How to Build a Data Analyst Portfolio That Gets Interviews

A data analyst portfolio that gets interviews is small, focused, and easy to scan: three to four projects that answer real business questions, each with a short write-up, visible SQL or code, and a dashboard a recruiter can open in seconds. Quality beats quantity. This guide shows you exactly what to include, how to structure each project, and where to host it.

What hiring managers actually look for

Recruiters and hiring managers often spend only a few minutes on a portfolio before deciding whether to call. In that short window they are checking three things:

• Can you frame a business problem? A project that starts with "Which product lines are losing margin and why?" beats one that starts with "I analyzed a dataset."
• Can you do the technical work cleanly? Readable SQL, a sensible data model, and accurate numbers.
• Can you communicate? A clear headline, a recommendation, and charts that need no explanation.

Everything in your portfolio should serve one of those three signals. If a project does not, cut it. Before you build, compare your plan with the list of BI analyst skills employers want so each project proves something specific.

Choosing the right projects

Aim for three to four, each showing a different skill

A strong mix looks like this:

Use data from a domain you understand

If you have worked in retail, distribution, healthcare, or logistics, build projects in that world. You will ask better questions, spot errors faster, and talk about the results with real confidence in interviews. Public sources include government open-data portals, Kaggle datasets, and sample databases such as Microsoft's AdventureWorks or Contoso. Never use a current or former employer's confidential data, even if it is anonymized.

Avoid the overused datasets

Titanic survival and the classic Iris flowers are fine for learning, but reviewers have seen them hundreds of times. A messy, realistic business dataset is far more memorable.

How to structure each project

Use the same layout for every project so reviewers know where to look:

1. Headline finding. One sentence at the top, for example "Three product categories drive most of the returns in the West region."
2. Business question and context. Who would use this, and what decision does it support?
3. Data. Source, size, time period, and any important limitations.
4. Approach. Cleaning steps, model design, and key calculations, with links to the code.
5. Results. Two or three screenshots or a link to the live dashboard.
6. Recommendation. What should the business do next?
7. What you would improve. Shows maturity and self-awareness.

Show a short, readable sample of your logic directly in the write-up. For a repeat-customer analysis, that might be:

WITH first_orders AS (
    SELECT customer_id,
           MIN(order_date) AS first_date
    FROM orders
    GROUP BY customer_id
)
SELECT COUNT(DISTINCT f.customer_id) AS customers,
       COUNT(DISTINCT CASE
             WHEN o.order_date > f.first_date
              AND o.order_date <= f.first_date + INTERVAL '90 days'
             THEN f.customer_id END) AS repeat_90d
FROM first_orders AS f
LEFT JOIN orders AS o
       ON o.customer_id = f.customer_id;

(The interval syntax shown is PostgreSQL. In SQL Server you would use DATEADD(day, 90, f.first_date).)

Where to host your portfolio

GitHub for code

Create one repository per project with a clear README that follows the structure above. Keep SQL in .sql files, Python in notebooks or scripts, and include a small sample of the data or instructions to download it.

A live dashboard link

Reviewers love clicking a working dashboard. Tableau Public lets you publish dashboards for free, though everything you publish there is public. For Power BI, "Publish to web" creates a public link, but your organization's settings and license type affect whether it is available, so check Microsoft's current documentation. If public publishing is not an option, record a short screen video walking through the report.

A simple landing page

A one-page site, a GitHub profile README, or a well-organized Notion page can act as your hub. It should include a two-line introduction, links to each project with its headline finding, and contact details. Link to it from your resume and LinkedIn profile.

Connect it to your resume and LinkedIn

Your portfolio only works if people find it. Add one line per project to the resume section where you list experience or projects, written as an outcome rather than a task. For example, "Built a Power BI sales dashboard that flags regions below target and reduced a weekly manual report to a single refresh" says far more than "Created a dashboard." On LinkedIn, use the Featured section to pin your best project and write a short post about each one when you publish it, explaining the question and the most interesting finding. Keep the tone practical and specific. Hiring managers searching for analysts in your area may see those posts, and each one is another chance to show how you think.

Finally, tailor the order. If you apply for a role that leans heavily on SQL, put your SQL project first on the landing page for that application.

Using AI to build faster (without faking it)

AI assistants can speed up portfolio work a lot, as long as the thinking stays yours. Good uses include brainstorming business questions for a dataset, explaining an error message, drafting a README outline, and reviewing your SQL for readability.

Be ready to explain every line in your projects. Interviewers often ask "why did you model it this way?" and it shows quickly when a candidate cannot answer. You can also mention in your write-up where AI helped, which signals honest, modern work habits. For more ideas, see our guide to ChatGPT and Claude for analysts.

Common mistakes that cost interviews

• Too many unfinished projects. Three polished projects beat ten half-done notebooks.
• No business context. Charts without a question or recommendation look like tutorials.
• Copied tutorial projects. Following a course is fine for learning, but change the dataset and the question before showing it.
• Broken links. Test every link from a private browser window before applying.
• Cluttered dashboards. Too many visuals and colors hurt more than they help.
• Unverified numbers. Reconcile your totals. A single obvious error can undo a strong first impression.

Next steps

Start this week with one project in a domain you know. Write the business question first, build the SQL and dashboard, then write the one-page case study. When it is published, ask a friend outside data to read the headline and recommendation and tell you what they understood. Then repeat with a project that shows a different skill. If you want a ready-made idea, our Power BI retail sales dashboard walkthrough is a solid first build that you can adapt with your own data and questions.

Watch: video lessons

▶

Frequently asked questions