Downloads
Free practice datasets you can download right now in CSV, Excel, JSON or SQL, plus notebooks and templates as we finish them.
Sample datasets, generated on demand
14 fictional datasets, 100 to 10,000 rows each. Made when you click, never stored.
Two years of orders from a fictional chain of Indian sweet shops, with Diwali and Raksha …
An HR file of employees with salary, satisfaction, overtime and whether they left: built for attrition …
Marks in five subjects with attendance and study hours, plus totals, percentage, result and CBSE-style grade.
Customer records with contact details, segment, orders and spend: ideal for cleaning, deduplication and RFM.
Monthly stock register for five warehouses: opening, received, issued, closing, reorder level and value.
Daily sessions, users, bounce rate and conversions by page, traffic source and device.
Simulated daily temperature, rain, humidity and AQI for Indian cities, with monsoon and winter smog seasons.
Daily open, high, low, close and volume for made-up companies (tickers start with SIM): practise returns, …
Personal bank statements with salary, rent, bills and UPI spends, plus a running balance.
Loan applications with income, credit score, loan size and term, and whether each was approved.
Users rating fictional films from 0.5 to 5 stars: build your first recommendation system.
10-minute readings from factory machines that slowly wear out, warn and fail: predictive maintenance practice.
Orders with distance, weather, traffic, prep and delivery times, and the customer's rating.
Survey answers on a five-point scale, an NPS score from 0 to 10 and short written …
Hand-prepared files: what is coming next
The datasets above are generated, so they are clean by design (unless you ask for mess) and safe to share. Alongside them we are preparing a second kind of download: files modelled on real project work, which take longer to get right. Here is why.
Where the data comes from
Most of what I teach on this site comes out of real work: sales reports for a sweets company, stock sheets, customer lists, invoice exports from accounting software, attendance registers. That's the data that actually teaches you something, because it's messy in the ways real data is messy. Dates typed three different ways. A "Total" row sitting in the middle of the sheet. Mobile numbers stored as numbers, so the leading zero is gone.
The problem is I can't hand that data out. It belongs to the businesses I work with, and some of it has people's names, phone numbers and GST details in it. Publishing it, even "just for practice", would be a breach of trust and in many cases a legal problem too.
So what we're doing instead
For every dataset that goes up here, we go through roughly the same routine:
- Keep the shape, drop the people. Names, phone numbers, emails, addresses and GSTINs get replaced with made-up ones. Customer IDs are re-numbered so nothing can be traced back.
- Shift the numbers. Amounts and quantities are scaled and jittered, so totals still look believable but don't reveal anyone's real revenue.
- Keep the mess. This is the part most "practice datasets" skip. We deliberately leave in the blanks, duplicate rows, mixed date formats and typos, because cleaning them is the skill you're practising.
- Test it against a lesson. Before a file goes live, I work through it myself with the matching lesson or blog post to make sure the exercise actually works and the answers come out right.
That last step is the slow one. A dataset that looks fine can turn out to be useless once you try to teach with it, so I'd rather publish five good files than fifty random ones.
What's coming first
- A monthly sales file (12 sheets, one per month) for the merge-and-clean exercise on the blog.
- A small customer list with duplicates and broken phone numbers, for the data cleaning checklist.
- Jupyter notebooks that go with the Pandas and Machine Learning tracks in Learn.
- A couple of plain Excel templates (GST invoice, stock register) that you can use as they are.
Everything will be free, with no sign-up needed. In the meantime the sample datasets, cheat sheets and code snippets are ready to use, and most lessons include small examples you can type straight into a notebook.
Have a dataset you'd like us to build for practice, or found a problem in one? Tell us. Suggestions from readers usually go to the top of the list.