# 04 — Accuracy and Scope Audit

Audit of `02_INFS3700_Final_Exam_Revision_EN.html`, `03_INFS3700_Final_Exam_Revision_CN.html`, `05_INFS3700_Exam_Sprint_Guide_EN.html` and `06_INFS3700_Exam_Sprint_Guide_CN.html`.

> **This audit is current throughout.** Sections 1–13 describe the pack as it now stands; §14–§18 record how it got there. No section contradicts another — where a revision changed an earlier finding, the earlier section was rewritten rather than left standing with a pointer.

---

## 1. Files reviewed

**293 files across 45 directories** were enumerated by a recursive programmatic walk of the course folder, **with the output folder `INFS3700_Final_Exam_Revision_Pack\` excluded from the walk** — nothing this pack produced is counted. The per-type counts sum to 293; the counting method is stated in full in `00_FILE_INVENTORY.md`. The figure has been recalculated at each revision rather than carried forward: 269 (a hand-written estimate, wrong) → 284 → **293**, the last increase being the eight transcripts and one audio file added before this revision. Of the 293:

| Category | Count | Treatment |
|---|---|---|
| Read in full, page by page | 21 files / ~540 PDF pages | Weeks 1–10 lecture decks, all tutorial preparation files, the tutorial case study, three assessment briefs |
| Read as text | 14 files | Week 8 and Week 9 tutorial transcripts, Week 9 preparation forum, tutor feedback, student forum answers, and the **eight newly added lecture and tutorial transcripts** (all read in full — see §15) |
| Read selectively | ~30 files | Student coursework — workshop deck, workshop plan v2, workshop transcript, individual portfolio, testing-plan and problem-definition workspace files |
| Enumerated and classified, not used | ~210 files | Duplicates, superseded versions, media, other groups' photographs, third-party ebooks, AI-generated working notes |

Full enumeration with authority tiers is in `00_FILE_INVENTORY.md`.

**Extraction method.** Every lecture deck was read in chunks of ≤18 pages by a dedicated agent instructed to reproduce the lecturer's exact wording, then **independently re-read and adversarially verified** by a second agent whose brief was to refute the first. Twelve source clusters were processed this way. Seven of the twelve verification passes returned `fidelity_ok: false` and produced specific corrections — all of which were applied before writing (§7 below).

---

## 2. Authoritative files used

| Rank | File | Authority for |
|---|---|---|
| 1 | `WEEK10\10 UX  SD in Industry - Lecture Slides T2 2026.pdf` | **Exam scope, format, mark split, inclusions, exclusions, practice questions.** Overrides everything else on these points |
| 2 | `assessments\INFS3700 T2 2026 Group Assignment.pdf`<br>`assessments\INFS3700 T2 2026 Individual Assignment May.pdf`<br>`assessments\INFS3700 T2 2026 Assignment Problem Spaces_May.pdf` | Assignment task wording, artefact requirements, participant requirements, problem-space scenarios |
| 3 | `WEEK1–WEEK9` lecture decks (8 files) | Concepts, terminology, definitions, frameworks, worked examples |
| 4 | `WEEK1–WEEK9` tutorial preparation files (8 files); `WEEK8/9 Tutorial case study.pdf` | P&P task wording; the Week 9 prep file supplies the five blueprint components as the P&P task names them |
| 5 | `WEEK9\新录音 92.txt`, `WEEK8\tut transcript wk8.txt`, `group workshop\tutor给 plan的反馈.docx`, and the eight transcripts in `lecture transcript\` and `WEEK10\tut transcript raw.txt` | Spoken clarification where the speaker and context are clear. **Supporting only** — a transcript never overrides the Week 10 slides, an official brief or a printed lecture slide |
| 6 | Student coursework | Chapter 16 example bank only, each validated and limitation-stated |

---

## 3. Duplicates and superseded files found

### Exact byte duplicates (MD5-verified)
`3700assignment backup files\3700\` is a complete **mirror** of the primary folders — 17 files exist twice with identical content (all seven lecture decks present there, both assignment briefs, both workshop whiteboards, the workshop guide, the Week 5 wireflow, the persona, the script, the Week 5 prep image). The primary copy is cited throughout; the mirror is ignored.

**One cross-week duplicate that changes an interpretation:** `WEEK8\08 SD Intro - Tutorial case study.pdf` and `WEEK9\09 - Tutorial case study.pdf` are **byte-identical**. One case study serves both weeks — Week 8 frames it with the 5Ps, Week 9 blueprints it. Cited as "Week 8/9 Tutorial case study".

Also identical: `AUTHORITATIVE_Slides_01_to_11_Final_Content.docx.docx` and `新建文件夹\s1-11.docx` (both student-generated, both unused).

### Superseded versions (near-duplicates)
| Used | Superseded |
|---|---|
| `week4_design_principles_forum_submit_v2.html` | `…_final.html`, `…_submit.html`, `week4_design_principles_forum.html` |
| `Workshop Plan 第二版.pdf` | `Workshop Plan第一版.png` |
| `INFS3700_Group5_Workshop_SLIDE_DECK_FINAL.pptx` | `…_SLIDE_DECK.pptx` |
| `group workshop\…_SLIDE_DECK_FINAL.pdf` (1233 KB) | backup copy (1241 KB) — **not identical**, older export |
| `INFS3700_Individual_Portfolio_Requirements_Fixed111111.pdf` | three earlier portfolio drafts |
| `最终的稿子walkthrough_script_short.md` | `walkthrough_script.md.pdf`, workspace `walkthrough_script.md` |
| `INFS3700_Group5_PPT_Handover_Guide_FINAL_v2.html` | three earlier handover-guide versions |

### Empty files
`group workshop\一些事实.txt` and `group workshop\小组成员情况.txt` are **0 bytes** in the current filesystem state. Contextual information about the workshop's actual facilitation split was recovered from other files instead.

---

## 4. Excluded topics — how each was handled

| Excluded topic (W10 s13) | Where it appears in the sources | Treatment in the guide |
|---|---|---|
| **Personas** | W3 slides 34–46 (13 slides); W3 tutorial prep task; Individual Assignment criterion 1a; student workspace `04_Personas\`; `persona_mia_zhang_v2.html` | **No examinable chapter.** Named in the Ch. 15 artefact table as excluded, and in Ch. 6 where the course's own method map places them. No construction taught. |
| **Journey maps** | W3 slides 47–54 (8 slides); Individual Assignment criterion 1b; student workspace `05_Customer_Journey\` | **No examinable chapter.** One structural mention in Ch. 6's method map, and the *pain-points definition* is cited from W3 s51 with an explicit note that the slide sits in the excluded section. |
| **Design patterns & design systems** | Listed on the W4 agenda (s4); referenced in W1 s19 and job ads on W1 s8 | Noted in Ch. 8 with the useful finding that **the Week 4 deck contains no slides for either section** — it jumps from s39 to s40. Nothing to revise. |
| **Service design capabilities** | Previewed as W9 topic 09.5 on W8 s45; "Service Capabilities" band in the case-study diagrams | Mentioned only to keep it distinct from the 5Ps and the blueprint. Ch. 5 additionally flags that the "Learning (capability) workshop" (W2 s57) is an unrelated topic that shares the English word. |
| **Mid/high-fidelity sketching** | W5 slides 23–25, 30–33; Individual Assignment criterion 3b | Definitions retained (Ch. 11) because they are needed to justify choosing low fidelity; **producing them is marked excluded** throughout. Ch. 16's EX-7 stops at the low-fidelity decision. |
| **Calculations** | W3 s67 feature-scoring table; W7 s35 probability figures | The scoring table is described for its criteria only; the Magic-of-5 figures are presented as facts to recall. Ch. 9 states explicitly that numbers are interpreted, never computed. |
| **Referencing** | — | No citation format is taught in the guide for exam use. The guide's own source labels are for revision accuracy, and Ch. 20 says so. |

**Roughly a third of the Week 3 deck (21 of 68 slides) is exam-excluded.** The guide deliberately does not let that material dominate Chapter 6.

