Uniformly random demographics produce charts that look wrong to anyone who has seen a real one. The fields here are weighted to plausible population shapes, so a bar chart built on this data has the lumps you would expect.
What you can control
- Every field is weighted rather than uniform — two-person households are the most common, doctorates the rarest.
- Pick only the fields you need, so a survey fixture does not carry columns your form never collects.
- Gender includes non-binary and prefer-not-to-say options, because production forms do and your charts should handle them.
- Attach a name to each profile to turn it into a full respondent record.
What this is not
The weightings are plausible rather than sourced from a specific census. Treat the output as realistically shaped test data, not as a statistical sample of any real population.
Questions
Where do the distributions come from?
They are hand-set to broadly resemble published population figures for developed economies. They are not drawn from a specific census and should not be cited as such.
Can I change the weightings?
Not from the interface. What you can do is pick which fields appear and generate a larger batch, then filter it down to the shape you need.
Is this suitable for training a model?
No. It will teach a model this generator's assumptions rather than anything about real populations. It is for exercising code paths and filling charts.