How to write a peer review that can be read aloud

Write your peer review as if the person will see all of it, word for word, because in most places they eventually do, in one form or another.

The form promises confidentiality. In practice that means your text gets lifted into a summary, and whoever writes the summary keeps your sentences whole, because rewording them is work and nobody has spare hours during a review cycle.

So the question worth asking before you submit is whether you would be fine standing next to every line. Honest is the lower bar. Plenty of honest sentences are things you would never say to someone's face, and those are the ones that do damage once the text starts moving.

Who actually reads a peer review

More people read your review than the form suggests, and the person you wrote about is often one of them.

The manager writing the summary reads it first. Then it travels.

• Whoever runs the calibration conversation, where lines get read out to a room.
• The person's next manager, when they change teams and someone opens the file.
• The person themselves, either directly or in a paraphrase close enough that they can hear you in it.
• Anyone reading back through the file if a decision about that person is ever questioned.

None of that is sinister. It is how documents behave.

The read-aloud test

Read your finished draft out loud as if the person is sitting opposite you, and rewrite anything you would not say in that chair.

One catch is the sentence you would never say to a face. Usually it is a diagnosis: checked out, not a team player, lacks ownership. You would not say those out loud because they are not things that happened. They are labels for a feeling, and nobody can act on a label.

The other is a true sentence written at the wrong temperature. "Handoffs arrive with no context and I lose most of a morning rebuilding them" is easy to say out loud. Typed into a box late at night it tends to come out as "careless", which is the same information with the evidence removed and an insult stapled on.

The test does not mean make it kinder. Plenty of hard sentences pass. "Dates move and I hear about it afterwards" is specific, sayable, and survives being read back to me.

Most peer reviews fail by being pleasant and empty

Empty praise wrecks more reviews than harshness does. That has held for every review set I have ever read.

"Great teammate, always willing to help" is the standard output. It costs nothing to type and gives the reader nothing to hold. In a room where people are being compared, it works as a shrug with a smile on it.

I think vague praise is a small theft. You had specific information about somebody's work, you were asked for it, and you handed back a greeting card.

Praise that survives the trip has a scene in it:

• One thing they actually did, described so someone who was not there can picture it.
• What it changed for the work, or for you, in plain terms.
• What it cost them, if it cost them something like a weekend or an awkward conversation.
• Whether this is a pattern or the best week they have had.

Four sentences of that beat a paragraph of adjectives, and they are the lines that get quoted when someone argues for that person in a room they are not standing in.

Writing the part that stings

Write criticism as behavior, effect, and the thing you would want instead, and keep it to one or two items.

Behavior means something that happened, not a trait. Traits cannot be fixed, so naming one is just a verdict on a human being. "Specs arrive with the acceptance criteria missing" points at something repairable. "Sloppy" points at a person.

Effect means what it cost the work. Leave it out and your note reads as a preference, and preferences get filed under personality clash.

The thing you want instead is the part most people skip, and it is the part that makes the note usable. If you cannot describe what better would look like, you may be annoyed rather than in possession of feedback. The same shape works on a draft or a design, for the same reason: the reader needs a target, not a mood.

I would rather read a review with one honest item in it than a tidy list where nothing carries evidence.

If your read of the person runs against the general opinion, say so plainly and give your reason. Disagreeing in writing works the same here as anywhere else: evidence first, verdict second.

There is one rule I will not bend. Nothing goes in a review that the person has not already heard from you. A review is a terrible place for first-time news: they cannot ask a question, cannot fix it in time for it to count, and now they know you were carrying it quietly while being friendly in the hallway. If you have not raised it, raise it first and then write it, or leave it out and accept that you sat on it.

Keep out anything you did not see yourself

Leave out anything you cannot tie to work you personally saw.

• Secondhand reports, which are gossip with your name attached to them.
• Personality, mood, and anything from their life outside work.
• Comparisons with other people, which read as campaigning even when you meant nothing by it.
• Opinions about pay or level dressed up as observations about the work.
• Old material that was raised, fixed, and is being brought back to add weight.

That last one is the meanest habit in review writing and it is everywhere. A problem someone fixed is evidence that they respond to feedback. Reusing it as ammunition teaches them that nothing they repair ever gets closed.

Anonymity is thinner than you think

Assume you are identifiable, because with a handful of reviewers and your own way of writing, you usually are.

Three or four people wrote about this person. One of them mentions the migration you both worked on. Your sentence rhythm is your sentence rhythm. The subject does the arithmetic in about a minute, and they are mostly right.

I think anonymous feedback boxes make writing worse rather than braver. Cover is not safety, and the belief in cover is what produces the line people later regret: sharper than they would ever be in person, vaguer than anything that could be checked.

Write it signed even when the form does not ask for a name. If a sentence only works while you are hidden, it is a complaint in a costume.

The sentence you keep deleting

The line you type and delete three times is usually the one worth submitting.

You know the one. It is the real observation, the thing the whole review has been circling, and you keep pulling it because it looks harsh sitting in a plain box.

The wording is the fixable problem. The observation is the point. Write it, then fix the temperature: swap the trait for the event, add what it cost, add what you would rather see, and delete the two hedges you added so you could bear to leave it in.

Hedging is what gets people into trouble here. Not cruelty. Fog. A vague negative sentence is the worst thing you can put in a review, because the reader quietly fills in something specific and usually worse than what you meant, and you never get to correct it, since you are not in the room when it is read.

Mine come out foggy on the first go, then overcorrect into blunt on the second. The version I would sign sits somewhere past both. A review box makes me nervous, and nervous is when the distance between what I mean and what my hands actually type is widest, so I run the draft through Unvent in the box and take the wording that lands between my two bad drafts.

One habit worth stealing. Keep a running note on the people you work closely with, and add a line the day something happens, good or bad. Then the review is transcription instead of invention, and the person who quietly did the hard part gets the credit my memory would otherwise hand to whoever was loudest last.