---

## 5. Coverage against every Week 10 *possible* topic

| Possible topic (W10 s12) | Covered in | Depth | Primary sources |
|---|---|---|---|
| **Participant criteria** | Ch. 5 Part A | Full — 5 recruiting questions, worked screener anatomy, suitable/unsuitable comparison, sample-size table, tutor's dimensions | W2 s67; W3 s5; W7 s26–27, s35; Group Assignment p5; tutor feedback |
| **Workshop planning** | Ch. 5 Part B | Full — definition, 4 types, agenda skeleton diagram, 4Cs with all 40 activities, tutor's 3 rules, poor/improved instruction pair | W2 s52–57; W2 s10 / W3 s8; W10 s8–9; Group Assignment p4 |
| **Interview planning** | Ch. 5 Part C | Full — question types, protocol a–g, discussion guide with timings, Beginning/Middle/End, Forming–Performing | W2 s41–47, s64–66, s69 |
| **Synthesis & analysis** | Ch. 6 | Full — sensemaking, IRA, affinity mapping with size rules, when to stop, problem framing, 6 traps, HMW with good/better/poor, prioritisation ×2 matrices | W3 s14–33, s55–67; W10 s27 |
| **Wireframes** | Ch. 11 | Full — definition, 6 things communicated, fidelity ladder, comparison table, wireflows, annotation, annotated original diagram, exam drawing checklist | W5 s7–33; W5 tut prep p1 |
| **UX Testing** | Ch. 10 | Full — concept vs usability, moderated/unmoderated, hypothesis recipe, task wording, prioritisation matrix, briefing script, 8 Rules, bias, reusable test plan template | W7 s6–51; Individual Assignment p4; W8 transcript |
| **Heuristics evaluation** | Ch. 7 | Full — all five with attribution, positive/negative framings, per-heuristic evidence/example/recommendation, 5-step finding structure, weak/strong pair | W1 s30–40; W10 s25–26; W5 s51–52 |
| **Research methods** | Ch. 4 | Full — qual/quant purposes, Risk × Problem Clarity, 14-method selection matrix, triangulation, survey craft | W2 s17–39; W3 s32–33; W10 s6–7 |
| **Design principles** | Ch. 8 | Full — all five with components, positive/negative framings, everyday-object examples, the second digital layer, error handling pair, three named laws | W4 s8–32; W10 s28–29; W5 s54–55 |
| **Storyboards** | Ch. 12 | Full — What/Why/Who/How, three acts, 8 elements, 3 principles, Heartline example, 6-frame template, storyboard vs wireframe vs blueprint | **W4 s43–50 (primary)**; W7 s10 (storyboards as a concept-testing artefact) |
| **Service design 5Ps** | Ch. 13 | Full — all five with sub-components, relationship map, diagnostic questions, worked GP application, the project-brief trap | W8 s38–44 |
| **Service blueprint** | Ch. 14 | Full — definition and Shostack origin, three characteristics, current/future state, all three lines, lane definitions, structure, six-step build process, zoom and fidelity, complete worked example, printable blank template, six service design principles, 9 common errors | **W9 lecture deck, 40 slides (primary)**; W9 tut prep p1 (P&P task wording); W9 transcripts (tutor's live corrections); W8 s40, s45 (framing) |
| *Format:* theoretical questions | Ch. 17 Parts B–C | 6 questions with model answers | |
| *Format:* case study questions | Ch. 17 Parts D, I | 3 case questions incl. a full 55-mark 5-part case | |
| *Format:* drawing out artefacts | Ch. 11, 12, 14; Ch. 17 Part G | 3 drawing questions + annotated worked artefacts + a printable blank template | |
| *Format:* using examples | Ch. 16 | 8 validated personal examples + 9 course examples + everyday alternatives + an allocation table | |

**All eleven topic entries printed on Week 10 slide 12 are covered, together with all four named formats.** The guides also cover **quantitative and qualitative data** as a twelfth revision area — it is not printed on slide 12, but Week 10 slide 30 revises it directly and Week 10 Practice Question 3 examines it. Nothing on slide 12 is absent or thin. Note that slide 12 prints eleven entries rather than twelve because **"Workshop / interview planning" is a single combined entry**; this guide keeps it as one area and covers both halves of it.

---

## 6. Uncertain content and how it is labelled

| Item | Uncertainty | Label used in the guide |
|---|---|---|
| Time per mark | Arithmetic, not published | "revision guidance" badge + explicit note |
| Answer structures per command verb | No official rubric exists | Chapter-level warning quoting W10 s15 ("We will not provide marking criteria") |
| Model answers | Written for this guide | "These are not past papers and there is no official rubric" |
| Blueprint swim lanes and lines | **No longer uncertain.** Sourced from the Week 9 lecture deck (lane definitions s16–18, lines s7, s8, s17, s23, s34) | Cited to the deck by slide number. The transcripts are now used only where they add the tutor's live corrections, and those are marked as the tutor's spoken description |
| "Line of internal interaction" | **No longer uncertain.** Named in the Week 9 lecture deck on slides 7, 8 and 23. The residual uncertainty is narrow: the deck labels it on diagrams but gives no prose definition | Taught as course terminology. The chapter states that the deck labels rather than defines it, and that the blank template on s34–35 draws only two lines — a presentation variation within the course's own material |
| F pattern / the second viewing pattern | **Partly resolved by the Week 4 recording.** The slide names neither; the lecturer names the **F pattern** aloud and describes its scan path, but the second name is lost in the transcription ("one is that, uh, pattern") | "F pattern" is taught as usable course terminology, cited to the recording. The second pattern is described rather than named, and the guide states that no source in this folder confirms it as "Z" |
| Four named biases | Named on W7 s39 with **no definitions** | "named, not defined" badge |
| Retain / revise / reject | Not the course's language | "nearest equivalent" badge on "Go / no go / refactor" |
| Pain-points definition | Sits on an excluded journey-map slide | Cited with the location flagged |
| Method matrix strengths/limitations | Some quoted, some generalised | Table note distinguishes the two; generalisations marked "supplementary" |
| Diagnostic questions per P | This guide's device | "revision device, not a course framework" |
| Worked GP 5Ps and blueprint | Assembled, not a course example | "worked example, assembled from course sources" |
| Storyboard 6-frame contents | Course template is unlabelled | "supplementary structure, not a course template" |
| Student examples | Not validated by staff | `student example` badge + a stated limitation on every one |

The guide uses the required gradations consistently: *explicitly included*, *explicitly excluded*, *identified by Week 10 as a possible topic*, *supported by tutorial activities*, *not confirmed as course terminology*, *supplementary explanation, not explicitly stated in the course files*.

