How much of a data analyst's day is spent writing SQL?

Most of it. On a typical day, 3 to 5 hours go into writing, testing, and fixing SQL queries. That's the bulk of the work, not a side skill. The queries aren't glamorous—you're pulling a table, checking if the numbers match what someone expects, joining it to another table, and formatting it so a non-technical person can read it.

Do data analysts use Python every day?

Not at the entry level, and often not later either. SQL and a spreadsheet tool cover most entry-level and mid-level analyst work. Python shows up more in data engineering, machine learning, or advanced analytics roles, not in the daily grind of a standard analyst seat.

What meetings does a data analyst attend?

A typical week includes a team standup, one or two stakeholder meetings where you present or explain a number, and the occasional call where someone wants to understand a dashboard they can't read on their own. The skill that matters most in these meetings is explaining a number in one sentence to someone who doesn't care how you got it.

Is being a data analyst boring?

Some of it, yes. Most days include stretches of repetitive query writing and data cleanup that nobody would call exciting. The interesting part shows up when a chart finally answers a question someone's been asking for weeks, or when you catch an error before it reaches a director's inbox.

What's the difference between a junior and senior analyst's day?

A junior analyst spends more time on execution: writing the queries, building the dashboard, fixing the data someone else flagged. A senior analyst spends more time deciding what should get built in the first place and pushing back when a request doesn't make sense. Both still write SQL most days. The senior version just spends less time on syntax and more time on judgment.

What tools does a data analyst use daily?

In order of frequency: SQL editor, a BI tool like Tableau or Looker, spreadsheets like Excel or Google Sheets, Slack or email, and Python (occasionally). SQL editor is used most, while Python ranks lower than many expect, especially at the entry level.

Should You Use Kaggle Datasets in Your Portfolio?

Yes, you can use Kaggle datasets in your portfolio, but not as the whole project. Kaggle is a fine place to find raw data. The problem is that a project built on an overused dataset, with no original question and nothing messy to clean, looks like every other applicant's. A Kaggle dataset works when you pick a less common one, ask your own question, and show the work a hiring manager can't get from a tutorial.

This post shows when Kaggle helps, when it hurts, the datasets to avoid, and 6 steps to make a Kaggle dataset your own.

Should you use Kaggle datasets in your portfolio?
What Kaggle gives you
Why hiring managers groan at some Kaggle projects
Datasets to skip
6 steps to make a Kaggle dataset your own
A quick audit query to start with
Licenses and credit
Where else to find data
How to write it up
FAQ

I didn't have a portfolio site when I applied for jobs. I had 3 project links on my resume: 2 Tableau Public dashboards and 1 Power BI project. I'm self-taught with no relevant degree, so those links were the only proof I had. What mattered was whether someone could open one in 10 seconds and follow it. Where the data came from mattered far less than whether the project said something.

Should you use Kaggle datasets in your portfolio?

Yes, with conditions. Think of a Kaggle dataset as ingredients, not a finished meal. The dataset is a starting point, and the analysis is what you're actually showing.

A Kaggle project is worth including when it does these 3 things:

Starts with your own question. You decided what to ask before you looked at what the dataset could answer.
Shows real work. You cleaned something, joined something, or checked something, and you can show it.
Ends with a recommendation. A stakeholder could act on what you found.

If it does none of those, it's a tutorial with a different name. Portfolio projects that don't look like tutorials explains the difference in detail.

What Kaggle gives you

Kaggle is useful because it removes the hardest early step, which is finding data you're allowed to use. It has real advantages:

Free and easy to download. Most datasets are a CSV file you can open in minutes.
A huge range of topics. You can find data on almost any subject, which helps you match a project to your target industry.
Documentation on the page. Many datasets describe the columns, the source, and the license.
Practice material. It's a good place to try a new SQL technique or chart type on low-stakes data.

Those are real benefits, especially when you're starting out. The trouble starts when the dataset becomes the entire project.

Why hiring managers groan at some Kaggle projects

The issue is sameness. Many applicants pull from the same short list of popular datasets, so a hiring manager sees the same project dozens of times. A few things make it worse:

The data is already clean. Many Kaggle datasets have been tidied by the uploader, so there's no cleaning story to tell.
The question is generic. "Predict who survives" or "analyze sales" answers nothing a business would ask.
The work is copied. Public notebooks make it easy to follow someone else's analysis step by step, and a reviewer can tell.
There's no context. Without a stakeholder or a decision, the numbers are just numbers.

None of this is Kaggle's fault. It's what happens when a project has no original angle. The fix is in how you use the data, not in avoiding the site. If you want a better way to choose, how to find datasets worth building on covers what to look for.

Datasets to skip

Some datasets are so common that using them as your main project works against you. Skip these unless you're doing something truly different with them:

Titanic. Kaggle's intro competition and probably the most used beginner dataset anywhere.
Iris. A classic teaching dataset with too few rows and columns to say anything about business.
House price prediction sets. Heavily copied, and they push you toward modeling instead of analysis.
Superstore-style sales samples. They appear in countless dashboards, so yours won't stand out.
Any dataset with thousands of public tutorials. If a search turns up 50 walkthroughs, so will your reviewer.

