My inclination is to suggest that the first round be as simple as possible (no reduction), and participants then comment on the results analytically (e.g., "tomato too prevalent", "too sour", whatever). If there is general consensus on one or more comments, attempt to identify the cause and iterate until there is general consensus that it is, at least, "OK". Then introduce a new variable (e.g., use of reduction) and see if there is general consensus that the results are now better than before, the same, or worse. Adjust again on that basis. Etc.
What do others think ?
** Phil.