---

## 7. Source conflicts found and how they were resolved

| # | Conflict | Resolution | Where flagged |
|---|---|---|---|
| 1 | **Develop vs Design** — W5 s5 labels the third phase *Develop*; W5 s36 labels it *Design* | Use **Develop** (six decks including Week 10) and state the conflict | Ch. 1 warning box; Ch. 19 §F |
| 2 | **Wireframe fidelity** — W5 s19 "medium to high"; W5 s8 and s26 "low / mid / high" | Present s8/s26 as the general rule, s19 as the wireframe's typical role. Do not state a single fidelity | Ch. 11 card; Ch. 19 §F |
| 3 | **Metric vocabulary** — W7 says "Success rate"; W5/W10 say "direct success / indirect success / confidence rating" | Use the Week 10 vocabulary (it is the revision deck) and give the Week 7 equivalent | Ch. 9 warning table; Ch. 19 §F |
| 4 | **Heuristic wording** — s35 "Match between system **&** real world" and "Don't waste my time**!**"; s39/s40 differ | Use s35 as canonical because the tutorial prep reproduces it exactly; note the variants | Ch. 7 table |
| 5 | **Two five-P lists in one deck** — W8 s39 (the 5Ps) vs W8 s35 (project brief headings) | Dedicated comparison box; the s35 slide never calls its list "the 5Ps" | Ch. 13 opening; Ch. 3 detail panel; Ch. 19 §C |
| 6 | **Two Beginning/Middle/End structures** — W2 s55 (workshop) vs W2 s69 (interview), different steps | Presented separately with an explicit warning not to merge | Ch. 5; Ch. 19 §C |
| 7 | **Two prioritisation matrices** — W3 s64 (Impact × Importance) vs W3 s65 (Impact × Effort) | Presented side by side in a warning box with the axis definitions | Ch. 6; Ch. 19 §C |
| 8 | **W3 lecture vs W3 tutorial prep** — milestone weeks (3,4,5 vs 4,5,6); naming convention (T18A-01 vs W18A-01); "written/verbal" vs "written" feedback | Administrative only, not examinable. Not reproduced in the guide | — |
| 9 | **Term labels** — W5, W7, W8 decks labelled **T1**, not T2; W7 tutorial prep labelled **T1 2025** | Flagged in the inventory and Ch. 19 §F. Content is consistent with the T2 decks; only in-deck dates are stale | Ch. 19 §F; `00_FILE_INVENTORY.md` |
| 10 | **Assignment briefs cite "T1 2026 Assignment Problem Spaces"** while the supplied document is titled T2 2026 | Stale cross-reference in the briefs. The T2 document is the one used | `00_FILE_INVENTORY.md` §3 |
| 11 | **4Cs slide numbering** — two items numbered "4", none numbered "5" | Reproduced as printed with a `[sic]` note rather than silently renumbered | Ch. 5 table note |
| 12 | **`tut transcript wk8.txt` is mislabelled** — the tutor opens by saying the session covers **Week 7** content | Used only as a Week 7 (UX testing) source. **Not** cited as evidence of anything about service design | Ch. 20 source table |
| 13 | **Blueprint lane naming** — the tutorial prep writes "Front stage / Back stage" (two words); the lecture deck writes "Frontstage / Backstage" (one word); the two embedded NN/g diagrams use two further namings | All four are course content and any of the four is defensible if used consistently. Reconciled in a comparison table. The tutorial-prep wording is still recommended **for the Week 9 P&P task specifically**, because that is the wording the task itself uses | Ch. 14 opening; Ch. 19 §F |
| 14 | **How many lines a blueprint draws** — the lecture deck **names three** (s7, s8, s23); some templates in the same deck **draw two** (s34–35); the tutorial prep names none | Teach three. All three names are the course's own terminology, so none of them needs a disclaimer. The two-line templates are a **presentation variation within the course's own deck** — they are not evidence that the third line comes from outside the course, and the guide says so. A student who draws a two-line blueprint matching the template is also correct | Ch. 14 lines table; Ch. 19 §F |
| 15 | **Pen or pencil for the hand-drawn artefact** — the Week 10 what-to-bring slide says a pen is preferred; in the tutorial the tutor read that line out and disagreed, preferring a pencil because it can be erased | Both are recorded. The tutor settled it himself — "Pen or penciling doesn't really matter that much… make sure that you can make your drawing **clear to us**" — so the guide says **bring both** and states that legibility in the uploaded photograph is the requirement neither source disputes. This is the one place in the pack where a transcript is allowed to qualify a printed slide, and it does so because the speaker explicitly addresses that slide | Ch. 0 warning box; sprint §G |
| 16 | **Lane order on the Week 9 elements slide** — the lecturer says the printed order is flipped | Recorded, not corrected: "but the very first one is customer action. Um, so I know that **order is flipped**, but it doesn't matter too much". The guide teaches the canonical top-to-bottom order and notes the slide's sequence differs | Ch. 14 |

### 7.1 The Week 9 lecture deck — final position, and the correction it forced

**Final position (this supersedes every earlier statement in this pack about Week 9):**

1. **`WEEK9\09 Service Blueprinting - Lecture Slides T2 2026.pdf` is the primary source for Chapter 14.** It is a 40-slide T2 2026 lecture deck. Every definition, lane description, structural element and process step in Chapter 14 is cited to it by slide number.
2. **All three lines are course terminology.** *Line of interaction*, *line of visibility* and *line of internal interaction* are all named in the deck — slides 7, 8 and 23. None of the three needs a disclaimer, and none should be labelled as the student's own addition.
3. **The two-line templates are a presentation variation, not a contradiction.** The blank template and worked example on slides 34–35 draw two lines. That is a simplification the lecturer made inside the course's own deck. It is **not** evidence that the third line comes from outside the course.
4. **The Week 9 tutorial preparation file and the tutorial transcripts remain valid supporting sources.** They supply the P&P task wording, its five-component list, and the tutor's live in-class corrections — none of which is in the deck. The tutorial-preparation lane terminology ("Front stage / Back stage") is still recommended **for the P&P task**, because it matches the task's own wording; it is not being presented as the only correct naming.

**The correction that produced this position.** The first build of this pack stated that the phrase **"line of internal interaction" does not appear in the course materials**, and instructed the student to label any third line as their own addition. That conclusion was drawn from the only Week 9 sources then available — a 2-slide tutorial preparation file and two tutorial recordings, which between them name only two lines. The lecture deck was added to the folder afterwards and it names all three. **The earlier guidance was wrong.**

What was done about it:

- Chapter 14 was **rewritten from the lecture deck**, which is now its primary source.
- The chapter carries a **visible correction notice** rather than a silent edit, so a student who read the earlier version is told explicitly that the advice changed.
- The blueprint diagram was redrawn to show all three lines, all three drawn and labelled as course terminology, with a caption noting that some templates in the deck draw only two.
- The "claims deliberately not made" entry in `01_SOURCE_MAP.md` was first **withdrawn and struck through**, and was then **removed from the active provenance table altogether** in the final targeted revision — an active claim table should state the current evidence position, not preserve an obsolete claim. Section 14 of the Source Map now opens with the settled position instead, and the history lives here. See §16.1.
- Chapter 19's terminology table lost the incorrect row and gained two accurate ones covering the lane-naming and line-count variation.
- Every remaining passage in this pack that described Week 9 as having no lecture deck, or the third line as absent, external, or the student's own addition, was located by full-text search and rewritten. See §14.

**Why the original error was reasonable but still an error:** the reasoning was sound given the evidence, and the uncertainty was flagged rather than hidden. But the guide stated a negative — *"does not appear in the course"* — from an incomplete source set. A negative claim of that kind needs the full source set behind it, and the correct framing at the time would have been "not found in the Week 9 sources available here", which is what the sourcing warning elsewhere in the chapter already said.

---

## 8. Student examples that required correction

| Example | Problem found | Correction applied |
|---|---|---|
| Week 9 blueprint (Preparation Forum) | Writes "Front-stage actions" / "Back-stage actions" with hyphens | Guide uses the slide's spelling — **no hyphens**. Correction stated explicitly in Ch. 16 EX-8 |
| Individual assignment testing plan | Uses the **Single Ease Question** (7-point, Sauro & Lewis) and **first-click accuracy** — neither appears in the course slides | Retained as an example but labelled "industry instruments, not in the course". Ch. 16 EX-6 gives the course's own vocabulary as the alternative |
| Individual assignment hypothesis | The "at least 4 of 5" threshold is the student's own judgement | Kept, with the limitation stated: the course publishes no threshold conventions |
| Group workshop deck | Claims the four 4Cs activities were "verified in the Encyclopedia" — but the group's own guide records that the Encyclopedia file could not be opened | Guide cites **activity names and numbers only** (which come from W2 s10 and are reliable) and states the limitation in Ch. 16 EX-1 |
| Group workshop dot voting | Deck reports "cluster support is directional, not strictly comparable" because mixed reaction types were used | No vote counts are quoted anywhere in the guide |
| Group workshop plan | Named participants before defining criteria; tutor flagged it twice and the group's own audit concedes it was "not clearly addressed" | **Reframed as a lesson**, not a model. Ch. 16 EX-2 presents it as a failure with a stated consequence |
| Second student HMW | "How might we give group members **one trusted view**…" embeds a solution | Retained as the **contrasting weak example**, with the fault named |
| Student causal chain | Presented as a finding; it is the team's inference | Guide states it is an inference, and notes that labelling inference as inference is itself worth marks |
| Group 5 workshop recording password | Appears on the deck's closing slide | **Not reproduced** in this pack |

