In classical project management, you guess when features will be complete, promise the delivery date to the customer, and now that your scope and schedule are fixed, whip up the team to get the work done on schedule. As reality hits you in the face, you constantly shuffle features, or split features, to project a plan that you can meet. The constant re-planning is “waste” and the changing plan is disruptive.

In my favorite kind of Kanban, the teams work the backlog in priority order and use Lean and Agile techniques to push up their productivity. Strategic leadership is implemented by ordering the backlog. Expectation are set by projecting based on backlog position and throughput. (backlog position / stories/week = weeks to delivery) No one has to guess anything! Story size averages out, but creating more consistent stories improves projection accuracy.

I envision a feature to support it looking like this:

1.   Define you r release schedule:
   

   • Continuous releases
   • Regular release intervals
   • A list of release dates.

2.   Track the throughput as work is done
   

   • Prime it with a guess for new teams.
   • Have a way to compensate for holidays and vacations
   • Have a way to compensate for team changes

3.   Present a Gantt chart of where the features are projected into releases. (or add and attribute to each story with a projected deliver date)
   

   • Adjust for dependencies
   • Show when the expected delivery date is not met.
   • Calculate buffers based on historic variation???

4.   Make It really easy to adjust the backlog order and see the revised Gantt chart. (or attributes)
   

   • Protect work already in progress, but allow it to be pushed back into the backlog, if necessary