The Capacity You Don’t Book Is Doing a Job

Consider two teams.

Both can comfortably complete forty units of work in a week.

The first is scheduled for thirty-nine.

The second is scheduled for thirty-two.

On Monday morning, both look very different.

The first appears efficient.

Almost every available hour has already been allocated.

The second appears to have eight units of unused capacity.

Then Tuesday happens.

A client changes something.

One task takes twice as long as expected.

An employee is absent for half a day.

A senior reviewer discovers an issue that needs correction before work can continue.

Something genuinely urgent arrives.

The first team has nowhere for any of it to go.

So it waits.

Or displaces something else.

Or creates overtime.

Or starts a chain of revised promises.

The second team has room.

Nothing magical happened.

It simply possessed something the first team did not:

the ability to recover.

That capacity may have looked unused on Monday.

By Friday, it may be the reason the week still worked.

Utilisation tells only part of the story

The attraction of utilisation is obvious.

Capacity costs money.

If people are available to work, a business understandably wants that capacity producing value.

In professional services, the logic becomes even stronger because time can often be connected directly to revenue.

So 70% looks worse than 80%.

80% looks worse than 90%.

And 100% can begin to look like theoretical perfection.

Every paid hour used.

No bench.

No idle time.

Maximum efficiency.

The difficulty is that most service work does not arrive or execute with perfect predictability.

Clients respond late.

Projects differ.

Tasks overrun.

Priorities change.

Questions appear.

Reviews uncover issues.

People get sick.

Work arrives in lumps rather than evenly spaced portions.

Queueing theory has studied this interaction between utilisation, variability and waiting for decades. As utilisation approaches the available capacity of a variable system, waiting does not increase proportionately; it can rise sharply because there is less spare capacity available to absorb fluctuations.

The mathematical insight is important.

But there is an operating implication underneath it that matters just as much.

The unused portion of capacity is not necessarily doing nothing.

It may be preventing variability from becoming delay.

Capacity has more than one job

Businesses often divide capacity into two categories.

Used.

Unused.

That is administratively convenient.

Operationally, it can be misleading.

Some capacity produces the planned work.

Some capacity absorbs the difference between the plan and reality.

Those are different functions.

Call the second one recovery capacity.

Recovery capacity is what allows an organisation to:

absorb a task that runs long;

deal with a client escalation;

correct work before it reaches the customer;

respond to an urgent opportunity;

cover an absence;

clear a temporary backlog;

help another function through a constraint;

or simply catch up after an unusually difficult day.

None of these activities necessarily appears in the original capacity plan.

That does not make them imaginary.

A team without recovery capacity has only one way to accommodate variation:

move the cost somewhere else.

Into waiting.

Overtime.

Reprioritisation.

Missed deadlines.

Context switching.

Customer disappointment.

Or tomorrow.

A calendar can be full while the system is under-capacitated

This produces a strange visual illusion.

The fuller the calendar, the healthier the resource plan appears.

Everyone has work.

There are no obvious gaps.

Nothing seems wasted.

Then delivery begins to feel slow.

Not because people are working slowly.

Because every new requirement joins a queue.

A ten-minute review waits until Thursday.

A small decision sits behind six larger ones.

A client clarification delays the project because the person who can answer it has no opening.

A piece of rework cannot begin because the team is already committed to everything else.

The amount of actual work required may be modest.

The elapsed time becomes large.

This is the difference between work time and flow time.

A task can require only thirty minutes of effort and still spend three days inside the organisation.

Most of those three days are not work.

They are waiting for capacity to become available.

The mistake is treating waiting as somebody else’s metric

Utilisation tends to be measured against the resource.

How busy is this person?

How many hours were billable?

How much of available capacity was allocated?

Waiting appears elsewhere.

Project lead time.

Customer turnaround.

Work in progress.

Missed dates.

Slow approvals.

Backlogs.

So one part of the organisation celebrates the utilisation figure while another absorbs the queue it created.

The metrics never meet.

That matters because the local optimisation looks excellent.

The person is fully utilised.

The team is fully booked.

The department has minimised idle time.

But the customer does not buy utilisation.