---

## 9. Unsupported claims deliberately omitted

Beyond the terminology list in `01_SOURCE_MAP.md` §14, the following were considered and left out:

- **Any statement about what markers reward, or about how marks are distributed inside a sub-part.** Week 10 slide 15 states that no marking criteria will be provided. The one adjacent statement the pack does carry is the tutor's own description of how he marks (bullets are fine; he works from the points, not the prose) — and it is labelled as a tutor explanation, not as a rubric.
- **Word counts per mark band.** Nothing in the course supports them; the guide gives proportions and time instead.
- **Nielsen's canonical phrasings** for principles the lecturer abbreviated (e.g. "Help Users With Errors").
- **Norman's canonical seven stage names.** The course presents seven *questions*.
- **Definitions of the four named biases.** The slide names them without defining them.
- **The Workshopper Playbook / Workshop Exercise Encyclopedia framings.** Third-party, and using them risks substituting external terminology for the lecturer's. The 4Cs content comes from W2 s10, the course's own reproduction.
- **The Thoughtworks five principles as an examinable framework.** Set reading, clearly labelled as such.
- **Individual mouse/phone caption text from W1 s24.** Flagged by the verification pass as below reliable legibility; described rather than quoted.
- **Detailed cell contents of the case-study platform diagram (p5).** Same reason.
- **Any inference from the four Week 9 student presentations' garbled transcript passages.** Several were flagged as unrecoverable ASR; only the two clearly-recovered structural observations were used, and both are attributed.

---

## 10. Technical confirmations

| Check | Result |
|---|---|
| **No source file modified, renamed, moved or deleted** | ✅ Confirmed by content comparison, not just by intent. All access was read-only; every file written is new and inside `INFS3700_Final_Exam_Revision_Pack\`. PDF page images were rendered to new PNG files, and the renderer never saved back to a source PDF.<br><br>**Verification:** six lecture PDFs exist in two places (the week folders and the `3700assignment backup files\3700\` mirror). They were MD5-verified identical *before* any rendering, and re-verified identical *after* all rendering — so the render pass altered zero bytes.<br><br>**One timestamp to explain:** `WEEK10\10 UX  SD in Industry - Lecture Slides T2 2026.pdf` shows a last-modified time during this session. Its **CreationTime is identical to its LastWriteTime**, which is the signature of a OneDrive cloud-only file being hydrated to local storage on first read — not of an edit, which would have preserved the original creation time. The file is intact: 1,377,038 bytes, 33 pages, and page 23 still reads "Exam Questions / Three questions, all with sub-parts / • Questio…". Week 10 is the only source file without a local mirror to diff against, which is why it is called out explicitly here. |
| **Both HTML files work offline** | ✅ Confirmed by automated validation: zero `<link href="http…">`, zero `<script src="http…">`, zero `@import url(http…)`, zero `@font-face`, zero CDN references, zero remote image hotlinks. All 36 images resolve to files inside `assets/course_screenshots/`; all original diagrams are inline SVG. Fonts are system stacks only. |
| **Every image resolves** | ✅ Automated check confirms every `<img src>` points to an existing file on disk, relative to the HTML. |
| **Every image has alt text** | ✅ Automated check enforces `alt="…"` on every `<img>`. |
| **Every figure has a caption with source** | ✅ 52 `<figure>` elements, 52 `<figcaption>`. Course screenshots name the file and page; original diagrams name the slides their content derives from. |
| **No orphaned assets** | ✅ 75 slide crops were rendered during production; the 39 not referenced by either guide were deleted. Every remaining file in `assets/` is referenced. |
| **EN ↔ CN structural parity** | ✅ Automated cross-check, re-measured against the current files: identical counts of `section` (21), `table` (83), `tr` (578), `th` (217), `td` (1543), `figure` (52), `img` (36), `details` (33), `svg` (16), `li` (463), and of every semantic class (`src` 37, `card` 142, `hi` 43, `warn` 31, `good` 21, `bad` 21, `qbox` 22, `badge` 60). Section ids identical and in the same order. Image path lists byte-identical and in the same order. All 16 SVG blocks byte-identical. No content dropped. |
| **Encoding** | ✅ Both files declare `<meta charset="utf-8">` and are written as UTF-8. 58,220 CJK characters in the CN file; no mojibake markers (`â€`, `é`, `�`, U+FFFD) in either. The 7 CJK characters in the EN file are the Chinese *filenames* it cites (`新录音 91.txt`, `tutor给 plan的反馈.docx`) — correct, not an encoding fault. |
| **Course terminology survives translation** | ✅ All 36 checked course terms verified present in English in the CN file — the five heuristic names, the five design principle names, the 5Ps, the five blueprint components, the Double Diamond stages, Front of House / Back of House, line of interaction, line of visibility, How Might We, direct/indirect success, confidence rating, wireflow, storyboard. |
| **Nav ↔ sections consistent** | ✅ 21 nav links, 21 sections, all matched, no duplicate ids, no orphans. |
| **Tag balance** | ✅ `<section>`, `<table>`, `<figure>`, `<details>`, `<div>` all balanced. |
| **Tables scroll rather than overflow** | ✅ All **83** tables in each full guide, and all **22** in each sprint guide, sit inside a `.tw` wrapper with `overflow-x:auto`. Containment is checked by walking the tag nesting, not by comparing counts — equal counts would not prove that every table is actually inside a wrapper. |
| **Print layout checked** | ✅ All four files rendered to PDF via headless browser — full guide **EN 175 pages, CN 185**; sprint guide **EN 43, CN 40** — and pages inspected. Nav, top bar and search hidden; collapsed panels expand; figures, tables and cards avoid page breaks; screenshots and SVG diagrams print sharply; source labels retained. **One defect found and fixed:** two annotation labels in the service-blueprint diagram were struck through by the lane divider lines; both were moved clear and the corrected SVG re-synced to the CN file. |
| **Search works locally** | ✅ Pure JavaScript, no network. Filters sections and nav, highlights matches, reports hit counts, clears on Escape. |
| **Active-section navigation** | ✅ IntersectionObserver with a scroll-listener fallback. |
| **Responsive** | ✅ Single-column below 1000px with an off-canvas nav; no horizontal page scroll; tables scroll within their own container. |
| **No dark theme** | ✅ White/near-white backgrounds only; no `prefers-color-scheme` rules. |
| **Selectable text** | ✅ All body content is text. Original diagrams are inline SVG with real `<text>` elements. Only the course screenshots are raster, and each is accompanied by the same content in body text or a table. |
| **Assets folder** | `assets/course_screenshots/` — 36 PNGs, 2.5 MB total, optimised (average 71 KB). `assets/original_diagrams/` — one printable blank blueprint SVG. `assets/external_visuals/` and `assets/student_examples/` — **empty by design**: no external image is used anywhere, and student work is reproduced as quoted text rather than images. |

---

## 11. Visual quality control

- Every screenshot was rendered directly from its source PDF at 170 dpi, cropped to the relevant content (the UNSW footer band is cropped off on slide crops), then downsampled to 1200 px wide and palette-optimised. Average file size 72 KB.
- Text inside screenshots was checked at normal browser scale and at print scale. Two low-legibility candidates identified during extraction — the Week 1 slide-24 caption strips and the case-study platform diagram — were **not** used.
- Captions were checked against the source: every one names the correct file and page.
- No image is decorative. Each either reproduces a framework in the lecturer's own layout, demonstrates an artefact, shows a good/poor contrast, or maps a process. Where a table or SVG in this guide already carries the same content in selectable text, the screenshot was **dropped rather than duplicated** — that is why 39 of the 75 rendered crops were pruned.
- Diagrams were checked against the surrounding text for contradictions; where a diagram makes a layout choice the lecturer did not (the 5Ps quadrant arrangement, the six-frame storyboard template, the faint third blueprint line), the caption says so.
- Both guides were rendered to PDF and inspected page by page. The only visual defect found — two struck-through labels in the blueprint diagram — was fixed and re-verified.
- The EN and CN versions reference the **same image files** and contain **byte-identical SVG**, verified programmatically — no visual divergence is possible.

