Revise from the top of the hierarchy down
Use this order:
- Truth: Is the central account accurate? Are facts, memories, sources, and uncertainties correctly represented?
- Argument: Does the conclusion follow from the evidence and mechanism? What is the strongest countercase?
- Reader value: Does the piece solve a real reader problem?
- Structure: Does each section move the reader through the intended change?
- Emphasis: Does the draft make the right idea feel most important?
- Evidence and source fidelity: Are claims supported, sources accurately represented, and attribution clear?
- Prose: Can the reader move through the language with low unnecessary effort?
- Mechanics: Grammar, punctuation, spelling, formatting.
Many writers work almost exactly backward, because the bottom of the list is easier to see.
Diagnose before rewriting
Your accepted editing practice has a strong constraint. First identify the problem and quote the evidence. Do not rewrite a passage merely because another version is possible.
A useful editorial sequence is:
exact passage → problem → why it matters → required decision → local edit only after the decision
This sequence prevents the editor from replacing the author while pretending to help.
Major revision comes before local revision
A major revision changes any of these:
- claim
- reader
- scope
- evidence
- order
- emphasis
- inclusion or exclusion of sections
A local revision changes any of these:
- wording
- sentence structure
- clarity
- repetition
- transitions
- grammar
If major revision is still possible, minimise local polishing.
Let time create distance when possible
A draft becomes hard to evaluate when you remember what you intended rather than read what is on the page.
Distance can come from:
- time away
- changing font or medium
- reading aloud
- a reverse outline
- a cold reader
- converting headings to a one-page map
- asking AI to report what the text appears to claim, without telling it your intention
The goal is to reduce your access to the invisible context in your own head.
Build a skeleton before you polish the body
The science-writing notes contain a useful drafting method for research-heavy pieces. Reduce the material to one or two lines per section before filling it out. The skeleton records what each section must accomplish, not how the prose will sound.
The method is especially useful after reporting, because it stops the order of your notes from becoming the order of the essay.
A section skeleton might read:
- open on the supervisor discovering the mismatch
- show why the original attendance control looked reasonable
- explain the exception path the control removed
- compare three sites to test whether the mechanism repeats
- return to the payroll decision and state the boundary
Only after that spine works should sections become "meatier."
Keep a cuts file so deletion does not feel irreversible
When a good paragraph is structurally unnecessary, move it to a separate cuts file. This can reduce the sunk-cost resistance to removing it.
The cuts file is not a secret appendix. It is a psychological tool. It lets the live draft become cleaner without pretending the discarded work never existed.
Reader confusion is evidence, but the reader's fix is optional
An outside reader may propose the wrong repair. That does not make the confusion imaginary.
Separate two signals:
- Diagnosis signal: the reader lost the thread, misunderstood the claim, or expected evidence that never arrived.
- Prescription signal: the reader wants a new example, a different order, more background, or a rewritten sentence.
Treat the first seriously. Evaluate the second independently. The reader owns their experience of the draft. The author still owns the solution.