8 Data Analyst Portfolio Projects That Get Interviews

Updated October 14, 2026 · 19 min read

Quick answer: The data analyst portfolio projects that get interviews each start from a question written as a sentence, use a real, messy public dataset that a stranger can download, and prove a different skill. The eight below cover cleaning, multi-table SQL, customer segmentation, pricing on a map, survey statistics, a dashboard, time-and-place patterns and an analysis of your own data. Build three to five, chosen by the job you want, and give each one a short write-up with a finding, its limits and a link to the files. To lay a finished project out as a page, follow the data analyst project page template; if you have no analyst job on your resume yet, start with the no-experience guide.

Image credit: Yan Krukau / Pexels

How this guide was put together: every dataset below was opened on its publisher's own page on October 4, 2026, and the facts quoted about it (file names, date ranges, licences, fields) come from those pages. The finding sentences are templates showing the shape of a good write-up; they are not results from these datasets, so the numbers are yours to find. The portfolio statistics come from Seera's data portfolio benchmarks, and the three example portfolios come from Seera's public data showcase, each re-checked against its live page. Sources are listed at the end. One limit: no published study ties a particular portfolio project to interview rates, so "gets interviews" here means passing the checks a reviewer can make from the page, not a measured outcome.

Key Takeaways

Each project proves one thing. Together they cover the skills a junior analyst job description asks for, and none of them depends on a tutorial dataset every other applicant has also used.

Table of Contents

• What makes a data analyst portfolio project get interviews?
• Which data analyst portfolio projects should you build?
• How many projects do you need, and which ones?
• Which projects suit beginners?
• Which SQL projects belong in a portfolio?
• How do you write up a project so it earns an interview?
• How do you put projects on GitHub?
• Three real portfolios that show the format
• Frequently asked questions
• Useful sources and further reading

What Makes a Data Analyst Portfolio Project Get Interviews?

A project gets interviews when a stranger can see, in under two minutes, what question you asked, what data you used, what you found and where your working is. The tool, the chart type and how famous the dataset is all matter less than those four things.

1. It starts from a question, written as a sentence. "Which complaint types take longest to close?" gives the reader something to agree or disagree with. "NYC 311 analysis" gives them nothing.
2. The data is real and arrives messy. Dates in two formats, blank cells, categories spelled five ways. Cleaning it is the part of the job that interviewers can only check by seeing it done.
3. It ends in a finding and a limit. One sentence on the result, with a number and its scope, and one on what it cannot tell you.
4. The files open for a stranger. A query file, a workbook or a public dashboard, linked from the project and not behind a login.

The fourth point is where most portfolios fall short. Seera's benchmark of 56 data portfolios published on the platform (grouped under the Data Scientist title, and skewed towards business intelligence) shows what people list and what they link:

Two readings follow. SQL and one BI tool are the safe pair to show, because most of these portfolios list them. And the gap between the last two rows is the opening: two thirds of these portfolios point to a profile, but only one in five attaches a link to an individual project, so most readers are asked to take the work on trust. A project with its files one click away is already unusual. The benchmark is anonymised and free to cite under CC BY 4.0; the full breakdown is on the data portfolio benchmarks page.

Which Data Analyst Portfolio Projects Should You Build?

Build the three to five that match the roles you are applying for from the eight below. Each is a self-contained spec: a question, a dataset you can open today, the skills it proves, what to publish and the question an interviewer is likely to ask about it.

The project ideas in our portfolio template guide use different datasets and are a good second list once you have finished one from this page.

1. NYC 311: how long do complaints take to close?

• Question: Which complaint types take longest to close, and does that change by borough?
• Dataset: 311 Service Requests from 2020 to Present on NYC Open Data. The portal updates it daily. Each row has created and closed timestamps, the responsible agency, a complaint type and location fields.
• Skills it proves: date arithmetic, handling requests with no closing date, grouping by category, and using a median where a few very slow cases would distort an average.
• Tools: SQL for the work, Excel or Python for the chart.
• Publish: the query file and one table of median days to close by complaint type, with a note on how many requests were still open when you downloaded the data.
• Interview question it answers: "How did you treat records that were never closed?"

