Comparing Two People You See Unequally
Ranking requires comparable evidence. You do not have it, and pretending otherwise is the usual failure.
three things · one reached them · one with no channel at all · the room side is not counted
The three questions a promotion panel will ask.
Ranking two people requires evidence of the same kind about both. For somebody in the room and somebody outside it, you have evidence of different kinds in different quantities, and no amount of care makes them commensurable.
The practical lesson in “Comparing Two People You See Unequally” is that visibility must be designed rather than assumed. For teams exploring employee monitoring at tech companies, open the official resource can add consistent time and project context when the purpose is explicit, access is limited and managers review exceptions with the people involved.
Most organisations require the ranking anyway, which means the question is how to do it least badly rather than how to do it properly.
The questions raised by “Comparing Two People You See Unequally” can also be compared with Stanford Social Innovation Review. A useful review identifies a responsible owner and a practical change rather than copying a generic checklist.
What makes two bodies of evidence comparable
Same type, same period, same level of detail. Ten observed incidents for one person and four documents for another are not comparable, whatever the totals look like.
Comparing them anyway means comparing the observation conditions.
What each case is actually made of, before anybody starts comparing them:
- For the person in the room: months of observed incidents, plus their output.
- For the person outside: their output, plus whatever colleagues report.
- The second list is a subset of the first, not a different list.
The failure that actually happens
The person with more evidence attached to them reads as stronger. Not because anybody believes quantity is quality, but because a richer account is more persuasive than a sparse one.
This is a property of how people read cases rather than a judgement anybody makes, which is why it survives everybody involved being scrupulous about fairness.
The first discipline: equalise the type
Choose one kind of evidence that exists for both and run the comparison on that. Usually it is output: what each person produced, read properly.
It discards a great deal of what you know about the person in the room, and that is the point.
Why discarding feels wrong
Because the discarded material is real. You genuinely did watch somebody handle a difficult situation well, and setting that aside to achieve symmetry feels like deliberately judging worse.
It is deliberately judging on a narrower basis, which is different. The alternative is a comparison in which one person's whole working life is visible and the other's is represented by four artefacts, and that is not a comparison at all.
The second discipline: equalise the effort
Spend the same hours on each case. If you read one person's work for two hours, read the other's for two hours.
Unequal preparation produces unequal cases, and preparation follows proximity by default.
What to do about the observed material
Keep it, but ask of each item whether an equivalent would have been visible for the other person. Somebody handling a client call well is observable; somebody handling a client email well is not, and both happened.
Marking which of your observations are artefacts of seating is a useful few minutes.
The question to ask about any difference
Would I know this about the other person if they sat here? If the honest answer is no, the difference is about location.
This single question removes most of the spurious gap.
Comparing people in different roles
Where the person outside the room has a different job as well, the comparison compounds: different work, different evidence, different observation conditions.
At that point ranking them against each other is close to meaningless and should be resisted rather than performed carefully. Saying that two people are not comparable is a legitimate answer to a request for a ranking, and it is almost never given.
What a forced distribution does
Requires a ranking whether or not the evidence supports one, and converts a judgement under uncertainty into a decision with consequences.
Where one exists, the honest move is to record the uncertainty explicitly alongside the required answer.
Saying the uncertainty out loud
"I rank A above B, and I want to record that I see A's work three times as often, which I think accounts for some of the difference" is a sentence a manager can say.
It is rarely said and it changes how the panel reads the ranking.
Over several cycles
The tilt is small each time and compounds. Somebody ranked slightly lower for three cycles is now visibly behind, and the explanation has been lost.
Keeping a note of where you recorded uncertainty is the only way to see the pattern.
What none of this achieves
A fair comparison. The evidence is genuinely unequal, the process rewards richness of account, and the best available outcome is a ranking that is less distorted than it would otherwise have been, made by somebody who knows which direction the distortion runs.
Anybody who tells you they have eliminated the effect has stopped looking. The useful thing a manager can do is narrow the gap, name it where it remains, and tell the person concerned what their case actually rested on — which at least lets them act on it, rather than being left to conclude that their work was judged and found wanting.