If you genuinely know a reasonable real-world number, you can use it. The goal is not to pretend you know less than you do.

State the number and make clear that you are using it from general knowledge. If the interviewer wants an estimation approach instead, they can redirect you.

The more important issue is whether that number materially drives the answer. You should still explain the logic of the model and sanity-check the result rather than treating one remembered statistic as proof the entire estimate is correct.

If you are uncertain about the exact figure, round it and say so. A transparent assumption is better than presenting a half-remembered number with false precision.

I usually say "I believe it's roughly X, so I'll use that unless you'd prefer an estimate." Never had an issue.

The dangerous part for me was remembering a number that was outdated and then treating it as exact.