Finding template: "Among [N] requests created between [date] and [date], [complaint type] took a median of [X] days to close in [borough], against [Y] days in [borough]; [Z]% were still open at download."

2. Olist: do late deliveries lower customer review scores?

• Question: When an order arrives after its estimated delivery date, do customers rate it lower?
• Dataset: the Brazilian E-Commerce Public Dataset by Olist on Kaggle: about 100,000 orders from 2016 to 2018 split across nine CSV files, including orders, order items, customers, sellers, products, payments and reviews. A free Kaggle account is needed to download it.
• Skills it proves: joining three or more tables without duplicating rows, calculating a delay from two date columns, a common table expression, and ranking sellers or categories with a window function.
• Tools: SQL in SQLite, PostgreSQL or BigQuery.
• Publish: one commented .sql file and a small diagram of which tables you joined and on which keys.
• Interview question it answers: "Walk me through this join. Why an inner join and not a left join?"

Finding template: "Across [N] delivered orders, those that arrived late averaged [X] review stars against [Y] for on-time orders, and the gap held in [N] of the [M] largest product categories."

3. Online Retail II: which customers are worth keeping?

• Question: Which groups of customers drive most of the revenue, and what do they have in common?
• Dataset: Online Retail II from the UCI Machine Learning Repository: 1,067,371 transactions from a UK online gift-ware retailer between 1 December 2009 and 9 December 2011, many of them to wholesale customers. Invoice numbers starting with "c" are cancellations, some rows have no customer ID, and the licence is CC BY 4.0.
• Skills it proves: removing cancellations and unidentified customers on purpose, calculating recency, frequency and monetary value for each customer (RFM), and turning scores into named segments.
• Tools: Python with pandas, or Excel with Power Query and pivot tables.
• Publish: segment definitions in plain words and one table with each segment's size and share of revenue.
• Interview question it answers: "How did you decide where a good customer starts?"

Finding template: "[N]% of customers, the [segment name] group, produced [X]% of revenue; they ordered every [Y] days on average and [Z]% of them were wholesale accounts."

4. Inside Airbnb: what explains price differences between neighbourhoods?

• Question: In one city, how much of the price gap between neighbourhoods remains once you compare the same type of listing?
• Dataset: Inside Airbnb publishes, for each city, detailed listings, a calendar, reviews, a neighbourhood list and a neighbourhood GeoJSON file, under a CC BY 4.0 licence. Each city shows the date it was collected.
• Skills it proves: spotting and explaining price outliers, comparing like with like by room type, joining a neighbourhood file to build a map, and stating which collection date you used.
• Tools: Tableau or Power BI for the map, Python for the cleaning.
• Publish: a public map dashboard and a short list of the rows you excluded and why.
• Interview question it answers: "What did you exclude, and how would the answer change without that decision?"

Finding template: "For entire-home listings in [city] on [collection date], the median nightly price in [neighbourhood] was [X]% above [neighbourhood]; the gap narrowed to [Y]% after excluding listings above [price]."

5. Stack Overflow Developer Survey: how much does pay rise with experience?

• Question: How does reported compensation change with years of experience, and how far can that comparison be trusted?
• Dataset: the annual Stack Overflow Developer Survey, which publishes anonymised responses for download. The survey data is licensed under ODbL 1.0, so read the attribution and share-alike terms before you publish a derived dataset.
• Skills it proves: reasoning about self-selected respondents, choosing medians over means when a few salaries are extreme, splitting multi-answer columns, and putting caveats next to the chart and not in a footnote.
• Tools: Python or R for the analysis, SQL for the filtering.
• Publish: a notebook and a short "what this does not show" paragraph.
• Interview question it answers: "Does this sample represent all developers?"

Finding template: "Respondents with [X] to [Y] years of experience reported a median of [Z] in [country]; the sample is self-selected, so the figures describe survey respondents and not all developers."

