5.4), which handles fast numerical arrays. Tutorials written for older versions may show outdated behavior, so check the version they use.
Use it when: your data fits in memory on one machine. That covers more projects than beginners expect.
8. Microsoft Excel
Microsoft Excel is still where many business datasets start and finish. Formulas, filters and pivot tables make quick checks fast, and Excel for the web is free with a Microsoft account. It struggles with very large datasets and repeatable pipelines. See our best Excel courses if your skills need work.
Use it when: you need a fast check or must hand results to non-technical colleagues.
Machine Learning and Deep Learning
9. scikit-learn
scikit-learn (1.9.1 on PyPI) is the standard Python library for classical machine learning: regression, classification, clustering and model evaluation. Its consistent interface makes it easy to compare models, and it is usually the right first choice for spreadsheet-style data.
Use it when: you have tabular data and need a reliable baseline before trying anything more complex. Our Udemy machine learning course picks teach it hands-on.
10. PyTorch
PyTorch (2.15.0 on PyPI, released 28 October 2026) is an open-source deep learning framework, and it is a common starting point for learners and researchers working with neural networks. Training large models usually requires a GPU, which you may need to rent.
Use it when: you work with images, text or other unstructured data where deep learning is justified.
11. TensorFlow
TensorFlow (2.21.0 on PyPI, last released March 2026) is Google’s open-source machine learning framework. It has a mature set of deployment tools, and many existing production systems use it. Our TensorFlow tutorials guide lists ways to learn it.
Use it when: your team or the system you maintain already runs on it. For a new learner with no constraint, compare it with PyTorch first.
Visualization and Dashboards
12. Matplotlib
Matplotlib (3.11.2 on PyPI) is the foundational Python plotting library, used for bar charts, histograms and scatter plots. Its defaults take effort to polish, and libraries such as seaborn build on it to make common statistical charts easier.
Use it when: you need charts inside a notebook or a report generated by code.
13. Tableau
Tableau is a drag-and-drop tool for interactive dashboards, and it needs no programming. Tableau Public is a free way to build and publish visualizations in the browser, and a free Tableau Desktop Public Edition lets you work offline, use private data and save locally. Paid Tableau products add company-wide sharing and governance; Tableau’s main site blocked our read this month, so we quote no paid price. Our Tableau courses guide covers how to learn it.
Use it when: your target employers list it, or you want a public portfolio of dashboards.
14. Power BI
Power BI is Microsoft’s dashboard tool. Microsoft lists a free account for building reports, Power BI Pro at $14 a user a month paid yearly for sharing them, and Premium Per User at $24. It fits well in companies already using Microsoft 365. See our best Power BI courses.
Use it when: your company runs on Microsoft. If it does not, compare Tableau before committing.
Big Data Platforms
15. Apache Spark
Apache Spark is an open-source engine for processing data across many machines, in both batches and streams. Version 4.3.0 was released on 30 October 2026. It has Python, SQL, Scala, Java and R interfaces, and PySpark lets pandas users move to larger data. Our Apache Spark tutorials guide lists beginner routes.
Use it when: your data no longer fits on one machine. Do not start here if a laptop can handle the job.
16. Databricks
Databricks is a commercial cloud platform for data engineering, analytics and AI, with Spark at its base. Its Free Edition replaced Community Edition, which was retired in 2025, and offers serverless compute under daily quotas for personal learning. Databricks says it comes with no guaranteed reliability or support, and it shuts down compute when you exceed quota.
Use it when: you want to practice Spark and a real cloud workspace without paying.
17. Apache Hadoop
Apache Hadoop stores and processes very large datasets across clusters of ordinary computers. It is still maintained, and 3.5.0 is the first stable release of its 3.5 line, but it is complex to run. New projects usually choose Spark with cloud storage instead.
Use it when: you are maintaining an existing Hadoop system, or a job specifically asks for it.
Commercial and Specialized Platforms
18. MATLAB
MATLAB is a commercial numerical computing environment used heavily in engineering, signal processing and control systems. Universities often provide licenses. MathWorks’ pricing page blocked our read this month, so we quote no price. Check it directly, along with any campus license.
Use it when: your engineering team or degree program already uses it. For general data science, Python covers the same ground for free.
