The Re-Analysis Loop:
How to Know When a Draft Is Actually Done

You submitted the first draft. The feedback came back: “it feels thin.” You revised. Added 400 words, fleshed out two sections, added an example in the middle. Sent it again. Same feedback: “still feels thin, needs more depth.”

At this point you start questioning yourself. You went back through the draft looking for what was missing. You Googled “how long should a blog post be.” You found a benchmark, hit the word count, and the next round of feedback was the same sentence, slightly reworded.

Is this your fault?

The feedback was real. Your editor's intuition that something was structurally incomplete is a genuine signal. The problem is that “depth” is not a structural description. It is a symptom report. And without a structural instrument, neither you nor your editor can diagnose what is actually missing.

The Editor Isn't Wrong. The Instrument Is Missing.

When an editor says a draft “feels thin,” they are picking up on something accurate: the content does not fully cover what the topic requires. Structural completeness is low. But “thin” describes a sensation, not a structure.

An editor who cannot name which specific concepts are absent, which relationships between ideas are unexplained, and which claims are made without the causal chain that would make them meaningful cannot give you actionable feedback. This is not a skill failure on their part. It is a tooling failure. The instrument that would let them say “the mechanism connecting X to Y is not explained” does not exist in most editorial workflows.

The result is a loop you probably recognize: feedback arrives, you interpret it, you add material based on that interpretation, the same feedback returns. The revision was real. The gap it addressed may or may not have been the actual structural gap. You have no way to know.

Word count benchmarks make this worse, not better. They answer a different question: how long is content in this category? They do not answer: does this draft contain the concepts this topic requires, at the depth the audience needs, with the relationships between those concepts explained? Those are not the same question. Adding words to meet a benchmark does not close a structural gap. It fills space.

What Structural Completeness Actually Means

A draft is structurally complete when it covers every concept the topic requires for the specific audience, at appropriate depth, and explains the relationships between those concepts: not just what they are, but how they connect, what causes what, what follows from what.

“Depth” is a proxy for this property. When your editor senses that depth is missing, they are often detecting that a relationship is unexplained, that a concept has been named but not integrated, that a causal chain has been assumed rather than demonstrated. These are specific structural conditions. They are diagnosable. But diagnosing them requires a structural map of what the draft actually contains.

Without that map, revision is positional. You look at sections that feel thin, by your own estimation or by word length, and you add material there. Sometimes you close the right gap by instinct. Often you add material around the gap without addressing it, and the next round of feedback is identical to the first.

This is the doneness problem: “done” has no structural definition in most editorial workflows. Done means the editor stopped complaining, or ran out of review cycles, or the deadline arrived. None of those are structural answers.

The Delta View: What Changes Between Drafts

A structural analysis of a draft produces a concept graph: a map of what the content covers, which concepts are fully integrated, which are partially integrated, and which are absent or mentioned without explanation. When you revise and run the analysis again, the output is a delta: a diff between the two structural states.

Eight concepts, two structural states. The number moved 29 points, but the actionable signal is which concepts changed state, which moved up, and which stayed where they were.

Here is what a delta output looks like at the row level between a first and second draft:

Concept: mechanism connecting rate limiting to downstream latency. State in Draft 1: absent. State in Draft 2: partially integrated, mechanism named but causal chain not demonstrated.

Concept: relationship between cache invalidation and consistency guarantees. State in Draft 1: partially integrated. State in Draft 2: fully integrated.

Concept: failure recovery sequence. State in Draft 1: absent. State in Draft 2: absent, no change.

This is a doneness signal. “Needs more depth” becomes: the causal chain connecting rate limiting to downstream latency is named but not demonstrated; the failure recovery sequence was not addressed in this revision at all.

The second revision now has a target. Not “make this section longer.” Specifically: demonstrate the causal chain from rate limiting to downstream latency, and address the failure recovery sequence which remains structurally absent.

A rescore from an aggregate tool tells you the number moved. It does not tell you which concepts your revision actually addressed, and which structural gaps remain open. That is true by product design: aggregate scores are designed to measure coverage in the aggregate, not to map which specific relationships changed state between versions. Both measurements are valid. They measure different things.

A note on honest scope. Sometimes “needs more depth” is an editorial voice judgment: the prose is thin in register, not structurally incomplete. The delta view addresses structural gaps. If the feedback is about voice, sentence rhythm, or register, that is a different problem. Know the difference before you run the analysis.

The Revision Protocol

This is the section worth bookmarking. Run this loop and “done” has a structural answer.

1. Run analysis on the draft before you revise. Get the structural map of what the current draft contains: which concepts are fully integrated, partially integrated, or absent. This is your baseline.

2. Identify the highest-priority structural gap. Look at what is absent or partially integrated. Cross-reference with your audience: which missing concept or unexplained relationship would most undermine comprehension for this reader? That is your revision target for this pass.

3. Revise against the specific gap. Do not revise the whole draft. Target the concept or relationship the analysis surfaced. Add the causal chain. Demonstrate the mechanism. Explain the relationship that was named but not integrated. Do not add material elsewhere to compensate.

4. Run analysis again and read the delta. What changed state? What moved from absent to partially integrated? What moved to fully integrated? What remained unchanged? The delta tells you exactly what the revision accomplished structurally.

5. Close the loop or identify the next target. If the delta shows the gap closed, move to the next structural gap from the baseline map. If the delta shows the gap partially closed, you know specifically what the next revision needs to do. Repeat until the delta shows no structurally significant gaps remaining.

That last condition, no structurally significant gaps remaining, is what “done” means structurally. It is a specific answer.

The Conversation the Editor Should Be Having

When an editor says “it feels thin,” they are trying to have a structural conversation without the instrument that would make it specific. The delta view is the missing conversation.

“The mechanism connecting X to Y is not explained” is what your editor would say if they could point to the structural map. They cannot say it because they are working from reading sensation, not from a concept graph. The feedback you receive is the best approximation available when the instrument is missing.

This is worth sitting with: your editor is not failing you. They are doing the most precise thing their tools allow. “Thin” is accurate. It is just not specific.

The delta view makes it specific. After the second revision, instead of re-reading the feedback and guessing at the gap again, you have a list: these concepts changed state, these did not, this relationship remains unexplained. The revision protocol runs against that list. The loop closes when the list is empty.

When structural completeness is the actual gap, “done” has a specific answer: the delta is closed.