6. College Scorecard: which fields of study pay off against the debt?

• Question: Which fields of study show the best earnings one year after graduation relative to debt at graduation?
• Dataset: the US Department of Education's College Scorecard data. Its field-of-study files cover award years 2014-15 through 2019-20 and include cumulative debt at graduation and earnings one year after graduation. The files download as ZIP archives and the page was last updated on June 10, 2026.
• Skills it proves: working with a very wide table, reading a data dictionary, handling missing values, building a ratio, and designing a dashboard around one decision.
• Tools: Tableau, Power BI or Looker Studio.
• Publish: a one-page public dashboard with one or two filters and a three-sentence summary that names its audience.
• Interview question it answers: "Who is this dashboard for, and what would they decide differently after seeing it?"

Finding template: "Among fields with at least [N] graduates in the data, [field] showed the highest ratio of first-year earnings to debt ([X]); [field] showed the lowest ([Y]). Fields with missing values were excluded."

7. Chicago traffic crashes: when and where do crashes cluster?

• Question: Which hours and weekdays see the most crashes, and which conditions are over-represented in the more serious ones?
• Dataset: Traffic Crashes – Crashes on the City of Chicago data portal: one row per crash on city streets under Chicago Police Department jurisdiction, with crash date, weather and lighting conditions, first crash type, primary contributory cause, a hit-and-run flag, a most-severe-injury field and ready-made crash hour and weekday columns. The city publishes it as shown in its crash-reporting system, so expect gaps.
• Skills it proves: bucketing timestamps into hour and weekday, comparing categories, and explaining why a count of crashes is not a measure of risk without traffic volume.
• Tools: SQL for the grouping, Power BI or Python for a heat map of hour against weekday.
• Publish: the heat map and a paragraph on what the data cannot say.
• Interview question it answers: "Does more crashes at night mean night driving is more dangerous?"

Finding template: "[N]% of crashes in [year] happened between [hour] and [hour] on [weekday]s. Without traffic counts for those hours this shows when crashes are reported, not when driving is riskiest."

8. Your own data: what does a year of your habits show?

• Question: Something only you can ask, such as how your spending, listening, training or commuting changed across a year.
• Dataset: an export you already own. Google Takeout lets you download data from your Google products, and many fitness, music and banking apps offer a similar download.
• Skills it proves: choosing a question yourself, cleaning an unfamiliar export format, and handling personal data responsibly.
• Tools: any.
• Publish: aggregated charts only. Never publish raw rows, and remove names, locations and account numbers before anything goes public.
• Interview question it answers: "Why this dataset?" A one-sentence personal reason is a legitimate answer and shows initiative.

This is the one project nobody else has, which matters when many applicants have used the same handful of famous datasets. Skip it if the data is sensitive: health and finance exports are better kept private.

How Many Projects Do You Need, and Which Ones?

Three to five. Seera's data portfolios feature 2.7 projects on average, and Dataquest's guide advises "3–5 polished, well-documented projects". Choose them so that one uses SQL, one produces a dashboard in a BI tool, and one ends in a written finding, then let the role you want decide the rest.

Finish and publish the project nearest your target role first. A page with one finished project is more use on an application than a plan for five, and you can add the next one while you wait to hear back.

If you are aiming at a BI analyst or reporting role, the dashboard stops being one project among several and becomes the product. The business intelligence portfolio guide covers what changes: showing the data model and measures behind each dashboard, and where to host Power BI and Tableau work so a reviewer can open it.

Which Data Analyst Portfolio Projects Suit Beginners?

Start with projects 8, 6 and 1. Each asks for one tool and one clear question, so you can practise the write-up before you take on a multi-table dataset. Projects 2, 3, 5 and 7 need joins, grouping or a statistics idea, and project 4 is the hardest because maps and outliers both call for judgement.

If you are switching careers and have no work history to show, pair one beginner project with one intermediate project, and read the guide to building a data analyst portfolio with no experience for how to write findings instead of impact.

Which SQL Projects Belong in a Data Analyst Portfolio?

Show SQL in at least one project and publish the query file, not a screenshot of its result. SQL is listed in 62 percent of the data portfolios in Seera's benchmark, so it is expected; a linked query is what proves you can write it. Each project above exercises a different SQL skill:

If you want a SQL project on event data, Google publishes an obfuscated sample of Google Merchandise Store events for November 2020 to January 2021 as the BigQuery public dataset ga4_obfuscated_sample_ecommerce. Its documentation says the free usage tier is enough to explore it, and warns that some fields hold placeholders and may not be internally consistent, which belongs in your Limits line.

