Sample dataset · CC0

A film catalog nobody owns.

Nine columns of film record spanning fifty years. Titles, directors and genres are invented, ratings run on a bell curve, and gross is derived from budget with a hit rate that means about a third of the catalog loses money.

9 columns · 100 or 1,000 rows as a file · up to 1,000,000 rows generated here · CSV, JSON, JSONL, Excel, Parquet

Take a file

The 100-row file is a good classroom size: small enough to read on a projector, wide enough to have something to sort by.

Format	Small	Standard
CSV	100 rows (8.1 KB)	1,000 rows (80.8 KB)
JSON	100 rows (18.7 KB)	1,000 rows (186.7 KB)
JSONL	100 rows (18.4 KB)	1,000 rows (183.8 KB)
Excel	100 rows (23.0 KB)	1,000 rows (155.0 KB)
Parquet	100 rows (6.2 KB)	1,000 rows (35.8 KB)

A bigger catalog

Bigger sizes recombine the same title pools, so at large row counts you will see repeated titles with different years and directors. Treat the title as a label rather than as a key.

The first eight rows

The first eight rows of the 1,000-row file. About one title in eight contains a comma and is quoted.

title	release_year	genre	director	runtime_min	rating	votes	budget_usd	gross_usd
The Silent Cartographer	2025	Science Fiction	Jennifer Allen	170	7.0	83933	191170000	202054728
The Crimson Promise	2008	Mystery	Larry Stewart	165	5.5	68151	29480000	20273289
The Crimson Archive	2025	Mystery	Mei Jones	109	6.0	340222	22460000	92001476
The Winter Cartographer	1981	Action	Emily Martin	134	8.2	49134	72060000	236885825
The Quiet Harbour	2020	Mystery	Eric Cook	107	7.2	247515	90750000	149231359
The Winter Circuit	2015	Drama	Carol Shah	177	6.4	42259	86540000	49780652
The Last Promise	1994	Horror	Donna Lopez	170	5.3	571887	177440000	124710671
The Second Cartographer	1979	Drama	Sofia Thompson	123	5.4	878748	124510000	20134226

Nine columns, and what each one holds

Column	Type	What it holds	Example
title	text	Invented title. Some contain a comma, which is the point of a quoted CSV field.	The Silent Cartographer
release_year	integer	1975 to 2026.	2025
genre	text	One of ten genres.	Science Fiction
director	text	Invented director name.	Jennifer Allen
runtime_min	integer	Runtime in minutes, 74 to 198.	170
rating	decimal	Audience score out of 10, one decimal.	7.0
votes	integer	Number of ratings, long tailed.	83933
budget_usd	integer	Production budget in whole dollars.	191170000
gross_usd	integer	Worldwide box office. Roughly one film in three loses money, which is about right.	202054728

What it models

A film reference table of the kind used in every introductory data course: a title, a year, a genre, a person, three measures and a rating. It is deliberately the friendliest dataset here, because the first thing you teach somebody should not also require explaining what a SKU is.

Everything in it is invented. Titles are built from an adjective and a noun with the definite article, so you get The Silent Harbour and The Broken Observatory rather than anything with a rights holder attached. About one in eight is a sequel with a comma in the title, which is where the CSV quoting comes in.

Why the money columns are related

Box office is not drawn independently of budget. Roughly a third of films return between 15 and 95 percent of what they cost, and the rest return between 1.05 and about 7 times, with the multiplier squared so the enormous hits are rare. The result is a profit distribution with a heavy loss cluster and a thin, very profitable tail, which is close to how the industry actually works.

That gives a scatter plot of budget against gross a real shape: a diagonal band with a cloud of flops below the break-even line. Draw the line where gross equals budget and roughly a third of the points fall under it. Most sample film data draws the two columns separately and produces a shapeless blob, which is fine for testing an axis and useless for teaching a correlation.

Ratings and votes

rating is a bell centered around 6.4 out of 10 with one decimal place, which is roughly where public film ratings actually cluster, and it means a histogram of it looks like something rather than like a rectangle. votes is heavily right skewed, because most films are rated by a few hundred people and a handful by hundreds of thousands.

The pair is a good subject for a weighted average. The mean rating across the catalog and the vote-weighted mean rating are different numbers, and the difference is exactly the popular-films-are-rated-differently effect. It is a small analysis with a real conclusion, which is what you want from a teaching file.

What people use it for

A first lesson, because the subject needs no explanation.
Scatter plots with a real relationship: budget against gross, with a break-even line.
Weighted averages, comparing mean rating against vote-weighted mean rating.
Sorting and top-N queries, on a catalog with a genuine long tail of grosses.
Testing quoted fields, on the sequel titles that contain a comma.

Open it somewhere useful

This is the friendliest dataset in the collection for a first lesson, because everybody has an intuition about films.

Open in Data Explorer

The full workspace: grid, filters, pipeline, SQL and charts, on this file.

Open in dashboard

Straight into the chart builder with the file already loaded.

Pivot it

Genre down the side, decade across the top, mean rating in the middle.

Profile it

The votes column is the clearest right skew in the collection.

Convert to Markdown

A hundred films as a table you can paste into a document.

Convert to JSON

Check that the titles with commas survive the round trip.

License, and the people in it

This dataset is dedicated to the public domain under CC0 1.0. Put it in a course, a paid product, a test suite, a bug report, a screenshot, a conference talk or a book. There is nothing to ask for, nothing to sign and no attribution required. A link back is welcome and is not a condition.

No real film, studio, director or performer appears in this file. Titles are assembled from adjective and noun pools, and director names come from the same fixed pools as the other datasets, so nothing here is a real person or a real work.

Questions people ask about this file

Are these real films?

None of them. Titles are assembled from an adjective pool and a noun pool, directors come from the same fixed name pools as the rest of the collection, and every number is generated. A sample film dataset built on real titles carries a rights question that nobody wants to answer, and it also tempts people to check the numbers against reality and find them wrong.

Is gross related to budget?

Yes, on purpose. About a third of films return less than they cost, and the rest return between about 1.05 and 7 times, with the large multipliers rare. That gives a budget against gross scatter plot a real diagonal band with a cloud of flops beneath it, so the break-even line means something and the correlation is worth computing.

Why do titles repeat at large row counts?

Because the title pools are finite: fifteen adjectives and fifteen nouns give 225 base titles, plus the sequel variants. At a thousand rows you will already see repeats. Treat title as a label rather than as a primary key. If you need a unique key at scale, use the row position or add one.

What is the rating distribution?

A bell centered near 6.4 out of 10 with one decimal place, which is roughly where public film ratings really cluster. That matters for a histogram lesson: a uniform draw produces a flat rectangle that teaches nothing about shape, while this produces something a class can describe out loud.