Reading Variance Without Blame
How to run a variance review that produces corrections rather than defensiveness, which questions to ask, and the three questions that guarantee nobody tells you anything useful again.
A variance review either produces corrections or it produces defensiveness, and which one it produces is decided in the first two minutes by what the person running it asks.
The variance in “Reading Variance Without Blame” requires an unchanged plan and a trustworthy record of actual work. Teams exploring gdpr employee monitoring for gdpr employee monitoring can strengthen the actual side of that comparison while preserving corrections, category changes and human explanations.
The mechanics are easy. The difficulty is entirely in keeping the exercise about the account rather than about the people in the room.
For an outside reference relevant to “Reading Variance Without Blame”, consult PMI project-management library. Compare the general guidance with the organisation’s own unit, period and decision rights.
What the review is for
To find out which of the account's figures should change. That is the whole purpose, and stating it at the start of each review is not ceremony — it is the thing that keeps the conversation pointed at the numbers.
A review that produces no change to any figure and no decision to keep watching a line has not accomplished anything.
The question that works
One question, asked of each person in turn, does most of the work:
Which line in this account was wrong, and what should it say instead?
It is answerable, it is about the document, and anybody in the room can answer it. Most of the corrections that matter arrive in response to that question asked of four or five people in turn.
The three questions that do not
"Why did this overrun?" invites a defence. "Who was working on this?" invites a name. "Can we do better next time?" invites agreement and changes nothing.
Each produces an answer, none produces a correction, and a review made of these three teaches everybody that the honest answer is the expensive one.
Separating the estimate from the estimator
A bad estimate is information about a class of work. It is almost never information about the person who produced it, because the same person estimating the same class of work next time will be wrong in the same direction unless the class's figures change.
So the correction belongs to the class — "migrations run 40% over" — and the person is the witness rather than the subject.
Where variance is genuinely about behaviour
It happens: a commitment made without authority, a claim deliberately left open, hours recorded against the wrong line to protect a budget.
These exist and they are management matters, handled separately and privately. Mixing them into the variance review destroys the review, because everybody present now knows what the review is for.
The order that helps
Supply lines first, then the committed position, then the claims. By the time the conversation reaches individual claims, two or three corrections have already been made to figures that belong to nobody, and the register of the room is set.
Starting with the overrunning claim sets the opposite register and the rest of the review is spent recovering from it.
Writing it down in front of people
Each correction entered into the account as it is agreed, visible on the screen, with its date.
This is the single most effective thing available: people who can see their correction being recorded bring the next one, and people whose correction disappeared into somebody's notebook do not.
What to do with a line nobody can explain
Mark it as watched and move on. A variance with no available explanation is a reason to look again next period, not a reason to keep the room for another twenty minutes.
Two or three periods later it is either a pattern with a cause or it was noise, and both of those are answers.
The review that takes twenty minutes
Once the account is in good order and the closure discipline holds, a variance review is twenty minutes: four or five lines moved, two watched, one claim closed late with a note.
A review that routinely takes two hours is usually reviewing the bookkeeping rather than the plan, and the remedy is upstream of the meeting.
The record the review leaves
A short note per period: which lines moved, which are watched, which claims closed late. Four or five sentences.
Read in sequence, those notes are the most useful document the planning function produces, because they show what the organisation has learned about its own work rather than what it intended to do.
When the variance is the planner's fault
It frequently is: a shrinkage figure left stale, a commitment recorded without checking authority, a demand line carried at the wrong rate.
Saying so first, in the review, costs nothing and sets the terms for everybody else. A planner who has never been wrong in a variance review is not reviewing honestly.
Who should not be in the room
Anybody whose presence changes what gets said. That is a judgement rather than a rule, and it is worth making honestly: a review with the wrong attendance produces a polished document and no corrections.
The corrections can be collected separately and brought in. The document is what has to be right, not the attendance list.