How Do You Write Up a Project So It Earns an Interview?

Use the same six lines on every project: question, data, method, finding, limits and files. A reader who sees only those six lines should know what you did and where to check it.

The finding line is where self-made projects most often go wrong. Nobody acted on a project built from public data, so it can report what the data showed but not an impact. Compare two versions of the same result:

• ❌ "Improved delivery performance by 20%."
• ✅ "Across [N] delivered orders, those that arrived late averaged [X] review stars against [Y] for on-time orders."

The first claim cannot be defended in an interview, because nothing was delivered by you. The second can be reproduced on request. If you want to suggest what a business could do with the result, label it as a recommendation. The full argument is in the guide to writing findings instead of impact.

The files matter as much as the sentences. Dataquest's project guide puts the stakes plainly:

Many hiring managers check GitHub before an interview. A repository with no documentation is almost as bad as no repository at all.

A README of one page is enough: the question, a folder map, the data source with its licence, and the one or two commands needed to rerun the work. The project page template covers how to turn that into the page a recruiter reads before they open the repository.

How Do You Put Data Analyst Portfolio Projects on GitHub?

Use one repository per project, make the six-line write-up its README, and pin the best ones to your profile. GitHub lets you pin up to six repositories and gists combined, and three to five is enough.

A repository holds the working and a portfolio page holds the summary. In Seera's benchmark 66 percent of portfolios link a GitHub profile or repository, but only 20 percent attach a link to an individual project, so link each repository from that project's entry on your page. GitHub can also publish a static site from a repository, at owner.github.io, which suits a minimal page. If you want a designed page with your own domain and view counts without building it by hand, the data analyst portfolio example shows the layout to copy.

Three Real Portfolios That Show the Format

Disclosure: Seera is our product, and all three portfolios were created with it. They appear here because anyone can open the live pages, and each illustrates a different part of the format above. More are in the gallery of data portfolios published on Seera.

Every project links to its file: Harsha Kumar Pasala

Harsha Kumar Pasala shows three projects, one each in Excel, SQL and Power BI, and the title of each one links to the working file on GitHub: a supermarket sales workbook, a customer behaviour write-up and a mobile sales Power BI file. That is the "files open for a stranger" test applied literally. What to copy: link the file from the project title, not from a general profile.

A named dataset and a written finding: Benedict Chidiebere

Benedict Chidiebere analysed sleep patterns across 62 mammal species using the Allison and Cicchetti 1976 dataset, links the code on GitHub and published the results as a public article. The project says which variable turned out to matter most, which is the finding line in one sentence. What to copy: name the dataset and its year, and publish the write-up somewhere public.

A dashboard on real data: P. Tara Bhramaramba

P. Tara Bhramaramba built Tableau dashboards on housing sale prices, property features, renovation trends and geographic patterns. It is the same shape as projects 4 and 6 above: a public dataset, a map or comparison view and filters that let the reader explore. What to copy: pick a dataset with a place dimension, because a map gives a non-technical reader an entry point.

You do not have to build the page from scratch. Upload a resume to Seera, even one that is mostly coursework and projects, and it drafts an editable portfolio you can publish; one portfolio is free. Start with the annotated data analyst portfolio example to see what a finished page contains, or the student portfolio builder if you are still studying.

Frequently Asked Questions

What are good data analyst portfolio projects?

Good data analyst portfolio projects start from one clear question, use a real public dataset that is messy enough to need cleaning, and prove a different skill each. Eight that work are NYC 311 response times (cleaning), the Olist e-commerce orders (multi-table SQL), Online Retail II (customer segmentation), Inside Airbnb (pricing and maps), the Stack Overflow Developer Survey (survey statistics), College Scorecard (a dashboard), Chicago traffic crashes (time and place patterns) and an export of your own data.

How many projects should a data analyst portfolio have?

Three to five. Seera's benchmark of 56 data portfolios found an average of 2.7 featured projects, and Dataquest advises three to five polished, well-documented projects. Choose them so that one uses SQL, one produces a dashboard in a BI tool such as Power BI or Tableau, and one ends in a written finding, then pick the rest by the role you want.

