Capacity planning for agencies and consultancies

Most capacity plans fail for the same reason: the demand side is guessed. Supply is easy — you know who works when. Demand is the hard half, and a spreadsheet filled in from memory ages badly within a fortnight.

Derive demand, don't type it

Demand should come from the work that already exists: assignments on live projects and tasks with estimated effort. Derived demand has a useful property — when the plan changes, the capacity picture changes with it, without anyone maintaining a second version of the truth.

This is why rough estimates beat no estimates. An imperfect number that updates automatically is more useful than a precise number nobody revisits.

Model supply honestly

• Use each person's actual working pattern rather than a flat full-time assumption.
• Account for the non-project time that genuinely exists — internal work, management, leave.
• Treat roles and skills as constraints: a free person who cannot do the work is not capacity.

Keep tentative work separate

Pipeline work is real information and terrible commitment. The answer is not to exclude it or to treat it as booked, but to hold it in its own layer that can be turned on and off.

With it off you see what you have promised. With it on you see the shape of the next few weeks if expected work lands. The difference between those two views is the conversation worth having with the people selling.

Act on the overlap, not the average

Utilisation averages hide the problem. A team at seventy per cent for the quarter can still be badly over-allocated in the week that matters. Look for the overlaps: the fortnight where three projects need the same person, or the phase that assumes someone who is finishing elsewhere.

Once an overlap is visible, the options are ordinary — move a phase, split the work, change the assignment, or tell the client early. All of them are cheap a month out and expensive on the day.

A simple weekly rhythm

1. Check the next four to six weeks for over-allocation, not the whole quarter.
2. Confirm the plans creating that demand are still current.
3. Toggle pipeline demand on and note what would break if it lands.
4. Resolve the nearest conflict properly rather than all of them roughly.