A manual stamping press with a blue lever above a blue card, beside a tray holding three cards.

The cost of leaving no capacity in reserve

A task can take minutes to complete and spend days waiting. When all capacity is already committed, everyday variation can turn an efficient schedule into a system that is slow to respond.

Alejandro Mayorga6 min read

A request arrives on Monday for half an hour of technical review. In this hypothetical scenario, the documentation is complete, the approval criteria are known, and the specialist has the authority to decide. She also supports several teams, and almost all her review time is already committed. Two requests arrive together; another needs more analysis than expected. The request gets a slot on Thursday and is completed in the estimated thirty minutes. No one lost it or forgot to respond. For almost its entire journey, the work was ready to move forward and had to wait.

In the specialist’s activity log, those days may look like time well spent. For the team that made the request, they are several days in which it cannot finish a deliverable. Both accounts can be true because they track different things: one records how time is occupied; the other, response time. Measuring only the thirty minutes of review would leave almost the entire elapsed time unexplained. The relevant question is why such a short task could not get access to the capacity it needed sooner.

Averages leave out what happens at the same time

The week’s workload can fit within the planned hours and still create waiting time. The total does not show when requests arrive or how long each takes. In our example, the problem arises when several demands converge on the same specialist review. A task that takes longer than expected delays the start of the next one; new requests arrive to find work still pending. Catching up requires a period in which more work can be completed than is coming in. If every future slot is already committed, catching up means moving one of those commitments.

Queueing theory helps clarify this relationship. The congestion analysis from MIT OpenCourseWare (opens in a new tab) considers variability both in arrivals and in the time needed to handle each task. In the models presented, under their assumptions of stable operation, waiting time grows nonlinearly as utilization approaches full capacity. A small increase in load near that limit can produce a much larger increase in waiting time. This relationship helps explain why average demand alone is insufficient to estimate response time. A full calendar does not demonstrate that level of utilization: scheduled commitments and actual use of capacity must be distinguished.

The model does not describe everything a person does. It isolates a mechanism in the flow of work. Applying it requires defining how much time is actually available for that review, after accounting for other responsibilities, and observing how requests vary. Counting the entire working day as if it could be devoted to a single activity would distort the starting point. Capacity planning needs to start with that actual availability; understanding the wait also requires showing how demand is distributed over time.

Before reserving capacity, understand what is waiting

Nor is it enough to find an empty slot in someone’s calendar. The relevant capacity has to be available at the point where the request needs it. A free afternoon on another team does not automatically substitute for the specialist’s judgment. Even her own spare capacity may be inaccessible if it is reserved for another priority. Operational availability includes these conditions of access: knowing how to carry out the review and being able to handle it within the required time are different commitments.

A person sands a wooden box with three other boxes lined up beside it on the workbench.

Leadership also decides how long the wait will be

Leaving capacity available has a visible cost. It means not committing all review time in advance, even when there is enough work to fill it. The alternative also has a cost: accepting that a new request will wait or that another will lose its place. In Monday’s scenario, declaring the request urgent could move it forward, but someone would have to decide which review gets pushed back. Changing the order of the queue may be necessary; it does not, by itself, create the missing spare capacity.

That decision belongs to those who can accept, limit, or change the commitments reaching that point. If several teams share the review function, they cannot each promise the same availability independently. Leadership needs to decide what response time it wants to protect and how much work it is willing to commit to in order to sustain it. Asking the specialist to find a slot after every slot has been assigned leaves that incompatibility unresolved and makes her negotiate it with each requester.

No single utilization rate is right for every kind of work. The reserve needed depends on variability, the alternatives for handling requests, and the cost of responding late; maintaining it also uses capacity that could serve other commitments. The decision requires observing real requests and reviewing the consequences. A reserve that never coincides with demand may be in the wrong place or at the wrong time. A reserve routinely absorbed by planned work raises the question of whether the committed workload really fits within the available time.

Protecting that reserve does not require someone to sit idle until a request arrives. It can accommodate work that can be postponed, provided this does not make the availability illusory or disregard the cost of interruption. What matters is that the reserve remains accessible when it is needed. If every unassigned slot is treated as waste and filled with another obligation, the organization ends up removing the capacity to respond that it had decided to retain.

Monday’s request lets us judge that choice without inventing a promise of improvement: we would still need to observe whether a well-placed reserve reduces the wait and what work is postponed to maintain it. But responsibility can already be made clear. Leaders who promise speed must account for the cost of making it possible. If they commit all available time without leaving room for variation, they must also take responsibility for any waiting time that decision may produce, even if it does not appear in the plan.