---

## 12. Remaining limitations

1. **Service blueprint is fully sourced.** The Week 9 lecture deck is Chapter 14's primary source and the chapter was rebuilt around it; see §7.1. Blueprint is now among the best-sourced topics in the guide. One narrow caveat remains: the deck **labels** the *line of internal interaction* on its diagrams (s7, s8, s23) but gives no separate prose definition of it, so the guide describes what the label marks rather than quoting a definition.

2. **Week 5, 7 and 8 lecture decks are labelled T1, not T2.** Their content is internally consistent with the T2 decks and with the Week 10 revision slides, but they may differ in detail from what was actually delivered this term. Weeks 1–4, 9 and 10 are T2 2026.

3. **All transcripts in this pack are automatic transcriptions of varying quality, and several corrupt course terminology.** The Week 1 file renders *intuitive* as "intrusive" throughout, which inverts the meaning; the Week 2 file has no speaker labels at all and is a single unbroken line; the Week 4 file never renders the second viewing pattern as "Z". Wherever this pack quotes a transcript it reproduces the ASR error rather than silently repairing it, and flags the intended word in brackets. No ASR spelling is treated as course terminology. Per-file limitations are tabulated in `00_FILE_INVENTORY.md`.

4. **The Week 9 tutorial transcripts are low-to-medium quality auto-transcriptions.** They are now used only for the tutor's live corrections and the exam warning — the lecture deck supplies the definitions. Where the guide quotes them, the wording has been lightly cleaned and is presented as the tutor's spoken description rather than as a verbatim quotation.

5. **Storyboards have the least slide coverage of any included topic.** The teaching content is the eight-slide block in the **Week 4 lecture (s43–50)**, plus **Week 7 slide 10**, which lists storyboards among the concept-testing artefacts. The chapter covers all of it. The Week 5 deck does not teach storyboards and is not cited for any storyboard claim.

6. **No past papers exist.** The three Week 10 in-class questions are the only genuine sample questions. The other 19 practice questions are written for this guide against course content and official problem spaces.

7. **No marking criteria are released for this exam.** Every answer structure, mark allocation suggestion and model answer is revision guidance.

8. **Audio and video were not re-transcribed.** Existing text transcripts were used. Approximately 218 MB of audio and 1.3 GB of video went unprocessed; if the Week 8 lecture recording (`新录音 85.m4a`) contains material not in the slides, it is not reflected here.

9. **The Workshop Exercise Encyclopedia was not read.** It is third-party, and the group's own notes record it exceeded the file-size limit for processing. The 4Cs activity names and numbers come from Week 2 slide 10 instead, which is the course's own reproduction and is reliable.

10. **The pre-existing `Course_Weeks_1-8_Complete_Guide_CN.html` was not used as a source.** It predates the Week 10 scope announcement, covers excluded topics, and its claims were not independently traceable. It was audited for contradictions only.

11. **The Chinese companion is a translation, not an independent write-up.** It mirrors the English structure exactly and preserves English course terminology throughout, but the English version remains the primary resource because the exam is answered in English.

---

## 13. Summary

| Measure | Result |
|---|---|
| HTML guides | 4 — full guide EN/CN, sprint guide EN/CN |
| Chapters (full guide) | 21 |
| Sections (sprint guide) | 19 — exam scope, process map, 12 exam-relevant revision areas, drawing sheets, answer skeletons, 2 mock exams, last-night checklist |
| Course-slide screenshots | 36, all captioned with file and page |
| Original diagrams | 16 inline SVG in the full guide + 4 in the sprint guide + 1 standalone printable template |
| Figures (screenshots + diagrams) | 52 in the full guide, 4 in the sprint guide, each with a captioned source |
| Tables | 83 in the full guide, 22 in the sprint guide |
| Practice questions | 22 (3 from Week 10, 19 written for this guide), plus 2 full mock exams in the sprint guide |
| Collapsible panels | 33 in the full guide, 11 in the sprint guide — model answers and optional detail only; **no definition, scope statement or source reference is hidden** |
| Week 10 topic coverage | **11 of 11** entries printed on slide 12, plus quantitative and qualitative data as a twelfth revision area (slide 30 + Practice Q3) |
| Week 10 exclusions respected | 7 of 7 |
| Source conflicts identified and flagged | 16 |
| Student examples corrected or limitation-stated | 9 of 9 |
| EN / CN structural parity | full guide: exact on all 33 measured element and class counts; sprint guide: exact on all 15. Verified after every edit |
| Printed length | full guide EN 175 pp / CN 185 pp; sprint guide EN 43 pp / CN 40 pp |
| Source files modified | **0** — verified by MD5 comparison against the in-folder mirror, before and after rendering |

---

## 14. Revision after external review

An external review of this pack was carried out and its findings were applied in place. The existing files were revised, not rebuilt: the 21-chapter structure, the white theme, the sticky navigation, the local search, all existing screenshots and diagrams, the EN/CN structural parity, the visible source labels, offline operation and print support were all preserved.

**The user confirmed that the existing screenshots display correctly. The review therefore focused on file content, source consistency, HTML structure and code-level issues.** No screenshot or diagram asset was replaced, regenerated, recompressed, deleted or reorganised. The only visual change is the wireframe annotation described in item 8 below, which was a wording change to an inline SVG label — its geometry was untouched.