19. SAS Viya
SAS Viya is SAS’s cloud analytics and AI platform, used in large regulated organizations such as banks, insurers and healthcare companies. SAS sells it by quote, offering a free trial and a pricing request rather than a public price.
Use it when: your employer or target industry runs SAS. It is rarely a sensible self-funded first choice.
20. BigML
BigML is a cloud machine learning platform with a point-and-click interface and an API. It lists a three-day free trial and a Standard Prime plan at $30 a month with 64 MB per task, with higher tiers for larger data. That dataset limit is low compared with what the open-source tools can handle on a laptop.
Use it when: you want to try machine learning concepts on small data without writing code.
A Practical Stack by Role
| Role | Start with | Add later |
|---|---|---|
| Data analyst | SQL, Excel, Tableau or Power BI | Python with pandas |
| Data scientist | Python, SQL, Jupyter, pandas, scikit-learn, Git | A dashboard tool, Spark |
| Machine learning engineer | Python, Git, PyTorch | TensorFlow, Spark, cloud deployment |
| Researcher or biostatistician | R or Python, Git | SAS if your organization requires it |
| Engineering analyst | MATLAB or Python | Git, scikit-learn |
Data Science Tools and Techniques: How They Fit Together
Tools are easier to choose once you map them to the steps of a project. A typical project moves through five stages:
- Collect: query a database with SQL or load files into Python.
- Clean: fix types, duplicates and missing values with pandas or SQL.
- Explore: summarize and chart the data in a Jupyter notebook with Matplotlib.
- Model: fit and evaluate a baseline with scikit-learn, then consider PyTorch if the problem needs it.
- Communicate: present the result in Tableau, Power BI or a short written report, and save the code in Git.
One small project through those five stages teaches more than a tour of 20 tools. Our list of data analytics projects has starting points, and time series courses help if your data is sequential.
How to Choose Data Science Tools
- Read 20 job listings first. Count which tools appear for the role you want, then learn those before anything else on this page.
- Match the tool to the size of the data. Do not reach for Spark if pandas or DuckDB can do the job on your laptop.
- Prefer free tools while learning. Paid platforms rarely teach skills you cannot get from the open-source stack.
- Check what a free tier excludes. Tableau Public publishes what you save to it, and Databricks Free Edition has quotas and no service guarantees.
Data Warehouses and Pipelines
Company data usually lives somewhere other than your laptop. Snowflake, Google BigQuery and Amazon Redshift are cloud data warehouses you query with SQL, and most charge by usage. We did not count them in the 20 because you normally inherit whichever one your employer chose, and your SQL carries across with small dialect changes. Open-source dbt (1.12.6 on PyPI) is worth knowing once you maintain tables other people rely on, since it brings version control and tests to SQL transformations. Airflow and similar schedulers run those jobs on a timetable, and that work tends to belong to data engineers.
Tools We Left Out, and Why
Longer lists pad their counts with products few learners will meet. These are real tools with real users, but they do not belong in a first toolkit:
- KNIME, RapidMiner and Alteryx: visual, drag-and-drop workflow tools. Learn them if your employer uses them. Python skills transfer to more jobs. See our best Alteryx courses if yours does.
- IBM SPSS: still common in social science and survey research departments, usually through a university or employer license.
- Julia: a fast language with a smaller job market and library ecosystem than Python.
- Keras: a high-level deep learning interface you will meet through TensorFlow or PyTorch work, so it does not need separate planning.
- D3.js: a JavaScript library for custom web charts, better suited to visualization developers than to most data scientists.
- Weka: a teaching-friendly machine learning workbench that some university courses still use.
What Data Science Tools Actually Cost
Learning the core stack costs nothing. The bills arrive in three places. Sharing is the first: Power BI needs Pro at $14 a user a month, paid yearly, to publish and share reports, and paid Tableau plans serve the same company need. Compute is the second: deep learning on a rented GPU and large Spark jobs can cost far more than any software license, so stop cloud resources when you finish. Commercial platforms are the third: MATLAB and SAS are usually bought by institutions, and SAS quotes prices on request. For a self-funded learner, the realistic budget is a course subscription, because the tools themselves are free.
