Agency Capacity Planning: Match Committed Work to Real Hours

Every agency has had the same week. Two projects that were supposed to finish in sequence finish at the same time. A client who went quiet for three weeks comes back wanting to start Monday. Someone is on holiday and nobody noticed until the standup. The work does get delivered, because it always does, but it gets delivered by people working late, and nobody planned that.

Capacity planning is the discipline that makes that week rare instead of normal. This guide covers what it actually is, the three numbers it runs on, how to build the plan, and the specific mistakes that make most capacity plans useless within a month.

What capacity planning actually is

Capacity planning answers one question, asked every week: given what we have already promised, what can we take on?

That is narrower than it sounds, and the narrowness is the point. It is not resource management in the enterprise sense, and it is not a forecast of revenue. It is a comparison between two numbers over a rolling window - the hours you have, and the hours you owe.

It is also not the same as project scheduling. A schedule says when a project's phases happen. A capacity plan says whether your team can absorb those phases alongside everything else already in flight. Agencies that have a schedule and no capacity plan can tell you when each project is due and still be genuinely surprised when three of them collide.

The horizon that works for most agencies is four to eight weeks. Shorter than four and you cannot react - you find out you are overloaded in the week you are overloaded. Longer than eight and the plan is fiction: clients change scope, deals slip, people leave. Plan the next two months seriously and treat anything beyond that as a rough shape.

The three numbers

Almost every failed capacity plan fails on one of these three, and usually on the first.

1. Available hours, not headcount

The most common capacity planning error is multiplying people by 40.

A full-time person does not have 40 billable hours a week. They have 40 hours of existence at work, from which you must subtract everything that is real but not client work: internal meetings, standups, admin, tooling, code review, hiring, sick days, holiday, the pitch you are writing, the retrospective, the hour after a difficult client call where nobody gets anything done.

What is left is available hours. For most agency roles the honest number lands somewhere between 25 and 32 hours a week, and it varies by role - a senior person with management responsibility has materially fewer available hours than a junior specialist, which is exactly why "we have six people so we have 240 hours" produces plans that fail.

Work out your own number empirically rather than adopting someone else's. Take a month of tracked time, divide client hours by total hours worked, and use that. It will be lower than you expect. That is the finding, not an error in the method.

2. Committed hours, not estimated hours

The second number is what you have already promised. Not what you hope a project will take - what you have contractually committed to deliver, converted into hours.

The distinction matters because estimates are optimistic by default and commitments are not negotiable. If a project was estimated at 80 hours and is 60 hours in with half the scope remaining, its committed hours are not 20. They are whatever it will actually take, and if you plan against the estimate you have already lost the week.

Update committed hours from tracked reality, not from the original estimate. A capacity plan built on estimates that nobody revisits is a plan that gets more wrong every day it survives.

3. Billable utilization

Utilization is the share of available time spent on billable client work, and it is the metric that tells you whether the first two numbers are healthy.

It is worth being careful about which utilization you mean, because the term is used for at least two different ratios. Scoro's breakdown of billable utilization and Asana's guide to utilization rate both walk through the formulas and where teams get them confused - most usefully, the difference between billable hours over available hours and billable hours over total hours, which produce very different-looking numbers from the same timesheet.

The trap is treating utilization as a target to maximize. It is not. A team at very high sustained utilization has no slack, which means no capacity to absorb a scope change, no time for the internal work that keeps the agency competitive, and no margin before people burn out. Mosaic's collection of professional-services utilization benchmarks is useful context for where firms actually land rather than where they aspire to.

Read utilization as a diagnostic. Consistently low across the team means you have sold too little or your available-hours number is wrong. Consistently very high means you are one surprise away from a bad month. Wildly uneven between people means your allocation is broken, which is the most fixable of the three.

Building the plan

Step 1: Establish your real available hours

Pull a month of tracked time. For each role, calculate the ratio of client work to total hours worked. Use those ratios, not a uniform assumption.

If you are not tracking time consistently enough to do this, that is the first problem to solve - capacity planning without time data is guessing with extra steps. Our guide to time tracking for agencies covers getting to reliable data without turning it into surveillance.

Step 2: Convert every commitment into hours by week

Take each active and signed project and spread its remaining committed hours across the weeks it will consume, by role. Not "Project A: 120 hours" - "Project A: 12 design hours and 20 build hours in week 3."

This is the step people skip, and skipping it is why the plan does not work. A total that is fine across two months can hide a week where you need three designers and have one.

