Plan Against Actual
How to compare a capacity plan with what happened, the three kinds of difference it produces, and why most variance analysis measures the wrong thing.
One quarter's plan against what the hours actually did
What the account published in January.
Two unplanned absences and one late start.
Against 1,840 planned: a 370-hour shortfall.
Three claims that arrived mid-quarter.
The allowance was 390 h; actual was 511 h.
Nothing left; the quarter closed exactly flat.
The plan was wrong in three separate places and the errors were different in kind: supply fell, unplanned work arrived, and the support allowance was too low. Only the third is a planning error; the other two are events. This is one employer's own quarter, not a benchmark.
At the end of a period, the plan and the actual can be compared. Almost nobody does it, and the few who do usually compare the wrong things: delivery against promise rather than hours against hours.
The variance in “Plan Against Actual” requires an unchanged plan and a trustworthy record of actual work. Teams exploring open the provider’s published overview for mouse jiggler detection can strengthen the actual side of that comparison while preserving corrections, category changes and human explanations.
A capacity variance is about hours. Whether the work delivered is a different question with different owners, and mixing the two produces an analysis that answers neither.
As a separate benchmark for “Plan Against Actual”, see Lean Enterprise Institute guidance. Keep source data, calculation, override and decision visible as distinct steps.
The three differences
Supply varied: the hours available were not the hours planned. Demand varied: the claims consumed more or less than their estimate. And composition varied: hours went to work that was not in the plan at all.
Three numbers, three causes, three different remedies. Collapsing them into one variance figure makes the remedy unknowable.
Supply variance
Unplanned absence, a late start, somebody leaving, an event nobody booked. These are events rather than errors, and the right response is usually to check whether the rate used was reasonable rather than to change anything.
A supply variance that is systematically negative, quarter after quarter, is not events: it is a shrinkage figure that is too low.
Demand variance
Claims that consumed more than their estimate. This is the one everybody measures and it is the hardest to interpret, because an overrun can mean a poor estimate, a scope change or work that was harder than anybody knew.
Recording which, in one word, at closure, is what makes the series useful. Without it the organisation learns only that estimates are wrong, which it already knew.
Composition variance
Hours spent on work that was not in the book. This is the most diagnostic of the three and the least measured.
A large composition variance means the intake process is being bypassed, and no amount of better estimating will fix it. It is also the variance most likely to be invisible, because unplanned work leaves no trace in a plan that never contained it.
Measuring the actual without timesheets
Timesheets are the obvious instrument and they are expensive, unpopular and frequently inaccurate. For capacity variance, a lighter method works: at the close, ask each claim's owner roughly how much went into it, in days, and reconcile against known availability.
Rough is adequate. The decision the variance informs is a change to a rate or an allowance, and those are not sensitive to a day either way.
The reconciliation that must close
Available hours should equal hours accounted for: planned claims, plus unplanned work, plus support, plus whatever is left. If the account does not close, something substantial is being spent on work nobody has named.
A persistent unaccounted balance of ten per cent or more is itself the finding, and it is usually either interrupts or a standing commitment nobody has written down.
What not to do with the variance
Not publish it by person. A capacity variance is a property of a plan and a period, and attributing it to individuals converts the whole account into a performance instrument, after which the numbers stop being true.
Team level, claim level, period level. Never name level.
How often
Every period, briefly, with one page a year that looks at the series. The single-period variance is noisy; the series is where the systematic errors live, and those are the only ones worth changing a rate for.
What the first year teaches
Usually three things: that the support allowance is too low, that one or two claimants systematically underestimate, and that between five and fifteen per cent of capacity goes to work that never entered the book.
All three are fixable, and none of them is discoverable without the comparison.
The six numbers, as a row
| Planned available | Actual available | Planned on claims | Actual on claims | Unplanned | Unaccounted |
|---|---|---|---|---|---|
| 2,061 | 1,881 | 1,840 | 1,470 | 290 | 121 |
One row per period. After four periods the columns start telling different stories, and the column that moves most is where the next correction belongs.
Asking for the actuals without a timesheet system
A single message at the close: "roughly how many days went into your claim this period — nearest five days is fine."
Most claimants answer within a day and the aggregate is good enough for every decision the variance informs. Precision here buys nothing and costs the cooperation that makes the whole exercise possible.
The line to record each period
Planned available, actual available, planned claims, actual on claims, unplanned work, unaccounted. Six numbers.
Over a year they produce the only honest picture of how an organisation actually spends its capacity, which is a different document from the one describing how it intended to.