Ensure you estimate the number of story points for defects being corrected. Some will be zero if they are very small, but there will be significant defects being worked on that consume available time, reducing velocity. It would be good to capture this data and show it on a stacked bar graph for each iteration, showing core productive velocity (# points for new features) and defect-related velocity (# points for defects). This allows management to better determine workload and capacity issues.

If you don't do this, total workload is hidden, and it is hard to see what the reason is for dropping velocity.

Finally, if a huge technical debt of defects keeps getting pushed out, you can see where you’ll have to reduce scope (or add an iteration) to accommodate if they are appropriately sized.