Whenever we are doing a quantiative risk analysis for a project we need to obtain uncertaintes from experts. How do you generally do? What distributions you end up using? How do you include discrete events? Saving Changes...
1. Use a combination of techniques - expert judgment, documented historical data, third-party industry data (if available)
2. Try to overcome individual bias by using tools such as Delphi method
When it comes to picking the "right" distribution, you could assume a normal distribution if the data appears to support that, but often times with service-focused time data, a lognormal distribution might end up being the better fit.
With discrete events, I usually would look to historical data assuming it is accurate and complete.