### 14.1 What the review found, and what was done

| # | Finding | What was done |
|---|---|---|
| 1 | **The file inventory's headline total was wrong.** It stated 269 files across 39 directories, while the per-type list beside it summed to 283 | Recounted programmatically. The correct figure is **284 files across 44 directories**; the script asserts that the per-type counts sum to the total before printing. The counting method is now stated in full, including the confirmation that the output folder `INFS3700_Final_Exam_Revision_Pack\` was excluded from the walk. The old figure is recorded as withdrawn so a reader who saw it knows it changed |
| 2 | **Storyboards were partly attributed to Week 5.** The inventory described the Week 5 deck as covering "wireframes, fidelity, storyboards", and one coverage row named Week 5 as the primary storyboard source | Corrected throughout. The **primary source is the Week 4 Lecture, slides 43–50**; the **Week 7 Lecture, slide 10** is the supporting application (storyboards as a concept-testing artefact). The Week 5 deck contains no storyboard material and is no longer cited for any storyboard claim. The chapter itself already used Week 4 — the error was in the inventory and coverage tables |
| 3 | **The Week 9 reconciliation was incomplete.** Passages written before the lecture deck arrived still described it as absent, or the third blueprint line as external | Every such passage was located by full-text search and rewritten. §7.1 now opens with a numbered **final position** that supersedes everything earlier in the pack: the Week 9 lecture deck is the primary source for Chapter 14; all three lines are course terminology and none needs a disclaimer; the two-line templates on slides 34–35 are a **presentation variation inside the course's own deck**, not evidence that the third line is external; and the tutorial preparation file, case study and transcripts remain valid supporting sources — the tutorial-prep lane wording is still recommended for the P&P task because that is the wording the task uses |
| 4 | **Malformed Markdown.** A row in the individual-assignment table was missing a cell; three rows of the source-conflicts table were missing their last cell; §5's sub-sections ran 5.1, 5.3, 5.2 | All fixed. A table-shape check now runs over all three markdown files and reports zero ragged rows |
| 5 | **The practice-question count was ambiguous**, reading as though 22 were additional to the course's 3 | Restated as **22 in total: the 3 printed on the Week 10 slides plus 19 written for this guide** |
| 6 | **Unsupported predictions about the paper.** The dashboard asserted that every sub-part is a written answer, and that Question 3 would be an extended case study with an artefact to draw | Removed. The guide now states what the slides state — the mark split — and notes that the slides do not say which sub-parts are written and which are drawn, that "drawing out artefacts" is one of the four named formats, and that planning should follow the published marks rather than a predicted topic |
| 7 | **Speculative phrasing presented as fact** — "most likely", "designed to lose marks", "free marks", "most likely to separate a strong answer" | Every instance was rewritten as an observation about the source rather than a prediction about marking. Where a genuine inference remains useful it is now labelled `Guide inference — not stated by the lecturer.` / `本指南推断——讲师并未明确说明。` Practical recommendations that follow directly from the slides are not disclaimed — over-labelling would be as misleading as under-labelling |
| 8 | **Reading-time guidance assumed you may write during reading time**, which depends on rules the course does not publish | Replaced with: *"During reading time, identify the mark value and command verb for each sub-part. Record notes only where the examination interface or invigilator explicitly permits it."* The surrounding list is labelled as guide inference, and states that whether annotation is allowed depends on the venue, the invigilator and the Inspera interface. One SVG annotation reading "Free marks." was reworded for the same reason |
| 9 | **The example-reuse rule was too strict.** The guide told the student never to use an example twice | Relaxed. One substantial project may support several sub-parts — the group workshop alone can evidence participant criteria, workshop planning, facilitation, synthesis and prioritisation. What must change each time is **the point the example is making**; retelling the same story for the same reason is what adds nothing |
| 10 | **The Chinese guide declared `<html lang="en">`** | Both Chinese files now declare `lang="zh-CN"`; both English files remain `lang="en"` |
| 11 | **Companion links were incomplete**, and the Chinese guide linked to itself rather than to the English one | All four guides now cross-link to the other three. Every internal link is checked against the filesystem by the regression script |
| 12 | **Chinese accessibility text was still English.** All 36 `alt` attributes and all 16 SVG `aria-label` descriptions were untranslated | All 52 translated, course terminology preserved in English. Navigation, search, previous/next and disclosure-panel labels were already Chinese and were verified. **The visible text inside the SVG diagrams is deliberately left in English** so that the two language versions remain visually identical and the diagrams keep the lecturer's terminology; the parity check now compares SVG geometry with `aria-label` excluded, and separately asserts that every Chinese `aria-label` and `alt` really is translated |
| 13 | **A regex bug in the search.** One global regular expression was shared between `.test()` in the TreeWalker filter and `.exec()` in the highlighter. A global regex carries `lastIndex` between calls, so the filter rejected roughly every second matching text node | Split into `testRe` (non-global, `'i'`) for the filter and `markRe` (global, `'gi'`) for wrapping, in all four files. Verified in a browser: searching "service" in the full English guide now marks **289 of 289** occurrences and "the" marks **4,043 of 4,043**, counted independently against the DOM. Three consecutive searches return an identical count |
| 14 | **Found while testing the above:** the two-character minimum search length made single-character Chinese searches silently do nothing — but one CJK character is usually a whole word | The minimum is now language-aware: **1 character for CJK, 2 for Latin**. Searching 的 in the Chinese guide now returns all 2,382 occurrences; a single Latin letter still does nothing, and the counter stays blank rather than misreporting "0" |

### 14.2 What was added

`05_INFS3700_Exam_Sprint_Guide_EN.html` and `06_INFS3700_Exam_Sprint_Guide_CN.html` — a condensed guide for the last few days, sharing the full guide's stylesheet and behaviour script verbatim so the four files cannot drift apart. Contents: the exam scope on one page; a one-page process map; all **12** Week 10 topics, each with a definition, purpose, timing, steps, one example, one mistake and one answer skeleton; three printable artefact drawing sheets; answer skeletons for the paper's five mark values (5, 8, 10, 12, 15); **two complete mock exams** matching the published structure — 25 (10 + 15), 20 (8 + 12), 55 (10 + 5 + 10 + 15 + 15) — each with a concise answer plan per sub-part; and a last-night checklist.

The mock exams carry an explicit warning wherever they appear: they were written for this guide, they are not past papers, they were not written or seen by course staff, and the only thing they copy from the course is the mark structure on Week 10 slide 23.

The full guides were **not shortened**. Both grew slightly as a result of the corrections above.

### 14.3 How the result was verified

- **All four HTML files were opened in a browser engine** (headless Edge) and driven through 34 automated checks each: rendering, nav/section consistency, exhaustive match marking against an independent DOM count, counter behaviour, next/previous stepping including wrap-around, whether the current match is actually inside the viewport, revealing a match buried in a collapsed panel, Shift+Enter reverse stepping, the no-results state, clearing, repeat-search stability, the print handlers, and broken images. **All four scored 34/34.**
- **All four were printed to PDF** and inspected: no navigation, top bar or search chrome leaks into the print; collapsed panels expand; page counts are full guide EN 171 / CN 181, sprint guide EN 40 / CN 36.
- **A regression script** re-checks tag balance, internal links, image existence and `alt` text, offline safety, `lang` attributes, companion links, EN/CN parity, Markdown table shape, and the presence or absence of every phrase this revision was meant to add or remove.

### 14.4 Known deviations from the review's brief

- **The sprint guides print at 43 and 40 pages, against a target of roughly 20–30.** The content the review specified — 12 topics × 7 elements each, plus an exam-scope page, a process map, three drawing sheets, five answer skeletons and two complete mock exams with answer plans — does not compress below this without dropping named items. A compact print mode was added (tighter margins, smaller type, long answer plans allowed to flow across page breaks), which brought the English guide from 49 pages to 40. Cutting further would mean removing required content, so it was not done.
- **The visible text inside the Chinese guides' SVG diagrams remains English**, as described in item 12 above. This is a deliberate carry-over of the original design, not an oversight.


---

## 15. Lecture and tutorial transcripts reviewed

Eight automatic transcripts were added to the course folder after the previous revision and all eight were **read in full**: `lecture transcript\wk1.txt`, `wk2.txt`, `wk4.txt`, `wk5.txt`, `wk7.txt`, `wk9.txt`, `wk10.txt`, and `WEEK10\tut transcript raw.txt`. They cover **Weeks 1, 2, 4, 5, 7, 9 and 10 lectures plus the Week 10 tutorial**. There is no transcript for Week 3, Week 6 or Week 8.

**Authority.** All eight are tier **A5** — supporting sources. Week 10 slides remain authoritative for exam scope, format, inclusions, exclusions and mark structure; official briefs remain authoritative for assignment requirements; lecture slides remain the primary source for definitions, frameworks and terminology. A transcript is used to clarify, expand or contextualise, never to overrule.

### 15.1 Method

111 candidate passages were extracted and each was then checked independently against three questions: **is the quoted string literally in the file**, **is the point already in the guide**, and **does it agree with the relevant deck**. **30 passages survived; 81 were rejected.** Every surviving quotation was re-verified by literal string search before it was written into a guide. Where the transcription garbles a word, the guide reproduces the garble and gives the intended word in brackets — it does not silently repair speech.