Step 3: Compare, weekly

Now the plan does its job. For each role, each week: available hours versus committed hours. Anywhere committed exceeds available is a decision you need to make now rather than a fire you fight later.

The decisions are always the same four, and it is worth naming them so the conversation is quick: move the work, add capacity (contractor or hire), reduce the scope, or move the deadline. There is no fifth option. "Everyone works harder" is how you discover the sixth, which is someone resigning.

Step 4: Include the pipeline, weighted

Signed work is not the whole picture. A deal that is 80% likely to close and starts in three weeks is real capacity pressure, and treating it as zero until the contract arrives is how agencies end up saying yes to something they cannot staff.

Weight pipeline work by probability and include it as a separate, visibly provisional layer. The point is not accuracy - it is that the plan shows you a week that is fine today and full the moment two proposals land.

Step 5: Re-run it every week

A capacity plan is not a document. It is a recurring thirty-minute meeting where you look at the next eight weeks, update what changed, and make the decisions the comparison surfaced. A plan built once and admired is worse than no plan, because it gives you confidence in numbers that stopped being true weeks ago.

Capacity planning for different agency shapes

The four-to-eight-week rolling model suits most agencies. Three variations are worth knowing.

Retainer-heavy agencies. A substantial proportion of capacity is committed before the month starts, which makes planning easier and creates a specific risk: reserved retainer hours get consumed by whichever project is loudest, and the retainer client - who is paying for responsiveness - gets the leftovers. Block retainer hours in the plan before anything else, and treat them as unavailable rather than as flexible.

Project-heavy agencies. Lumpier, and the plan needs to look further ahead at the pipeline, weighted by probability. The specific risk is the gap between projects, which is the largest single drain on utilization in most project businesses. Planning the next engagement's discovery to start before the current one finishes turns a gap into an overlap.

Agencies with many small clients. The plan is dominated by coordination overhead rather than by delivery hours. Ten clients at four hours a week is not forty hours of work - it is forty hours of work plus ten context switches, ten sets of updates and ten relationships. Build a per-client overhead into the model or the plan will consistently overstate capacity.

What to do when the plan says you are overloaded

The plan surfacing a problem is the plan working. Four responses, and there is no fifth.

Move the work. Push a start date, resequence phases, delay a non-urgent internal project. Cheapest option and the one most often available.

Add capacity. A contractor for the peak, covered in working with contractors. Fast, more expensive per hour, no commitment.

Reduce the scope. Deliver less, by agreement. Usually requires a client conversation and is frequently easier than expected - clients often have a view about what matters most that nobody has asked for.

Move the deadline. The most honest option and the one agencies avoid longest. Raised in week two it is a scheduling conversation; discovered in week seven it is a trust problem. Our guide to difficult client conversations covers the wording.

The option that is not on the list is "the team absorbs it." That is what happens by default when none of the four is chosen, and it is a decision made by omission with the cost paid by someone who was not consulted.

The mistakes that kill capacity plans

Planning at 100%. If every available hour is allocated, the first scope change breaks the plan. Leave deliberate slack - many agencies plan to somewhere around 80% of available hours and treat the remainder as the shock absorber it is. The slack is not waste; it is the thing that lets you absorb a change without a crisis.

Treating people as interchangeable. Six people with 30 available hours each is not 180 fungible hours. Your senior strategist cannot take the overflow front-end work. Plan by role, and by named person where the skill is genuinely scarce.

Forgetting the non-project work. Pitches, internal projects, onboarding a new hire, the rebrand you keep postponing. It is real work done by the same people, and a plan that ignores it will be wrong by exactly that much.

Letting the plan drift from tracked reality. If the plan says a project has 20 hours left and time tracking says it has burned through its budget, the plan is wrong and everything downstream of it is wrong too. This is the argument for capacity planning that reads from the same system as your time tracking rather than from a spreadsheet someone updates when they remember.

Confusing capacity with productivity. Capacity planning tells you whether the work fits. It does not tell you whether the work is being done well or efficiently. Using it as a performance metric is the fastest way to get people padding their estimates, at which point every number in the plan becomes fiction.

Capacity planning and hiring decisions

The plan's most valuable use beyond week-to-week scheduling.

Sustained overrun across a rolling eight-week view, with pipeline behind it, is the hiring signal. Not a busy fortnight, not a feeling. Three months of committed exceeding available, by role, with signed work continuing.

