Capacity planning sounds like something that starts at fifty people. In practice the agencies that need it most are the ones at eight to twenty, where every allocation decision is made in someone's head and the first sign of a problem is a person quietly working weekends.
You do not need software or a resourcing manager to fix this. You need one number per person per week and the discipline to look at it before saying yes.
Start with real capacity, not contracted hours
The most common mistake is planning against a 40-hour week. Nobody delivers 40 hours of client work.
Take contracted hours, then subtract:
- Holiday and public holidays, spread across the year
- An allowance for sickness
- The non-billable time your agency genuinely requires — internal meetings, admin, line management, new business
At a healthy 65–80% utilization, planning capacity for a full-time person is roughly 26–30 hours a week, not 40.
Plan against 40 and you will be over-committed permanently while your plan insists everything is fine. This single correction fixes more resourcing problems than any tool.
Plan in weeks, not days
Day-level resourcing is precise, expensive to maintain, and wrong by Tuesday. Week-level is coarse enough to survive reality and accurate enough to answer the question you actually have: can we take this on.
One row per person, one column per week, hours committed in each cell. Eight weeks ahead is plenty for most agencies; twelve if you sell long projects.
Commit less than capacity
The number that matters is what percentage of capacity you allow yourself to commit in advance.
Plan to 80–85% of real capacity. The remainder is for the things that reliably happen: work overrunning, a client bringing a deadline forward, someone being ill, the urgent request from your best account that you are not going to refuse.
An agency planned to 100% has no absorption. The first overrun does not cause a delay to that project — it cascades, because the only way to recover is to take time from whatever was next.
Use actuals, not estimates, for the numbers
This is where time tracking earns its place in resourcing.
When you allocate 30 hours to a project, that number came from an estimate. Estimates are systematically optimistic. If your historical data says this kind of project runs 40% over — and for most agencies it does, consistently, by a stable factor — then plan the real number.
Keep estimated against actual hours by project type. After a year you will know your own multiplier, and capacity planning stops being a wish.
The weekly rhythm
Fifteen minutes, same slot every week, whoever makes commitments:
- What changed last week? Overruns, delays, work that finished early.
- Is anyone over 100% in the next two weeks? Fix it now, while there are options.
- Is anyone under 50% for more than two weeks? That is a pipeline or assignment problem and it is cheaper to spot early.
- What is landing that we have not allocated? Signed work that is not on the plan is the most common cause of a bad month.
The three signals to act on
Someone consistently above 90%. Not a hero, a risk. Sustained overload is the most reliable predictor of resignation, and replacing an experienced person costs more than the margin their overwork produced. See why high utilization is not a win.
A whole team above 85% for a month. You have taken on more than you can deliver. Either something moves or something slips, and choosing which is considerably better than finding out.
Bench time appearing three weeks out. The most actionable signal in the whole exercise, because three weeks is enough notice to do something about it — chase a proposal, bring forward internal work, start the conversation with a contractor. Discovering it on the Monday is not.
When to get proper tooling
A spreadsheet stops working somewhere around twenty people, or sooner if projects routinely span several people and change weekly. The tells are that nobody trusts the sheet, that it is updated once a fortnight, or that two people maintain different versions.
At that point the thing to buy is something that reads your actual tracked hours rather than asking you to re-enter them, because a resourcing plan that has to be maintained by hand alongside a timesheet is a plan that goes stale. Until then, the spreadsheet is genuinely fine — the discipline is what produces the value, not the tool.
