Slots, Seats and Benches
Capacity that is not made of people: why a slot behaves differently from an hour, the simultaneity constraint, and how to hold two accounts without adding them.
Some capacity is not made of people. A theatre list, a bay, a bench, a bed, an appointment, a licence seat, a delivery window — each is a quantity of simultaneous things rather than a quantity of time, and the difference changes the arithmetic completely.
The reconciliation in “Slots, Seats and Benches” is strongest when assumptions and current work records can be examined separately. A team considering see how this process is supported for employee time tracking can add time and project context, with transparent settings, limited access and a clear correction route around consequential decisions.
Plans that convert these into hours lose the only property that constrained them, and then cannot explain why the schedule does not fit.
Additional context for “Slots, Seats and Benches” is available from McKinsey people and organisation insights. Use external material to prompt a specific validation check rather than replace local judgement.
What a slot actually is
A unit of simultaneity: one thing can happen in it, during a defined window, and no amount of effort makes two things happen in one slot.
Hours are fungible across time; slots are not. Four hours of bench time spread across a week does not help a job that needs four hours on Tuesday, and the plan in hours cannot see the difference.
The two numbers a slot account needs
How many slots exist in the period, and how many of them are in the window the demand actually needs. The second is almost always smaller and almost never recorded.
An operation with twelve bays and demand concentrated into two hours a day does not have twelve bays of capacity; it has twenty-four bay-hours in a window that needs sixty.
Where the two accounts meet
People and slots constrain each other: a bay without a technician is not capacity, and a technician without a bay is not capacity either.
The usable arrangement is two accounts, each reconciled separately, with a third short line recording the binding one for each period. Adding them produces a number that describes nothing.
The utilisation trap
Slot accounts attract utilisation targets, and high utilisation of a slot resource produces queues at a rate that surprises everybody who has not met queueing behaviour before.
Running a slot resource near full occupancy means any variation propagates, because there is nowhere for it to go. The practical limit is well below a hundred per cent and the right number depends on how variable the work is.
What breaks first
| Occupancy | What is observed | What it means |
|---|---|---|
| Below 70% | Slots visibly idle | Spare, and criticised for it |
| 70–85% | Smooth, occasional waiting | The working range for variable work |
| Above 85% | Queues lengthen quickly | Variation has nowhere to absorb |
| Near 100% | Everything waits on everything | The plan is a list of hopes |
The table describes a tendency rather than a law, and the thresholds move with how variable the arriving work is. What does not move is the shape: the cost of occupancy is not linear.
The licence seat, which is the modern version
Software seats behave exactly like benches: a fixed number of simultaneous users, a renewal date, and a procurement process that is slower than the plan.
They belong in the slot account with their renewal dates, and they are the capacity constraint that most commonly surprises a technical team mid-quarter.
Borrowed and shared slots
Where a resource is shared with another team, the available slots are not the total slots, and the share is usually informal.
Writing down the share — even an arbitrary one — converts a weekly negotiation into a planning input. Both teams can then plan honestly, which neither can do while the split is decided by who asks first.
Seasonal and windowed capacity
Some slots exist only part of the year or part of the day: a clinic that runs on Tuesdays, a yard that floods, a machine shared with production at month end.
Those belong in the account with their windows, because a plan that treats them as always-available will be wrong at exactly the busiest moment, which is when the window closes.
The slot that is really a person in disguise
Occasionally what looks like a room or a machine constraint is a qualification constraint: the bay is free, and only one person is signed off to use it.
Checking which of the two is binding takes one question and changes the remedy entirely. Buying a second bay when the constraint is the sign-off is the most expensive way to make no difference at all.
Measuring occupancy honestly
Count the slot as occupied when it cannot be used by anything else, not when work is visibly happening in it. Setup, cleaning, turnaround and waiting for the next job all occupy the slot.
Measured that way, most operations discover they are running ten to twenty points higher than their reported figure, which usually explains the queue that the reported figure said should not exist.
Keeping the slot account
One row per resource type: how many, in what window, shared with whom, and what makes a slot unusable. Four fields.
It is normally half a page and it explains a large share of the schedule failures that the people account cannot account for, which is why it is worth keeping even where the organisation thinks of itself as planning people.