One of my job now is to design a database for a survey. Let's say we have a series of questions (web-based), in which one page contains one question. Not every person will be given the same questions, those are based on their previous answers and also randomness.

I would like to know whether it is better to have database like this

user   question    answer
userX  question1   answer1A
userX  question2   answer2C
userX  question5   answer5F
userY  question1   answer1B
userY  question3   answer3B
userY  question6   answer6D
...

or

user   q1  q2    q3    q4    q5    q6
userX  1A  2C    null  null  5F    null
userY  1B  null  3B    null  null  6D
...

My idea here is, using the second approach seems better, however I would like to know whether updating the table is (much) slower than inserting a new row? Also with the first approach, I can omit having some null answers.

The total questions given are fix, the client wont add any more question later on.

So my question is, what will you do if you were me?