### 15.2 The additions that changed the guide most

| # | Addition | Where | Why it mattered |
|---|---|---|---|
| 1 | **Two of Question 3's five sub-parts require a hand-drawn answer.** The lecturer also indicated, tentatively, that students were expected to receive four sheets of paper | Ch. 0; sprint §A | The slides give only the mark split. This turns "a sub-part may ask you to draw" into a specific, plannable fact, and makes the artefact chapters and drawing sheets compulsory revision rather than optional |
| 2 | **The artefact is drawn on paper, photographed and uploaded, and the two hours includes all of it** | Ch. 0; sprint §A | A scheduling constraint no slide states |
| 3 | **An interview transcript is a source, not a data type; open-ended ≠ qualitative** | Ch. 9; sprint topic 12 | The tutor corrected exactly this while marking student answers to Week 10 Practice Q3 — the same question a student is likely to face |
| 4 | **The 5Ps are applied to one specific service, never to a whole industry** | Ch. 13; sprint topic 10 | Explains *why* generic 5P answers fail, which the slide's five definitions cannot |
| 5 | **All four Cs sit inside the first diamond only** | Ch. 5; sprint topic 02 | Answers an integration question the slides leave open, and the lecturer raised it precisely because a past student got it wrong |
| 6 | **"F pattern" is spoken course terminology** | Ch. 8, Ch. 19; sprint | **Corrects this pack's own earlier guidance**, which had treated both viewing-pattern names as outside vocabulary |
| 7 | **Bullets are welcome; essay style is not expected** | Sprint §D | Answers a question students actually asked, with the tutor's own words |

### 15.3 Conflicts found

| Conflict | Resolution |
|---|---|
| **Pen vs pencil.** The Week 10 what-to-bring slide prefers a pen; the tutor read that line aloud and disagreed, preferring a pencil because it erases | Recorded openly in Ch. 0 and sprint §G. The tutor closed the question himself — "Pen or penciling doesn't really matter that much… make sure that you can make your drawing **clear to us**" — so the guide says bring both and names legibility in the photograph as the real requirement. Logged as conflict 15 in §7 |
| **Week 9 elements slide lane order.** The lecturer says the printed order is "flipped" and that customer actions comes first, adding "it doesn't matter too much" | Recorded, not used to rewrite the slide. Logged as conflict 16 in §7 |
| **Survey question order.** The Week 2 slide says ask general questions first; the lecturer describes doing the opposite in practice, then concedes "in general textbook version p ask more general question first" | **Not added to the guide.** He does not retract either position, so there is no unambiguous correction to record, and the topic is peripheral to the examinable list |
| **"Design thinking… It's not an iterative process"** (Week 1) | **Rejected as a probable misspeak or ASR error** for "not a linear process" — the surrounding sentence says it "will become more linear towards the end". Adding it would have contradicted the course's own iterative framing on the strength of one garbled clause |

### 15.4 Passages deliberately excluded, and transcript-quality limits

- **81 of 111 candidates were rejected.** The large majority were rejected because the guide already made the point from the slides — the transcripts largely restate the decks, which is the expected result for a guide already built from them.
- **The Week 2 DVF passage** ("which one is visible visible, which one which one is viable, which one is desirable… that's the desirability viability and visibility") is mangled beyond safe reconstruction and is quoted nowhere.
- **Guerrilla testing's alternative name** is rendered "link testing", which is not recoverable; the concept is described without asserting the name.
- **The Week 1 file renders *intuitive* as "intrusive" throughout**, which inverts the meaning. Nothing from those passages is quoted as terminology.
- **The Week 2 file has no speaker labels at all** and is a single 88,330-character line. Every attribution from it is an inference from content and is labelled as such.
- **The Week 4 file never renders the second viewing pattern as "Z"**, which is why this pack now treats that name as unconfirmed rather than simply forbidden.
- **`WEEK10\tut transcript raw.txt` has no speaker labels**; the tutor is identified from context (he runs the session and critiques posted answers). Where a passage might be a student speaking, the guide says so.
- **The Week 10 tutorial audio (`WEEK10\新录音 11.m4a`, 109 MB) was not re-transcribed.** The supplied `.txt` was used.

**Unresolved ambiguity worth stating.** The Week 10 lecturer says the drawing questions relate to artefacts practised "in the tutorials throughout the 9 weeks" but explicitly declines to name them — "I'm not gonna give away what questions that might be". This pack therefore prepares **all three** drawable artefacts (wireframe, storyboard, service blueprint) rather than guessing which two are meant.

---

## 16. Final targeted revision

A second external review was applied after the transcript work. Each item below was inspected against the actual current wording and the course files before any change was made; where the review's suggestion did not survive that check, the decision and its evidence are recorded in §16.2.

### 16.1 What was done

| Item | Outcome |
|---|---|
| **Reading-time guidance reconciled** | The corrected wording now reads identically in all four guides: *"During reading time, identify the mark value, command verb and relevant case evidence for each sub-part. Record or annotate anything only where the examination interface or invigilator explicitly permits it."* Chapter 18's "Managing the paper" list still told the student to *write* mark values and *underline* case evidence — that was the last surviving instance and it is gone. The principle now also appears in the sprint guide's mock-exam instructions |
| **Example-reuse guidance reconciled** | Chapter 16's opening still said "one per concept, so you never use the same one twice", and Chapter 18 still said "allocate your personal examples so none is reused". Both replaced. The guide now states that the same project may support several questions provided each use demonstrates a different concept and the analysis is adapted — with the workshop's own spread across participant criteria, workshop planning, facilitation, synthesis and prioritisation as the worked case |
| **Unsupported marking predictions removed** | **43 passages** rewritten across both languages: "where most of the marks sit", "the marks are in", "earns marks", "free marks", "a cheap mark", "why it scores better", "what a marker can see and reward", "recover more marks", "a reliable few marks", "the annotation that carries the marks", and their Chinese equivalents (分数在…、真正拿分、能拿分的、稳拿的、更值分). Each became a statement about answer completeness, or a labelled guide inference, or was deleted. **Two flagged phrases were deliberately kept**: "the highest-value moment in the funnel" and "the highest-value point in the journey" both describe the *business* case in a worked example, not exam marks |
| **Mock Exam 1 blueprint reasoning corrected** | Q3(d) previously located the failure "at the line of visibility". It now reads: the failure **begins in the support process**, when check-in is treated as equivalent to confirmed shelf availability; that premature status propagates across the line of internal interaction into backstage, through the backstage-to-frontstage handoff, and across the line of visibility into the customer-visible notification. A second bullet states plainly that the line of visibility marks where the fault becomes *visible*, not what *caused* it |
| **Mock Exam 2 bias terminology corrected** | The answer had labelled "moderator effect" and "recruiting only active users" as the two biases. Week 7 slide 39 was checked directly: it names **Social desirability bias · Hawthorne effect · Culture bias · Confirmation bias** and gives no definitions. The answer now leads with **two course-named biases that genuinely fit the scenario** — social desirability (club volunteers testing their own club's app) and the Hawthorne effect (observed persistence concealing the abandonment the case is about) — and keeps moderator influence and sampling as **explicitly labelled practical risks that are not among the four named** |
| **Twelve-area framing clarified** | Slide 12 prints **eleven** entries, because "Workshop / interview planning" is one combined item. Headings changed from "The 12 possible topics" to "12 exam-relevant revision areas", with a note stating that eleven are listed on slide 12 and that quantitative and qualitative data is included because slide 30 revises it directly and Practice Question 3 examines it. Applied to the sprint heading, the process-map lead, both mock-exam preambles and the area table's footnote, in both languages |
| **Obsolete Source Map claim removed** | The struck-through "line of internal interaction" row is gone from the active provenance table. Section 14 now opens with the settled position — all three lines are in the Week 9 deck, some templates draw two, the variation is internal to the course, and no line should be labelled a student addition — and points to this audit for the history |
| **Audit contradictions removed** | The "where §14 and an earlier section disagree" override is gone. Sections 1–13 were rewritten to match the current files: the census, the files-read count, the authority table, the uncertainty table's F-pattern row, the omitted-claims list, the limitations list and the summary table |
| **Print heading pagination improved** | The reported defect was reproduced: in the pre-fix render, `Topic 01 · Assignment tasks` sat at 90% of page height on page 3 of the English sprint guide, with its heading on the next page. Print CSS now keeps `.sechead` whole and glues every heading to the block after it, in both the legacy `page-break-*` and modern `break-*` properties. After the fix, **no heading in either sprint guide sits in the bottom 12% of a page**, and no section kicker is separated from its heading. Cost: two pages in English, one in Chinese |
| **Embedded U+FEFF removed** | **7 embedded U+FEFF characters** were found inside document bodies — 1 in the English full guide, 6 in the Chinese. Cause traced: PowerShell's `Set-Content -Encoding utf8` writes a BOM, and a BOM on any fragment after the first lands mid-document when fragments are concatenated. Nine fragment and markdown files were de-BOMed, and **both build scripts now read with `utf-8-sig`, strip any remaining U+FEFF, and assert that none survives into the output**. Current count in all seven deliverables: **0** |
| **EN/CN parity verified** | Every correction above was applied in both languages. Parity is asserted programmatically after each build across 33 element and class counts for the full guides and 15 for the sprint guides, plus section ids, image paths, SVG geometry, and the requirement that every Chinese `alt` and SVG `aria-label` really is translated |
| **Horizontal overflow fixed (found during validation, not in the review)** | Driving the page sideways in a real browser viewport showed the two full guides could be scrolled **35&nbsp;px horizontally** at phone width — the sprint guides could not. No element's content overflowed it and every table was correctly inside a `.tw` scroller, so the cause was residual grid/sub-pixel overflow. Fixed with `overflow-x:clip` (not `hidden`, which would have created a scroll container and broken the sticky top bar) plus `min-width:0` on grid children. **Scoped to `@media screen`** after the first attempt added 45 pages to the printed full guides. Re-measured: `maxScrollX` is now 0 at both 1400&nbsp;px and phone width in all four guides, and print length is unchanged |
| **Search and print regression** | Reported in §16.3 |

