How to Build a Weekly Capacity Plan That Survives Real Client Work

Every resource manager has the same Monday. The plan you left on Friday no longer describes reality. A client pushed a milestone, a hire delayed their start, someone took two weeks of PTO you did not know about, and a partner promised a discovery workshop that needs three people you do not have. The spreadsheet is out of date before your coffee is cold.

A weekly capacity plan is supposed to absorb that motion. Most cannot, because they are built for a static world. This is the version that survives real client work.

What is a weekly capacity plan and why does it fail?

A weekly capacity plan is a rolling forward view of every billable person on your firm, their percentage of allocation per week for the next 12 weeks, and the projects those allocations belong to. In theory it lets you see bench, overbooking, and pipeline risk in one grid. In practice, three failure modes account for almost every dead capacity plan.

• Ownership drift. Delivery leads all edit their own rows, the resource manager reconciles at month-end, and no version of the plan is ever fully current.
• Wrong granularity. The plan sums to the month, so a two-week overallocation followed by two weeks of bench averages to 100% and hides both problems.
• No decision loop. The plan shows red cells, but nobody owes anyone a decision by a specific day, so the red persists for weeks.

If any of those is true, the grid becomes decorative. You are back to running staffing calls off a delivery lead's memory.

What data does the capacity plan actually need?

The minimum useful row is 10 columns. Anything more clutters the grid, anything less strips out the decision.

The column most teams miss is PTO known. Without it, the plan looks fine right up until someone takes a week off that the resource manager did not have on the grid.

How do you set the planning cadence?

Weekly, with named ownership on named days. Cadence is what turns a static grid into an operating rhythm.

• Monday, 9 to 10am. Resource manager reviews the current week and the next four. Any red cell, over or under, needs a decision by Wednesday.
• Wednesday, end of day. Delivery leads confirm scope changes and staffing requests for the following week.
• Thursday. Resource manager rebalances the grid. Anyone changing rows on Friday is late.
• Friday, close of business. The plan freezes for the following week. Changes after that require an explicit re-plan.

That cadence exists so the plan is trusted for one week at a time. If someone shifts an allocation Friday afternoon and the delivery lead only sees it Tuesday, the plan has already lost.

What is the right allocation granularity?

Weekly percentages. Not daily hours. Not monthly averages.

Daily hours look precise, but they create noise. Whether someone billed 6.2 or 7.4 hours on Tuesday is a time tracking question, not a capacity question. The capacity plan cares about whether the week can absorb a new engagement.

Monthly averages are worse in the other direction. They smooth over the two overallocated weeks in the middle by pulling capacity from the bench weeks on either side, so the plan says 95% when reality was 130% then 60%. Both weeks were problems.

A weekly percentage, floored at 5% increments, is honest without being noisy. It lets the plan say "Priya is at 90% next week, 100% the week after, and 60% after that," which is what a delivery lead needs to plan.

How do you handle scope changes without redoing the whole plan?

Have one clear rule: allocations do not move without a scope statement. Anything else and the grid becomes a wishlist.

• Small change, under 20%. Delivery lead updates the allocation with a one-line note in the row. Resource manager reviews at the next Monday standup.
• Medium change, 20 to 40%. Delivery lead files a re-plan request by Wednesday. Resource manager rebalances with the affected people copied.
• Large change, above 40% or a new engagement. Full staffing request, run through the normal intake, with a decision by the following Thursday.

Without this ladder, every scope wobble triggers a full replan, and the plan drowns in edits. With it, the noise stays local and the grid stays trustworthy.

How do you turn allocations into a forecast?

Every allocation row already has a person, a percentage, a week, a bill rate, and a project. Multiply and sum, and the plan is a revenue forecast.

• Planned hours per week per person equals target weekly hours times allocation percentage.
• Planned revenue equals planned hours times bill rate.
• Bench revenue lost equals hours below the target utilization multiplied by the person's bill rate.

The forecast is only as accurate as the plan. But that is the point: if two functions are looking at the same grid, resourcing and finance stop maintaining two conflicting numbers. When a partner asks how a delayed close changes Q4 revenue, the answer takes 10 minutes, not two days.

What does a healthy capacity plan look like?

Three quick tests. If any fail, the plan is decaying.

1. Reality check. Pick five people at random. Ask them what they will be working on the week of two Mondays from now. Match against the plan. Fewer than four matches means the plan is stale.
2. Bench check. The plan should show a small, believable amount of bench, usually 8 to 15% below target. Zero bench means the plan is optimistic. Above 20% means the pipeline has a problem the plan is quietly warning you about.
3. Decision check. Every red cell more than two weeks old is either a decision nobody made or a decision made verbally that nobody wrote down. Both are corrosive.

A healthy plan is boring. It updates weekly, shows a manageable amount of red that resolves within two review cycles, and nobody argues about whose version is current.

The mistake to avoid

Most services firms treat the capacity plan as a snapshot they refresh when someone asks. That framing guarantees decay. Treat it as an operating rhythm instead: one resource manager owns the grid, delivery leads owe updates by named days, allocations move only with a scope statement, and every red cell earns a decision within one review cycle. Everything else, including which tool holds the grid, is a detail.