The Judgment Gap three reasons more reviewers will not close it › The headcount that would close it, and why you cannot buy it 12 steps, no labs
07 — part 3, three reasons more reviewers will not close itsourced

The headcount that would close it, and why you cannot buy it

Work out how many people it would take to close the gap by checking harder, and then notice which kind of people they would have to be.

Somebody is going to propose adding reviewers. It is the obvious move, it sounds responsible, and it is the first thing that will be suggested in the meeting where you raise any of this. So it is worth having done the sum before then.

the checkers you would need, and the checkers that exist work produced checkers it would take to keep up, which grows with it and every one of them has to know the domain well enough to recognize a wrong answer. which is the scarcest thing you have. the people who can actually do it this row does not grow, and it is not hired quickly. That is the arithmetic, and it is a sum about your own team.
The checkers you would need grows with production. The people who could actually be those checkers does not.
What moves: bars of work produced rise while rows of checker dots grow in proportion beneath them, beside a much smaller fixed group who know the domain well enough to judge.

The sum, on your own numbers

Take your ratio from step three. To close the gap by checking harder, examined has to rise to meet produced — so the checking capacity you need is your current capacity multiplied by that ratio. With a handful of people doing the examining now and a ratio in the low single figures, that is something between two and four times the reviewing capacity you havemodeled: your own ratio from step three, times your current examining capacity.

Review capacity is roughly linear in reviewers, and reviewers are not interchangeablecode review practice studiesunverified, which matters here because it means the capacity does not come from process improvements. It comes from people.

And then notice which people

A checker has to know the domain well enough to recognize a wrong answer. That is the same expensive knowledge from step two, which means the people who can do this job are exactly the people you already cannot get enough of — and they are currently doing the work you would be taking them away from.

where the capacity could come fromwhat it costs
hire more people who know the domainthe slowest hire you have, and you are competing for them with everybody
move existing senior people into checkingthe work they are doing now, which is the work you would be protecting
train people into itreal and slow, and it is the remedy in part four rather than this one
have people check outside their domainnothing, and it produces checking that cannot catch anything — which is step nine

The last row is the one that actually happens. Faced with the first three costs, an organization quietly takes the fourth: more people are assigned to review, they do not have the domain knowledge to judge what they are looking at, and the examined count goes up while the amount genuinely examined does not. The ratio improves on paper.

Every term this page uses

produced-to-examined
How much work your team produced, divided by how much of it somebody actually looked at.
examined
Looked at by somebody with enough context to have caught a problem. Not the same as having passed through a review step.
remainder
Work that was produced and not examined. It does not go anywhere — it merges into what everything after it is built on.
judgment
Deciding whether a piece of work is good, which needs knowing the domain well enough to recognize a wrong answer.
inspection
Checking finished work at the end, rather than changing how the work is made.
checkable
Able to be judged without understanding the whole of it. The property that decides how much needs judgment at all.
constraint
The thing that limits output. Adding more of anything else does not help until it moves.
rework
Work that had to be revisited after somebody called it done.
vigilance
Staying alert for a rare problem across a long stretch of time.

Taken as already known, and so not defined here: agent, model, review, throughput, headcount. That list is a claim about who is reading, and it is printed so it can be argued with.