Useful feedback closes the distance between the draft in front of you and the piece it is trying to be. Vague reactions do not close it. A comment earns its place when it names where the problem sits, what it costs the reader, how serious it is, and what the writer must now decide.
- Read the full draft before writing your first comment.
- Declare which editing level you are working at and stay there.
- Give every comment a location, a problem, a reader consequence, and a severity.
- Ask what a section is for before rewriting it for the author.
- Say each objection once, in one thread, and then let the writer decide.
Read the whole thing before you write anything
Comment as you read and you will describe symptoms in the order you met them rather than in order of importance. The same discipline applies when you are on the receiving end, which working with an editor covers. A confusion in the second paragraph is often answered on the third page. Read once from end to end, then decide what needs saying.
After that first read, write two sentences for yourself: what the piece is trying to do, and the largest thing standing in its way. Every comment you leave should trace back to one of them. The ones that do not trace back can be dropped.
Name the level you are editing at
Feedback lands badly when the writer expected one kind and received another, which is why comments that stay attached to the draft beat a thread of loose notes. Say at the top whether you are responding to the argument, the structure, the paragraph craft, or the copy. Then hold your comments at that level, even when a stray comma catches your eye.
Mixing levels is the most reliable way to make a review exhausting. Ten notes about punctuation buried among three structural objections make the objections easy to miss, and the volume signals that the draft is closer to finished than it really is.
Name the problem precisely and you have already done most of the writer's diagnostic work.
Describe the experience, then the cause
Begin with what happened to you as a reader: where you slowed down, stopped believing a claim, lost the thread, or skipped ahead looking for the point. That part of the comment is not arguable, because you are the evidence for it.
Then offer your diagnosis, and mark it as a diagnosis. Saying that you lost the thread, and that you suspect it is because the example arrives before the claim it supports, gives the writer a symptom and a hypothesis they remain free to reject.
Mark severity so the writer can triage
Use three levels and use them the same way every time. Blocking means the piece should not publish in this state. Important means the change would make it meaningfully better. Optional means you would have done it differently and the writer may ignore you.
Without severity the writer treats everything as equal, then either does all of it or resents all of it. Most reviews should carry very few blocking comments. If half of yours are blocking, the draft needed a conversation about its premise, not margin notes.
Ask before you prescribe
When you cannot tell what a section is for, ask. The question of what it is meant to establish beats a rewrite that quietly assumes an answer. The writer often knows something you do not: a constraint, or a promise made earlier.
If you do supply replacement prose, label it as an illustration of the fix rather than the fix itself. The writer should be able to solve the same problem in their own rhythm. Prose handed over without a reason gets pasted in, and the piece loses its author.
Write comments a revision can be built from
A useful comment holds four things: the location, the problem, the consequence for the reader, and the priority. It does not need a solution. Supplying one is optional generosity, but a comment missing the consequence is only a reaction with a page number attached.
Keep each comment to one problem. Bundled comments get half-answered, because the writer fixes the part they agree with and resolves the thread with the rest still inside it. One thread should carry one decision, which also makes the round easy to close.
Handle disagreement without a stalemate
If the writer declines a change, work out whether you disagree about the goal or about the means. Disagreement about means is theirs to settle, and pressing it further is a preference dressed as a standard. Disagreement about the goal goes back to the brief.
Say each objection once, clearly, in writing. Repeating it across three threads does not make it stronger. It makes the review adversarial and pushes the writer into defending the draft rather than improving it, which is the opposite of a reviewer's job.
The same review, rewritten
Comments left down the margin of a 4,000-word report draft: Paragraph 2: "Confusing." Section 3: "This feels long. Tighten?" Section 3, paragraph 4: "Not sure about the tone here." Section 5: "Love this!" Section 6: "Can we add something on pricing? Also the second sentence is clunky, and I'd cut the third paragraph, it doesn't land for me. Maybe this section should come earlier?" Conclusion: "Needs more punch."
Pass: structure only. I have not looked at sentences yet. Section 3 - blocking. I stopped believing the argument here. You claim the migration reduced support load, and the only evidence is a quote from the team that ran it. A reader deciding whether to copy this needs a number or a named constraint. Section 6 - important. This answers a question the reader has by page two but arrives on page nine. Move it straight after Section 2 and cut the recap it opens with. Conclusion - optional. It restates Section 5 instead of saying what to do next.
The first set names feelings and locations but never the cost to the reader, so the writer must guess. It also stacks four editing levels inside one comment on Section 6. The second declares the pass, gives each note a severity, states the reader consequence, and leaves the prose decision to the author.
Common mistakes
| The mistake | Why it costs you, and what to do instead |
|---|---|
| Commenting in reading order | You leave a note the moment you feel something, and half turn out to be answered two pages later. Read the whole draft first, then write your comments in order of importance. |
| Praise that carries no information | Telling a writer you love a passage gives them nothing they can repeat. If something works, say what it does: it makes the stakes concrete, or it answers the objection you were forming. |
| Rewriting instead of diagnosing | Supplying your own sentence is faster than explaining the problem, so reviewers do it constantly. The writer then accepts prose that does not sound like them, or reverts it and loses the underlying point. |
| Treating taste as a defect | You would have written it differently, so you flag it. Mark those notes optional and let them go. A review that corrects style preferences trains the writer to write for the reviewer rather than the reader. |
Your checklist
- Read the draft end to end before leaving a single comment.
- State the pass and the editing level at the top of your review.
- Label every comment blocking, important, or optional.
- Name the reader consequence inside each substantive comment.
- Ask a question wherever a section's purpose is unclear.
- Delete the comments that later parts of the draft already answer.
Key terms
- Severity label
- A tag marking a comment as blocking, important, or optional, so the writer can tell what must be fixed now, what improves the piece, and what is preference.
- Reader consequence
- The specific cost a problem imposes on the intended reader, such as losing the thread or doubting a claim. It is what turns a reaction into an argument for change.
- Line pass
- A revision pass focused on sentences: rhythm, precision, and word choice. It comes after the structure has settled, because line work on material you later cut is wasted.
Getting a draft read by someone else?
Frequently asked questions
How much feedback is too much?
+
If the writer cannot hold your review in their head, they will work through it mechanically instead of thinking. Aim for the few problems that would most improve the piece, usually three to seven on a long draft. Volume signals thoroughness; selection signals judgment.
Should I fix typos while reviewing structure?
+
Fix them silently if the tool allows it, and mention once that you did. Do not raise them as comments during a structural pass. They compete for attention with the objections that decide whether the piece works, and they make the review look longer than it is.
What if I think the whole piece is wrong?
+
Say so directly, once, at the top, and then stop commenting inline. Margin notes on a draft you believe should not exist waste both people's time. Ask for a conversation about the premise before anyone writes another version of the same argument.
How do I give feedback to someone more senior?
+
The same way as to anyone else: reader experience, consequence, severity. Describing what happened to you as a reader is an observation, not a status claim. Skip the hedging, which reads as uncertainty about the observation rather than as politeness.
Can an AI model review a draft for me?
+
Strut AI is good at listing unsupported claims, spotting repetition, and finding sections that drift from their stated purpose. It cannot tell you whether the argument will convince your particular reader. Use it to clear mechanical problems before a human spends attention on the draft.
Should feedback ever be given in a meeting?
+
Use a call when the disagreement is about the premise, when the writer is stuck, or when written comments have gone around twice without moving. Write the outcome down afterwards, because a decision that lives only in a call gets relitigated within a fortnight.