The customer experiences elapsed time.

The organisation ultimately earns from completed work, not from the aesthetic satisfaction of a full calendar.

Recovery capacity is not the same as a bench

This distinction is important.

The argument for recovery capacity can easily turn into a defence of permanent underutilisation.

That would be equally crude.

Capacity sitting unused because sales failed to generate enough demand is not automatically strategic slack.

Nor is poor scheduling.

Or mismatched skills.

Or a resource sitting idle for three weeks between projects because nobody saw the gap coming.

Those are different operating problems.

Bench capacity exists because there is currently no suitable work for the capacity.

Recovery capacity exists because the operating system deliberately avoids committing every usable unit before reality has had a chance to vary.

The numbers might occasionally look similar.

The operating intent is different.

One should often be reduced.

The other may need to be protected.

The more variable the work, the more valuable recovery becomes

Consider two environments.

The first performs highly standardised work.

Demand arrives predictably.

Tasks take roughly the same amount of time.

Few customer decisions interrupt execution.

Very little specialist judgment is required.

That system can operate at relatively high utilisation without becoming unstable.

Now consider a service team.

Projects differ.

Customers respond unpredictably.

Work depends on approvals.

Certain decisions require scarce specialists.

A small change can create several hours of additional effort.

Deadlines overlap.

The variability is fundamentally higher.

The same utilisation target does not mean the same thing in both systems.

Operations research captures precisely this interaction: waiting depends not only on how heavily capacity is loaded, but also on variation in arrivals and processing time. Kingman’s well-known approximation expresses waiting as a function of variability, utilisation and processing time rather than utilisation alone.

That makes “What utilisation should we target?” the wrong first question.

A better one is:

How variable is the work this capacity has to absorb?

Not every role needs the same reserve

This is where blanket utilisation targets become particularly dangerous.

An organisation may decide:

Everyone should be 85% utilised.

Why?

A production role doing repeatable work may tolerate that load easily.

A client-facing role receiving unpredictable requests may not.

A creative role may require time for iteration that cannot be forecast precisely.

A senior operator who reviews decisions across ten projects may become a system-wide constraint long before their own workload appears impossible.

The cost of fully loading that last person is especially easy to miss.

One senior reviewer has forty hours.

The resource plan allocates thirty-eight.

Excellent utilisation.

Then five projects require unexpected decisions.

The senior person does not necessarily work more slowly.

But every project now waits for access to the same two remaining hours.

That queue propagates.

The scarce resource has been optimised locally and made the wider system slower.

The most dangerous resource to fill is often the shared one

Some capacity sits inside one workflow.

Other capacity touches many.

That distinction matters.

A specialist.

A founder.

A commercial approver.

A creative director.

A senior engineer.

Finance during month-end.

A person whose judgment several projects depend on.

Loading one of these resources to the edge creates a different risk from fully loading someone whose work is relatively self-contained.

Because when shared capacity becomes constrained, the queue does not belong to one project.

It spreads.

Project A waits.

Project B waits.

Project C starts working around the delay.

Project D escalates because its client deadline is closer.

The resource now spends additional time reprioritising the queue.

The queue itself creates work.

This is why reserve should not necessarily be distributed equally.

The organisation may need more recovery capacity precisely where variability converges on a shared constraint.

Slack in the wrong place does not make the system faster

There is an important counterpoint.

Imagine Delivery is overloaded.

Marketing has spare capacity.

The company technically has slack.

That does not help.

Capacity is not perfectly interchangeable.

A copywriter cannot necessarily absorb financial approval work.

A junior employee cannot automatically perform a senior technical review.

Someone free next month cannot resolve a bottleneck today.

This is why aggregate resource numbers can create false comfort.

“We are only at 76% utilisation overall.”

Perhaps.

But the three people everything depends on may be at 98%.

The system experiences the constraint, not the average.

The more useful capacity question is therefore not:

How much spare capacity do we have?

It is:

Where does the system need room in order to recover?

The first thing recovery capacity buys is not leisure. It buys choice.

A fully allocated team has very few choices when reality changes.

Something must move.

Something must wait.

