Sorry to hear that. These things are sometimes very unpredictable.
I don't remember the exact numbers I did that day, but It is an estimation problem. I started with US population- 300 million...cards will not be issued for ppl under 15 and old people, lets say 20%, so remaining 2.4 M. Similarly, subtract X% for people with bad credit, y% for people who bought card within last year, since they are unlikely to get another..and so on. Essentially, i started with max limit and subtracted an estimated % for the filters I thought were applicable.
I remember in the end, the interviewers were not even looking at the final number I came up. So they probably just wanted to see how I approached the problem.