Test data should look like data
A fixture full of Test User 1 hides an entire class of bug. Real names bring
apostrophes, hyphens, double-barrelled surnames, and non-Latin characters — all
of which reach your validation, your sort order, your column widths and your
email templates.
Two things break most often:
- Width. A Cyrillic full name is typically 20–30% longer than its English equivalent. Table columns and fixed-width cards fail here first.
- Sorting.
Ёsorts unpredictably unless a collation is set explicitly.
Gender agreement in Russian
Russian surnames inflect. Both of these are wrong in a way a native reader spots in under a second:
| Wrong | Right | | --- | --- | | Анна Иванов | Анна Иванова | | Сергей Смирнова | Сергей Смирнов |
This generator pairs a feminine given name with the feminine surname form, so seeded data reads as though a person entered it.
Real names, honest use
Every name here belongs to someone. Seeding a database, filling a mockup, and load-testing a form are all fine. Attaching one to a quote nobody said is not.
Questions
Where do these names come from?+
The English list is drawn from computing history — Lovelace, Hopper, Dijkstra, Liskov. The Russian list uses the most common given names and surnames. Both are real names of real people, which is exactly why they belong in fixtures and not in anything that pretends a person exists.
Why do Russian surnames change?+
Because in Russian they agree with gender: Иванов becomes Иванова for a woman. A generator that pairs Анна with Иванов produces text a native reader flags immediately, which defeats the point of realistic test data.
Can I use these in a public demo?+
For screenshots and seeded databases, yes. Do not present them as testimonials, reviews, or real customers — that is fabrication regardless of where the name came from, and it is the kind of thing that ends up in a regulator's complaint.
Do I need unique names?+
Repeats are possible and usually desirable — real data has duplicates, and code that silently assumes names are unique should fail in testing rather than in production.