Capacity planning usually starts by counting available hours and allocating work against them. The plans fail consistently, and the reason is that hours are not the unit the work is actually done in.
Available time is not usable time
A working day contains meetings, questions from colleagues, escalations and administration, none of which appear in a plan built from headcount and hours.
What remains is broken into intervals, and demanding work requires intervals long enough to hold the whole problem in mind before progress begins.
So a day with four hours of scattered availability delivers considerably less than a day with four continuous hours, even though both look identical on a resourcing sheet.
Interruption costs more than its duration
A short question suspends whatever was in progress, and resuming requires rebuilding the state that was lost, which takes longer than the interruption itself.
The person interrupting sees only the minutes they used, while the person interrupted absorbs a much larger cost that is invisible to everyone else.
Because the cost is invisible, nothing in the organisation naturally limits interruption, and it accumulates until someone measures output rather than availability.
High utilisation removes the buffer that absorbs variation
Planning a team to full utilisation assumes work arrives evenly and takes the estimated time, and neither holds in practice.
With no slack, a single unexpected item has nowhere to go, so it delays something else, and the delay propagates through every dependency behind it.
Systems run near capacity develop long queues from small disturbances, which is why a fully allocated team can miss dates that a slightly under-allocated one would have absorbed.
Estimates describe the work, not the environment
People estimate how long a task takes under reasonable conditions, because that is the only version of it they can picture.
The environment then adds waiting for answers, rework after a change of requirement, and competing demands that the estimate never contemplated.
Comparing estimated against elapsed time repeatedly gives a multiplier that describes the environment, and that multiplier is more useful for planning than any individual estimate.
Protecting blocks works better than adding people
Adding staff to a saturated team increases coordination demand and can reduce output in the short term, particularly where the new people need supervision.
Reserving parts of the week where interruption is not permitted increases usable capacity without changing headcount at all.
The approach only holds where a manager absorbs the escalations during those periods, since protection that depends on individual refusal collapses the first time something urgent appears.