A possible bias with on$-$line research is that

a$.$ it is slow

b$.$ it draws responses mostly from trend leaders

c$.$ does not have a wide reach

d$.$ only computer professionals respond