Strengthen backlog refinement, so work is ready before it starts.
Improve production stability (fewer bugs/incidents driving surprise work).
Introduce WIP limits to avoid over commitment.
Create clear policies: what interrupts the team vs what waits. Which can be a political minefield.
Introduce roles like a rotating “on-call” or “expediter” - we had a support person per week who worked on lower level items that could be dropped if needed.
Reduce context switching (a major hidden cost)
The biggest shift: stop treating these percentages as a people problem.
AKA W. Edwards Deming - most issues are caused by the system, not the individuals.