One thought I’ve had many times (open to feedback on this philosophy)
Estimates for completion dates in software should be given as a probability distribution instead of a date.
Instead of: I estimate the release will be shipped by 03/01/2025
I tell my manager: I estimate the release will be shipped by 03/01/2025 +- 10 days at one standard distribution confidence level. Plan accordingly based upon your risk tolerance for missing the deadline.
I've had mixed results with this approach. What happens when I try it is:
1. I communicate DATE and CONFIDENCE_INTERVAL
2. My manager hears DATE and CONFIDENCE_INTERVAL.
3. My manager tells project-manager "probably somewhere around DATE"
4. Project-manager hears "DATE"
5. DATE becomes what people thought I said.
Estimates for completion dates in software should be given as a probability distribution instead of a date.
Instead of: I estimate the release will be shipped by 03/01/2025
I tell my manager: I estimate the release will be shipped by 03/01/2025 +- 10 days at one standard distribution confidence level. Plan accordingly based upon your risk tolerance for missing the deadline.