Planning Capacity From Time Data

Capacity planning is usually done from headcount multiplied by contracted hours, which overstates available effort by a wide margin. Actual data does better.

Why the theoretical number is wrong

Contracted hours are not available hours. Leave, sickness, training, internal meetings, administration and recruitment all consume time before any project work starts.

The overhead is larger than people assume and varies by role: senior people spend more time on supervision and less on delivery, which theoretical models rarely capture.

Effort is not evenly available. Someone on three projects is not a third available to each.

Building the realistic figure

Take actual project effort per person per month, over a year.

Express it as a proportion of contracted hours. That proportion is your planning factor and it is usually well under what anyone expected.

Segment by role and seniority, because they differ substantially.

Use the median and the spread, not the mean, since a few unusual months distort it.

Re-derive annually, because it drifts with the mix of work.

Using it

Planned work is committed against realistic capacity, not theoretical.

New work is assessed against what is actually free, which is much less than a resourcing spreadsheet shows.

The difference between the two figures is the honest conversation an organisation needs to have about why everything runs late.

Where the gap is large, that is the finding: either the overhead is excessive and should be reduced, or the commitments are excessive and should be fewer.

What historical data cannot tell you

What a new person will deliver, which is lower for a period and the ramp is worth modelling explicitly.

What unfamiliar work will take, which returns to the estimate factor.

Whether the team is sustainable at the observed level. Utilisation that was achieved may have been achieved by overwork, and the hours-beyond-contracted measure is what distinguishes the two.

Whether people will still be there. A capacity plan built on a level that is producing attrition is planning with a number about to change.

The allocation trap

Assigning someone across five projects at twenty percent each is arithmetic rather than planning.

Switching costs are real and appear in the data as effort that fits no project neatly.

Effort concentrates: people work on what is urgent, not on the allocation.

Plan fewer concurrent commitments per person and measure whether elapsed time improves, which it usually does without any increase in effort.

What to report

Realistic capacity by team, derived from actuals, with the factor stated.

Committed work against that capacity.

Concurrent commitments per person, median.

Ramp position of anyone new.

Hours beyond contracted, which is the check that the capacity figure was achieved sustainably rather than borrowed from people's evenings.

Deriving the factor

One query that replaces a theoretical capacity model.

Actual project effort per person per month, over a year.

As a proportion of contracted hours.

Median and spread, by role and seniority.

That proportion is your planning factor, and it is usually well below what any resourcing spreadsheet assumes.

Re-derive annually, because it drifts with the work mix.

The gap between this and the theoretical number is the honest conversation about why commitments overrun.

Concurrent commitments

The allocation variable with the largest effect on elapsed time and the least attention.

Assigning someone across five projects at twenty percent is arithmetic, not planning.

Switching costs are real and show up as effort that fits no project cleanly.

Effort concentrates on whatever is urgent, so the allocation is fiction within a week.

Measure the median and maximum concurrent commitments per person.

Reduce them and measure elapsed time, which usually improves with no increase in total effort.

Connect policy to configuration

The practical choices behind this note can be compared with the resource-planning example. Keep the organisation's written purpose in control of the setup, enable only the data needed for that purpose and review the result with affected users.