# Worked Example — Statement Coach

A gold-standard run of this skill: a generic CS-master's statement diagnosed line by line, with mining questions doing the work a ghostwriter would have faked.

## The brief

Amara is applying to the MS in Computer Science at Northgate University (fictional), focus in human-computer interaction. Prompt: "Describe your purpose in pursuing graduate study" — 800-word limit; her draft is 640 words. The draft opens: "I have always been passionate about technology and its power to change lives." It tours her resume (dean's list, two internships, a senior project building an accessibility browser extension), calls Northgate "a prestigious program with world-class faculty," and closes: "I am confident my dedication and hard work will make me an asset to your program." Buried in paragraph four, one sentence: "My extension was tested by my uncle, who has low vision, and his feedback changed my design." She mentions no specific faculty. Her real story, extracted in conversation: she rebuilt the extension three times because her uncle couldn't use versions one and two, and that's when she decided accessibility research was the field.

## The output

### Statement Diagnostic: Northgate MS CS (HCI) — draft of April 2

#### Scores (with the reason, not just the number)
**Specificity: 3/10** — five sentences fail the 200-applicants test outright; the one specific asset (the uncle test) is buried in a subordinate clause. **Trajectory: 4/10** — the draft lists dots (dean's list, internships, project) but never the line between them; why HCI, why now, why research? **Fit: 2/10** — "prestigious program with world-class faculty" names no faculty, no lab, no course; it would survive find-and-replace with any school, which is the test it must fail. **Voice: 6/10** — paragraph four briefly sounds like a person; the rest sounds like a template defending itself.

#### Line-level feedback
> "I have always been passionate about technology and its power to change lives."
Generic — the canonical 200-applicants sentence; committees report reading it hundreds of times per cycle. The replacement material already exists in your draft: the uncle test. Openings that show a Tuesday beat openings that claim a lifetime.

> "My extension was tested by my uncle, who has low vision, and his feedback changed my design."
Alive — this is the essay. It's currently 19 words in paragraph four. "His feedback changed my design" is doing enormous compressed work: what feedback? What changed? This is where the mining questions aim.

> "…a prestigious program with world-class faculty."
Generic and risky — template language is where wrong-school-name accidents live. Fit must name real things: a professor whose work you can discuss, a lab, a course sequence.

> "I am confident my dedication and hard work will make me an asset to your program."
Generic closer — asserts qualities instead of showing them, and "asset" frames you as inventory. The closer should point forward: the specific work you intend to do at Northgate.

#### Material-mining questions (answer in writing before revising)
1. What exactly could your uncle not do with version one? Describe the moment you watched him fail with it.
2. What did versions two and three change, concretely — and what did you have to unlearn between them?
3. When did this stop being a course project and become the field you want? Was there a specific day?
4. Which Northgate faculty's work is closest to what you built? Have you read their papers — which one, and what would you ask them?
5. What accessibility problem do you want to work on that you *can't* solve with your current skills — what does the MS unlock?
6. Your internships: did either touch users directly? Any moment there that rhymes with the uncle story?

#### Revision plan
**Pass 1 — restructure:** open with the uncle scene (from Q1's answer); rebuild the middle as trajectory — extension → three rebuilds → the decision (Q2, Q3); cut the resume tour to one sentence, the CV already holds it. **Pass 2 — fit:** replace the prestige sentence with the faculty/lab material from Q4–Q5. **Pass 3 — cut to limit:** the new material will overshoot 800 words; cut the closer's assertions first, then any surviving adjectives doing evidence's job.

**The one thing:** move the uncle from paragraph four's subordinate clause to the opening scene, and let the whole statement be the story of the three rebuilds. Everything else is editing; this is the essay.

## Why it's shaped this way

- **Every generic flag quotes the actual line and names the test it fails** — the 200-applicants test, the find-and-replace test — rather than vaguely calling the draft "generic," per the quality checks.
- **The diagnostic found the alive sentence, not just the dead ones** — the uncle line gets flagged as the essay's real center, because coaching that's only demolition leaves nothing to build with.
- **Mining questions target diagnosed gaps** — Q1–Q3 excavate the buried scene, Q4–Q5 manufacture real fit material, Q6 checks for corroborating scenes — not general biography ("tell me about your childhood").
- **No output paragraph is written to be pasted in** — the coach describes what the opening should *do* and asks the questions that produce it in Amara's words; the integrity line from the framework is structural, and at interview it's Amara who has to own every sentence.
- **Fit feedback references the actual program** — it demands named faculty and sends her to their papers, because "prestigious program" is both weak and the classic wrong-name-in-the-template hazard.
- **The one-thing is structural, not cosmetic** — reordering the essay around its only real scene — matching the rule that the recommendation most move the essay, not most polish it.
- **Scores come with reasons** — a bare "3/10" teaches nothing; the reason is what makes the revision plan followable, and the praise-pad anti-pattern is avoided by scoring honestly (a 6/10 voice is the kindest true number here).