Variance That Is Really a Claim Nobody Closed
Why a large share of apparent overrun is bookkeeping rather than delivery, the three patterns that reveal it, and the close that prevents it.
A meaningful share of what looks like overrun is a claim that should have been closed. The work finished, the claim stayed open, and the hours that went somewhere else were recorded against it because it was the open row.
The commitments described in “Variance That Is Really a Claim Nobody Closed” become easier to reconcile when planned work can be compared with a consistent record of project time. A team reviewing view the current reporting information for stealth computer monitoring software can add current operational evidence while keeping the claimant, approval state and decision owner explicit.
That is a bookkeeping artefact, and it is indistinguishable from a real overrun unless somebody looks.
Additional context for “Variance That Is Really a Claim Nobody Closed” is available from INFORMS operations-research resources. Use external material to prompt a specific validation check rather than replace local judgement.
The three patterns
A claim that consumes hours after its deliverable was accepted. A claim whose hours continue at a steady low rate forever. And a claim that absorbs unrelated work because it is the only open row with budget in it.
Each produces an overrun in the variance and none of them is the work being harder than estimated.
The steady trickle
The most recognisable: a claim that was supposed to finish in March consumes four hours a week indefinitely. That is support, and it has become support without anybody deciding.
The remedy is to close the claim and open a standing commitment with an owner and a size, which is both accurate and much easier to question later.
The absorbing claim
Where several pieces of work are in flight and only one has an open line, the hours go to the open line. Nobody is being dishonest; there is nowhere else to put them.
This shows up as one claim that is wildly over and several that came in exactly on estimate, which is an implausible pattern and a reliable signal.
Telling it from a real overrun
Three questions. Was the deliverable accepted before the extra hours were spent? Are the extra hours steady rather than concentrated? And does any other work in the period lack a claim to be recorded against?
- Was the deliverable accepted before the extra hours were spent?
- Are the extra hours steady over months rather than concentrated in a push?
- Is there other work in the period with no claim of its own to be recorded against?
Any yes points at bookkeeping rather than estimating, and the remedy is a closure rather than an estimating conversation.
Why it matters
Because the remedies are completely different. A real overrun is an estimating signal; a bookkeeping overrun is a closure failure. Correcting estimates to fix a closure failure inflates every future estimate and fixes nothing.
It also matters to the claimant, who is otherwise carrying a reputation for overrunning that the data does not support.
The close that prevents it
At every period close: for each claim, has its deliverable been accepted? If yes, close it and open whatever remains as a new line with its own name.
Five minutes. It is the same closure step described earlier in this collection, and this is the second of its two payoffs.
The claim that should have been several
Sometimes the opposite: one claim covering three distinguishable pieces of work, all of which overrun or not independently.
Splitting it at the point of entry is better, and splitting it retrospectively at the close is acceptable where the hours can be reasonably attributed. Both beat carrying one large opaque row.
What this does to the estimating series
Cleans it. An estimating record that contains three closure failures per year is noise-dominated, and the conclusion drawn from it — that estimates are unreliable — is itself unreliable.
Removing the bookkeeping artefacts usually reveals that estimates are better than anybody thought, with one or two genuine and specific exceptions.
Who should look
The account owner, at the close, with the claimant's confirmation. Neither alone: the owner sees the shape, the claimant knows what happened.
Two minutes a claim, and the output is a variance record that describes delivery rather than filing.
Closure is not a judgement on the work
Worth saying out loud, because the first few closures feel like an accusation. Closing a claim says the deliverable was accepted; it says nothing about whether the estimate was good or the work was well done.
Where that distinction is not made explicit, people start keeping claims open defensively, which is the behaviour the exercise was meant to remove.
The standing commitment that results
Most closure failures convert into a standing commitment: the support that the finished work now generates, sized and owned.
That is a better description of reality than an open project claim, and it has the useful property of appearing on the demand side where it can be questioned, reduced, or paid for.
What the record looks like afterwards
Fewer rows, each with a shorter life, and a visible support line that grows as the estimate of finished work accumulates.
A plan with that shape is harder to argue with, because the growth in the support line is the arithmetic consequence of the delivery everybody was pleased about.
The annual finding
In most first years of doing this, between a quarter and a half of recorded overrun turns out to be closure failure.
That is a large correction to the organisation's view of itself, and it arrives from five minutes of bookkeeping per period rather than from any improvement in how anybody estimates.