I think the important part here is how you move from those individual estimates to the global estimates. I could never get this part right. Whenever starting from time estimates,someone in the end was always adding these "23 days", 15 days" and saying "ok, we'll be done in March"https://twitter.com/ChristerEricson/status/1027976350131339265 …
Say you have tasks task_i estimated as days_i. When sum{1,N-1}(days_i) shows task_N start 3 months after you have first (data or code) dependencies on task_N, you can with high probability assume you have a scheduling or scoping problem. Without estimations, you can't say shit.