Maybe it's not a good idea to conclude universal "facts" about people's preferences based on what three individuals indicate they like.

But it would sure help polling companies, market research, medical trials, etc. if such small sample sizes were adequate.