Which role to hire is visible in the same data. If design is consistently at 110% and development at 70%, the answer is not "a person" - it is a designer. Agencies frequently hire the role they enjoy hiring rather than the one the plan points at.

The plan also tells you when not to hire. Consistent overrun with no pipeline usually means estimates are wrong rather than capacity being short, and hiring against that multiplies the underlying error. Our guide to hiring your first employee covers separating a capacity constraint from a process one.

And it sizes the hire. A full-time delivery person adds roughly 1,000-1,100 billable hours a year. If the overrun is 400 hours annually, that is a contractor, not a hire.

How this connects to everything else

Capacity planning is the hinge between the parts of an agency that usually get managed separately.

It is how you prevent scope creep from being absorbed silently - when a client asks for "one more thing", the capacity plan is what turns that into a visible trade-off instead of a quiet extra evening. It is what makes managing multiple client projects a system rather than a juggling act. And it depends completely on reliable time tracking, because every number in it is derived from what actually happened rather than what was supposed to.

It is also the most honest sales tool an agency has. Being able to say "we can start you in three weeks, not next Monday, and here is why" is more credible than an unqualified yes, and it is very much better than a yes you cannot keep.

The weekly meeting that runs it

Capacity planning is not a document, it is a recurring half hour. What happens in it determines whether the model is useful or decorative.

Ten minutes: update what changed. Projects that finished, scope that grew, people unavailable, new work signed. This is the part that keeps the plan honest, and it is why the meeting has to be weekly - a plan updated monthly is describing a business that has moved on.

Ten minutes: look at the next eight weeks by role. Where does committed exceed available? Anywhere it does is a decision, not a discussion.

Ten minutes: make the decisions. Move work, add capacity, reduce scope, or move a date. Each one gets an owner and, where a client conversation is required, a date by which it happens.

Two rules make the meeting work. The same people attend every week - whoever can commit resource and whoever owns client dates. And the meeting ends with decisions rather than observations, because a capacity meeting that only surfaces problems is a worry session with a spreadsheet.

Making the plan visible

A capacity plan that lives in one person's spreadsheet solves half the problem. Making it visible to the team solves the other half, and it costs nothing.

People self-regulate when they can see the shape of the coming weeks. Someone who knows week three is heavy will not agree to an extra commitment in it. Someone who cannot see it will, entirely reasonably, and discover the collision later.

It makes the trade-off conversation concrete. When a new request arrives, "we're at 96% in the week that would land in" is a fact both sides can look at. "We're quite busy" is an opinion that invites negotiation.

It changes how the team talks about lateness. A visible plan showing a client dependency slipping by nine days moves the conversation from "we're behind" to "the input arrived late and here is the consequence" - which is accurate and considerably better for morale.

The practical form matters less than the visibility. A shared view of committed hours against available hours, per role, for the next eight weeks, updated weekly, is enough. Elaborate resourcing tools are optional; the weekly update is not.

Capacity planning and the sales conversation

The connection agencies most often miss, and the one with the largest effect.

A capacity plan is only useful if it influences what you sell. That means the plan has to be consulted before a date is promised, not after - which requires whoever is selling to look at it, and whoever is delivering to be asked.

Two habits make this work:

Delivery reviews every proposed timeline before it goes out. Ten minutes. It catches the dates that were optimistic and the ones that collide with something already committed. This is the single highest-return intervention available and it costs almost nothing.

Quote start dates from the plan, not from enthusiasm. "We could start you on the 14th" said because it sounds responsive, when the plan says the 28th, creates a problem that lands entirely on delivery. Being able to say "our next start date is the 28th" is also, incidentally, a stronger commercial signal than instant availability.

The failure mode this prevents is the most common cause of both overrun and burnout: dates agreed by people who were not going to deliver them, against capacity nobody checked.

Where to start

If you have never done this, do not build the full model. Do this instead, for four weeks:

1. Pick your five largest active projects.
2. For each, write down remaining committed hours by role.
3. Spread them across the next four weeks.
4. Write down each person's honest available hours - the number from tracked data, not 40.
5. Compare, once a week, for a month.

That is a capacity plan. It fits on one page and it will surface at least one collision you did not know about. Build the more elaborate version only once the simple one has proven it changes decisions - and if it does not change any decisions, the problem is not the model, it is that you were not going to act on it anyway.

The agencies that stay calm are not the ones with the most sophisticated planning. They are the ones who look at the same simple comparison every week and act on what it says.