Once a question is split into a tree, it is tempting to work through every branch in order. That is slow, and most branches turn out not to matter.
Pick the hypothesis your managers already suspect, or the one your own data can check in a day. Test it first. If it holds, you have the start of the answer; if not, you have ruled out the obvious.
Mark the result on the tree the same day. A ruled-out branch is worth as much as a confirmed one, as long as everybody can see it.