Ask only what changes the match
Every question on a formation survey should change who ends up with whom. If an answer would not move a single student between teams, cut it. Long surveys lower response rates and rarely improve the teams.
The inputs that earn their place
A few things consistently matter: the skills and experience someone brings, when they are actually free to meet, the kind of work or problem they want to take on, and a small number of people they would like to work with. Most good formation comes from these.
Be careful with free text
Open questions feel thorough but are hard to act on at scale. If you use them, keep them focused, one on strengths, one on working style, and have a plan for turning the answers into something you can actually compare across a cohort.