Data Science Tools vs Data Analytics Tools
The two lists overlap on SQL, Python, Excel and dashboards. The difference is emphasis. Analysts spend more time in SQL, spreadsheets and Tableau or Power BI, answering what happened and why. Data scientists spend more time in Python, statistics and scikit-learn, building models that predict what happens next. Many job titles blur this line, so read the listing before deciding which set to learn. Our data analytics courses guide covers the analyst route.
Where AI Coding Assistants Fit
AI assistants, including GitHub Copilot and general chatbots, now write much routine pandas and SQL. They speed up work you can check, but they can also produce a plausible query that joins tables incorrectly, or a chart built on the wrong column. Treat them as a fast junior colleague: useful for first drafts and for explaining errors, and in need of review. Learn the fundamentals first, because you cannot catch a mistake in code you cannot read. Check your employer’s policy before pasting company data into any outside service.
Mistakes to Avoid When Picking Tools
- Collecting tools instead of finishing projects. Employers would rather see one complete analysis than a resume listing 20 products you opened once.
- Starting with deep learning. Many business problems on tabular data are handled well by scikit-learn, at far lower cost and with results that are easier to explain.
- Learning a tool from an old tutorial. pandas, Spark and Databricks have all changed recently, so check the version a tutorial uses, and prefer the official documentation when the two disagree.
- Ignoring SQL. Python skills will not help if you cannot get the data out of the company database in the first place.
How to Learn These Tools
A structured course saves time if you are starting from zero. The IBM Data Science Professional Certificate on Coursera is a 12-course series rated 4.6 from 151,401 course reviews, and Coursera estimates four months at 10 hours a week. If you have never programmed, Python for Everybody is a gentler first step, with more than 2 million enrollments. For analyst roles, the Google Data Analytics Certificate covers spreadsheets, SQL, Python and Tableau.
Coursera subscriptions cover those programs, and Coursera Plus costs $239 a year (normally $399). Our best data science courses guide compares these with other platforms, and how to become a data scientist without a degree sets out the career path.
Frequently Asked Questions
What are data science tools?
Data science tools are the software used to collect, clean, analyze, model and present data. A typical set includes a language such as Python or R, SQL for querying databases, a notebook such as Jupyter, libraries such as pandas and scikit-learn, and a dashboard tool such as Tableau or Power BI.
What are the most important data science tools?
Python with pandas and scikit-learn, SQL, Jupyter and Git. Add one dashboard tool, Tableau or Power BI, based on what your target employers use. Those cover the daily work in most entry-level roles.
Do I need to know how to code to be a data scientist?
Yes, for most roles. Python and SQL appear in most data science work. Some analyst roles rely more on Tableau, Power BI and Excel, but coding widens your options considerably.
What are the best free data science tools?
Most of the core stack is free and open source: Python, R, Jupyter, pandas, scikit-learn, PyTorch, TensorFlow, Matplotlib, Spark and Git. Tableau Public, the Power BI free account, Excel for the web and Databricks Free Edition also cost nothing, each with limits.
Which data science tools should a beginner learn first?
Start with SQL and Python, using pandas in a Jupyter notebook. Learn Git once you have a project worth saving. Leave deep learning frameworks and Spark until a project actually needs them.
Is Excel a data science tool?
For small datasets, yes. It is useful for quick checks and for sharing results with non-technical colleagues. It is a weak fit for large data, repeatable pipelines and machine learning, which is why data scientists pair it with Python and SQL.
Is Hadoop still used in data science?
Yes, but mainly in existing big data installations. Apache still ships releases, including 3.5.0, but new data science projects usually start on Spark, a cloud warehouse or a platform such as Databricks.
Is data science worth learning in 2026?
It can be, if you enjoy statistics and messy problems. BLS reported a May 2025 median wage of $120,230 for data scientists, but entry-level competition is strong, and many people start as analysts. The tools on this page also transfer to analyst and engineering roles, which lowers the risk of learning them.
What is the difference between data science and software engineering tools?
They overlap on Python, SQL, Git and code editors. Data science adds notebooks, statistics and machine learning libraries, and dashboards. Software engineering adds application frameworks, testing, deployment and infrastructure tools. BLS reported May 2025 median wages of $120,230 for data scientists and $135,980 for software developers.