Someone must work longer.

Someone must disappoint somebody.

A team with reserve has options.

That difference has strategic value.

An attractive new customer enquiry appears.

Can the business respond?

A competitor creates an unexpected market opening.

Can someone investigate it?

A recurring delivery problem becomes visible.

Can the team stop long enough to fix the cause?

A manager notices an employee needs development.

Is there time to coach them?

An automation opportunity appears.

Can anyone build it?

These activities are often described as non-billable.

They may also be the work that improves future capacity.

A system permanently filled with today’s work can become unable to make tomorrow’s work easier.

Full utilisation can preserve the very problems consuming capacity

This creates a particularly nasty loop.

The team is overloaded.

Therefore nobody has time to improve the process.

Because the process does not improve, the team remains overloaded.

So utilisation stays high.

Management concludes more capacity is needed.

More people arrive.

The same operating problems spread across a larger team.

The organisation has added capacity without creating leverage.

Recovery capacity can interrupt that loop.

Not by giving people vague “thinking time.”

By deliberately creating enough room to remove recurring demand on the system.

Fix the handoff.

Clarify the decision right.

Automate the repeatable step.

Resolve the recurring exception.

Train somebody else to make the judgment.

Remove obsolete work.

The capacity used for improvement may temporarily lower utilisation.

If the intervention works, it increases effective capacity later.

Reserve should have an operating purpose

This is where the argument for slack can become lazy.

“Keep 20% free.”

Why 20%?

Free for what?

Across which roles?

Under what variability?

There is no universal percentage worth defending here.

The appropriate reserve depends on the nature of the work, arrival patterns, skill constraints, response expectations and consequences of delay.

Instead of setting an arbitrary slack target, define what the reserve needs to protect.

For example:

urgent customer response;

unplanned rework;

senior review;

project overruns;

business development;

process improvement;

coaching;

or predictable seasonal variation.

Now the unused space has a reason.

And if it is consistently not needed, the plan can tighten.

If it is consistently exceeded, the organisation has evidence that the underlying demand or variability requires attention.

Reserve becomes an operating decision rather than an ideological preference.

Look at what happens when the plan is wrong

Capacity planning is often evaluated at the moment the schedule is built.

Everything fits.

Everyone has work.

No obvious conflicts.

Plan approved.

A better test happens later.

When a task overruns, where does the extra work go?

When a client changes something, what moves?

When an employee is absent, who absorbs it?

When a senior review becomes necessary, how long does the work wait?

When an urgent request enters, what gets displaced?

When a project finishes early, can that capacity be redeployed quickly?

These questions reveal whether the capacity model can recover from reality.

A plan that works only when every estimate is correct is not a robust plan.

It is a prediction with no tolerance for being wrong.

Measure recovery, not only loading

Utilisation still matters.

A professional-services business cannot ignore how much expensive capacity is producing billable work.

But it should sit beside other questions.

How long does new work wait before it can start?

How quickly does a backlog clear?

How often does urgent work displace committed work?

How often do deadlines move because the required skill is unavailable?

Which resources repeatedly accumulate queues?

How much work is simultaneously in progress?

How frequently does planned utilisation turn into overtime?

These signals tell a different story.

They reveal whether high utilisation is being achieved without sacrificing flow.

That is the real operating trade-off.

A fast team needs somewhere for reality to land

The ambition is not low utilisation.

It is not empty calendars.

It is not permanent excess staffing.

It is a system capable of absorbing ordinary variation without turning every surprise into a queue.

That requires enough capacity producing today’s work.

Enough flexibility to handle deviations.

And enough room to improve how tomorrow’s work gets done.

The apparent inefficiency is that some capacity may not be allocated in advance.

The actual efficiency is what that reserve prevents.

Waiting.

Rework.

Overtime.

Constant reprioritisation.

Missed opportunities.

And the slow accumulation of unfinished work.

A business that books every available hour has removed one kind of waste.

It may also have removed its ability to recover.

So when an empty slot appears on the resource plan, the question should not automatically be:

How do we fill it?

Sometimes the more useful question is:

What is that space allowing the system to survive?