What similarity percentage on a formal lab report looks suspicious?

If a group of students do a lab report on the same experiment at university, what per centage similarity rating do you think would look suspicious?