How do you decide what your team can commit to next quarter?

Start from real capacity, not headcount. Subtract on call, support, holidays, interviewing, meetings and a realistic allowance for unplanned work, which historically runs twenty to thirty percent. Use your team's actual recent throughput rather than an idealized number, commit to less than full capacity, and rank the plan so the last items are the known ones to fall off. Reforecast mid quarter, in public.

Headcount is the wrong starting number. Eight engineers is never eight engineers of feature work: on call effectively removes one person, support and escalations take around fifteen percent, then holidays, interview load and the meeting tax. Historically my teams land at roughly fifty five to sixty five percent of theoretical capacity on planned work, so that is what I plan with, and I can show the last four quarters to back it up when somebody thinks I am sandbagging. Then I commit to about eighty percent of that and rank everything, so it is clear in advance which two items fall off if something goes wrong, rather than picking under pressure in week ten. I also flag single points of failure at planning time: if a project depends entirely on the one person who knows that subsystem, that is a risk on the plan, not a surprise in March. And I reforecast at the midpoint in public, because the first six weeks of a quarter always tell you something.