It depends on the product, so the best approach is to know the demand of the product, how many people use/buy the product. Once you get that number you will be able to estimate.
- The customer has a problem.
- The customer understands he or she has a problem.
- The customer is actively searching for a solution and has a timetable for finding it.
- The problem is painful enough that the customer has cobbled together an interim solution.
- The customer has committed, or can quickly acquire, budget dollars to solve the problem.
If you know that there is 1000 individuals having the "problem" and you get 100 surveys you will know what the 10% of these people think about the "problem".
Out of that 10% (100 individuals) you can get the average like: 1 out of 10 think that bla bla bla.
You must know the starting number (how many individuals have the problem). If the number is low your percentage of surveys will be high and vise versa, but there will be cases where you will need to have a minimum that works for you, is realistic, and doesn't affect or misrepresent the truth.
Individuals with the problem Surveys
1 to 1000 1 to 100 -> 100%
1000 to 10.000 100 to 1000 -> 10% This would be your minimum
Let's say 100.000 individuals have the "problem". It may not be viable for you to survey the 100.000 individuals (100%) so, you may try to survey 10% (10.000 individuals). But, still this may not be viable, so what to do in this case? Reform the survey so the number of individuals having the problem is more flexible to your capabilities or hire more people for the surveys :)