Using one for practice is fine. Leaving it as a portfolio centerpiece is where it hurts. A quick test: if a hiring manager could guess your whole project from the dataset name, pick a different one.

6 steps to make a Kaggle dataset your own

You can turn almost any dataset into something distinctive. Follow these 6 steps:

Pick a less popular dataset. Sort by recent or niche topics, and favor ones with few public notebooks.
Write your question first. Frame it for a stakeholder, like "which product categories should the team stop discounting?"
Add a second source. Join the data to something else, like a population table, a calendar of holidays, or weather data. The join is the work.
Audit before you analyze. Check for duplicates, nulls, odd values, and anything that doesn't add up, and write down what you find.
Do the analysis in SQL. Show real queries with joins, aggregations, or window functions, not only a chart.
Finish with a recommendation. State what the stakeholder should do and what the data can't prove.

Step 3 does the most. A dataset joined to a second source can't be copied from a public notebook, and it proves you can combine data, which is most of the job. If you need help picking where to practice the SQL side, where to practice SQL for free lists good options.

A quick audit query to start with

Even a "clean" dataset deserves a check. This query shows row counts, duplicate IDs, and missing values in one pass:

SELECT COUNT(*) AS total_rows, COUNT(DISTINCT listing_id) AS unique_ids, SUM(CASE WHEN price IS NULL THEN 1 ELSE 0 END) AS missing_price, MIN(price) AS min_price, MAX(price) AS max_price FROM listings;

If total_rows and unique_ids don't match, you have duplicates. If the minimum price is 0 or negative, something is off. Writing down what you find, and what you did about it, is the cleaning story that most Kaggle projects are missing. Handling NULL values in SQL covers the next step once you find missing data.

Licenses and credit

Check the license on every dataset before you publish a project that uses it. Each Kaggle dataset lists a license on its page, and the terms differ. Some allow reuse freely, and others limit commercial use or require credit.

A few habits keep you safe:

Read the license on the dataset page before you start.
Credit the source and the author in your write-up, with a link.
Don't republish the raw data in your project if the license doesn't allow it.
Say what you changed, so nobody confuses your work with the original.

Crediting your source also reads well. It tells a reviewer you're careful about where data comes from, which is a real part of the job.

Where else to find data

Kaggle is one source of many, and mixing sources often makes a stronger project. Other places worth looking:

Government open data portals. City, state, and national sites publish data on transit, health, budgets, and more.
Public datasets in cloud warehouses. BigQuery hosts a set of public datasets you can query directly.
Your own data. A budget, a workout log, or a small business can make a project nobody else has.
Your current job. If your employer allows it, a sanitized version of real work is the strongest option. Using your current job as a portfolio project explains how to do it safely.

Industry-specific data helps too. If you're targeting retail roles, portfolio projects for retail and ecommerce shows what to build.

How to write it up

Treat the write-up as part of the project. A reviewer reads it before they open anything else.

Cover these points in about 1 page:

The question, framed for a stakeholder.
The source, with a link and the license.
What you cleaned or checked, including what was wrong.
The finding, in plain language.
The recommendation, and the caveats.

How to write a project README has the structure. For where to put it, see where to host your data analyst portfolio. And if you're wondering how many projects to include, how many portfolio projects you need has the answer.

The 90 days at Analyst Hive include project work with a clear structure, so you finish with something worth linking on your resume.

What people ask about Kaggle datasets in a portfolio

Is it okay to use Kaggle datasets in a data analyst portfolio?

Yes, as long as the project has your own question and real analysis. Kaggle is a legitimate source of data. A project that only runs a dataset through a standard tutorial adds little, but one that asks an original question and ends with a recommendation works well.

Do hiring managers care if my project uses Kaggle data?

They care about what you did with it. The source matters less than the question, the cleaning, the SQL, and the conclusion. An overused dataset with no original angle can hurt, and a niche one with a clear recommendation can help.

Which Kaggle datasets should I avoid?

Avoid the ones with thousands of public tutorials, like Titanic, Iris, and house price predictions. They signal that the project is a copy. If a quick search turns up dozens of identical walkthroughs, pick a different dataset.

How do I make a Kaggle project stand out?

Start with your own business question, join the data to a second source, audit and clean it, and do the analysis in SQL. Then write a short recommendation. The second source does the most, since it can't be copied from a public notebook.

Do I need to credit Kaggle datasets?

Yes. Check the dataset's license, credit the author and source with a link, and say what you changed. Some licenses limit reuse, so read the terms before you publish anything or redistribute the raw data.

Should I enter Kaggle competitions for my portfolio?

It's optional. Competitions focus on predictive modeling, which is more data science than analytics. For an analyst portfolio, a business-style project with a clear recommendation usually shows the right skills better than a leaderboard score.

If you want a day-by-day plan for building projects like this, join Analyst Hive and work through it in order.