It’s fun to put my hands on data gathered by a team I don’t directly work with. In such case every record is equally meaningful, and at the same time meaningless, for me. I’m not biased. I don’t know what to look for. It means that any hypothesis I make is based only on what I learn from the data.
Sometimes, however, the process is a bit frustrating. It is so when I find data incomplete.
I don’t need much. Usually three dates are enough: when a task hit backlog, when a team started working on it and when it was done. If the team uses sizing it also proves useful.
The problem is when I don’t have this data I can make something up, which doesn’t make sense, or delete incomplete records, which cripple the data as a whole. Either typhus or malaria.
So if you gather data to measure the work please, please, please, get it complete. If you have problems to gather a specific class of information just get rid of the whole class. It’s not useful anyway. And then get the rest complete.
Otherwise you will find it difficult to learn much from your measures.