Where can I find datasets for data analyst portfolio projects?

Start with public sources that publish their own documentation: city open-data portals such as NYC Open Data and the City of Chicago data portal, the UCI Machine Learning Repository, Inside Airbnb, the US Department of Education's College Scorecard, the annual Stack Overflow Developer Survey and Kaggle. Prefer datasets that arrive messy, and check each licence before you publish your own copy of the data.

Which tools should my data analyst portfolio projects use?

SQL, a BI tool and either Python or Excel. In Seera's benchmark of 56 data portfolios, Python is listed in 79 percent, SQL in 62 percent, Power BI in 57 percent, Tableau in 30 percent and Excel in 29 percent. A balanced set of three projects uses SQL in one, a dashboard tool in one and Python or Excel in one.

How do I show data analyst projects to recruiters?

Put the files on GitHub, host any interactive dashboard on Tableau Public, or publish it from Power BI, and link both from one portfolio page that gives each project a question, a finding and a limit in a few lines. Only 20 percent of the 56 data portfolios in Seera's benchmark attach a link to an individual project, so a page that links straight to the working files is already unusual.

What are good data analyst portfolio project ideas for beginners?

Start with your own data, the College Scorecard dashboard and the NYC 311 response-time analysis. Each asks one question with one tool and no multi-table joins, so you can practise writing up a finding, a limit and a link to the files. Add an intermediate project such as the Olist delivery analysis once the first is published.

What SQL projects should I put in a data analyst portfolio?

Choose projects that exercise different SQL skills and publish the query file for each. The Olist orders show joins across tables, NYC 311 shows dates and elapsed time, Online Retail II shows grouping and customer segmentation, and Chicago traffic crashes shows rates against a denominator. SQL is listed in 62 percent of the data portfolios in Seera's benchmark, so a linked query is what sets yours apart.

How do I put data analyst portfolio projects on GitHub?

Use one repository per project and make the question, data source, method, finding and limits its README. Add a folder of named SQL files, a small data sample or download script, and the notebook or workbook. Pin up to six repositories to your profile, then link each one from the project's entry on your portfolio page, because only 20 percent of the portfolios in Seera's benchmark link an individual project.

Related Reading

• Data Analyst Portfolio: 3 Project Templates That Get Hired
• Financial Analyst Portfolio: What to Show When Your Work Is Confidential
• Business Intelligence Portfolio: Examples, Templates and How to Show a Dashboard
• Data Analyst Portfolio With No Experience
• 7 Data Science Portfolio Examples Worth Studying
• Data Analyst Portfolio Example — Annotated Walkthrough
• Data Scientist Skill Benchmarks (2026)
• Real Data & ML Portfolios Published on Seera
• How to Make Your Portfolio Stand Out to Recruiters
• Turn Your Resume Into a Portfolio Website: Step-by-Step Guide

Useful sources and further reading

• Dataquest: Data Analyst Projects for Beginners — source of the quoted README guidance and the "3–5 polished, well-documented projects" advice
• NYC Open Data: 311 Service Requests from 2020 to Present — dataset for project 1
• Kaggle: Brazilian E-Commerce Public Dataset by Olist — dataset for project 2
• UCI Machine Learning Repository: Online Retail II — dataset for project 3, with the transaction count, date range and CC BY 4.0 licence
• Inside Airbnb: Get the Data — dataset for project 4, with the file list and CC BY 4.0 licence
• Stack Overflow Developer Survey — dataset for project 5, with the ODbL licence terms
• US Department of Education: College Scorecard data — dataset for project 6, with file coverage and update date
• City of Chicago Data Portal: Traffic Crashes – Crashes — dataset for project 7
• Google Takeout — export your own Google data for project 8
• Tableau Public — free hosting for interactive dashboards that open without an account
• Seera: Data Scientist Skill Benchmarks — source of the portfolio statistics, aggregated from 56 portfolios and free to cite under CC BY 4.0
• Seera: Resume to Website — the tool referred to above; it turns a resume into an editable portfolio page