### 16.2 Reviewer suggestions modified or not applied

| Suggestion | Decision | Evidence |
|---|---|---|
| Replace "moderator effect" and "recruiting only active users" with course-named biases, **or** describe them accurately as practical risks | **Both, rather than either.** The answer now names two course biases that fit the scenario *and* keeps the two practical risks with an explicit statement that they are not among the four named | Week 7 slide 39 read directly: it names four biases and defines none. Naming only the practical risks would have left a scenario-appropriate course term unused; replacing them entirely would have discarded two real threats to that test's validity |
| Remove language matching "most likely", "highest value", "line of visibility", "active users", "moderator effect" | **Applied selectively.** "Highest-value moment in the funnel" and "highest-value point in the journey" were kept — both describe business impact inside a worked example, not exam marks. "Line of visibility" occurs 100+ times as legitimate course terminology and was corrected only in the one place it was misused as a root cause | The reviewer's list is a search net, not a list of defects; each hit was read in context |
| Target 20–30 printed pages for the sprint guides | **Not met, and not forced.** They print at **43 pages (EN)** and **40 (CN)** | The content the reviewer specified — 12 areas × 7 elements, an exam-scope page, a process map, three drawing sheets, five answer skeletons and two full mock exams with plans — does not compress below this without dropping named items. The heading-cohesion fix and the new transcript findings each added pages. The review itself says to prefer readability; type sizes were left legible |
| Ensure no `U+FEFF` remains, "a BOM at the absolute beginning is acceptable" | **Applied more strictly than asked.** Leading BOMs were removed too, not just embedded ones | A leading BOM on a *fragment* is exactly what produced the embedded ones. Removing them at source is the only fix that holds |
| Verify the reading-time and example-reuse wording is not left old "in one chapter" | **Confirmed by full-text search over the fragments, not the built files.** The built HTML is generated, so a fragment is the only place a stale phrase can survive a rebuild | Both searches now return zero outside deliberate quotation of what was removed |

### 16.3 Validation performed after the edits

- **All four HTML files were opened in a browser engine** (headless Edge) and driven through the full behavioural suite: rendering, nav/section consistency, exhaustive match marking counted against an independent DOM walk, counter behaviour, next/previous stepping with wrap-around, whether the current match is inside the viewport, revealing a match buried in a collapsed panel, Shift+Enter reverse stepping, the no-results state, clearing, repeat-search stability, print handlers, and broken images. Results in §16.4.
- **The two-regex search fix is intact** in all four files — `testRe` non-global for the TreeWalker filter, `markRe` global for wrapping — and the regression script fails the build if `var re=new RegExp` reappears.
- **All four were printed to PDF** and inspected for chrome leakage, panel expansion, heading orphans and page counts.
- **A content sweep** re-checked every phrase the review listed, and every hit was read in context rather than removed on sight.
- **Horizontal overflow was driven, not inferred**: the page was scrolled to `x=9999` in a real browser viewport at 1400 px and at phone width, and `window.scrollX` was read back. This is how the 35 px overflow was found, and how the fix was confirmed.
- **Screenshots** were taken of the changed sections (Chapter 0's new transcript panels, the Chinese data chapter, the sprint drawing sheets, Mock Exam 1, and the Chinese sprint scope page) and inspected.

### 16.4 Results

**Search and behaviour:** all four guides scored **34/34** on the behavioural suite, run against the live DOM in headless Edge — exhaustive match marking checked against an independent DOM walk (289/289 for "service", 4,043/4,043 for "the" in the English full guide), next/previous with wrap-around, Enter and Shift+Enter, Escape, clearing and re-querying, matches revealed inside collapsed panels, and the language-aware minimum search length (1 character for CJK, 2 for Latin).

**Structure:** all checks passed — valid parsing, no duplicate ids, no broken internal anchors, all companion links present and resolving, no remote CSS/JS/fonts/images, every local image resolving with non-empty alt text, correct `lang` metadata (`en` / `zh-CN`), balanced tags, every table inside a scroll wrapper, **0 embedded U+FEFF**, and **no horizontal page scroll at either 1400 px or phone width**.

**EN/CN parity:** exact on all 33 measured element and class counts for the full guides and all 15 for the sprint guides, plus identical section ids, identical image paths, identical SVG geometry, and every Chinese `alt` (36/36) and SVG `aria-label` (16/16) verified as actually translated.

**Printed length:** full guide EN **175** pp / CN **185** pp; sprint guide EN **43** pp / CN **40** pp. No chrome leaks into the print, collapsed panels expand, and no section kicker or heading is stranded at the foot of a page.


---

## 17. Final consistency and transcript-certainty pass

Every item below was checked against the current files or the underlying course materials before anything was changed. Counts were **measured**, not carried forward.

### 17.1 Inconsistencies confirmed and corrected

| # | What was wrong | Correction |
|---|---|---|
| 1 | **§10 claimed 81 tables**; the current full guides contain **83** (two were added with the Week 10 transcript panels). The same row understated `tr` (570 → **578**), `th` (211 → **217**) and `td` (1525 → **1543**) | All re-measured from the built files and corrected. The table-containment check was also strengthened: it now walks the tag nesting to prove every table is *inside* a `.tw` wrapper, rather than comparing counts — equal counts would not have proved containment |
| 2 | **§10 still recorded "EN 160 pages, CN 169 pages"** — two revisions out of date | Corrected to the measured **EN 175 / CN 185**, and the sprint guides (**EN 43 / CN 40**) added to the same row |
| 3 | **§12 said the sprint guides print at 40 and 36 pages** | Corrected to **43 and 40** |
| 4 | **§5 said "All twelve named topics… are covered"** and **§13 said "12 of 12"** — both imply slide 12 prints twelve entries | Rewritten. Slide 12 prints **eleven** entries because *Workshop / interview planning* is one combined item. The audit now says **11 of 11 printed entries**, plus quantitative and qualitative data as a twelfth revision area justified by slide 30 and Practice Question 3 |
| 5 | **§7.1 said the obsolete Blueprint claim was "withdrawn and struck through rather than deleted"** — no longer true. The final targeted revision removed it from the Source Map's active table entirely | Corrected to describe both steps: struck through first, then removed. The Source Map now contains **no `~~` strikethrough and no "withdrawn" row**; it opens section 14 with the settled position instead |
| 6 | **The source-conflict table ran 13, 15, 16, 14** | Reordered to ascend: 13, 14, 15, 16 |
| 7 | **The A5 tier said "tutor statements" only**, although seven of the eight new transcripts are **lecture** transcripts | Widened in both the Inventory and the Source Map to *spoken course clarification from a lecturer or tutor, plus tutor written feedback — supporting rather than primary evidence* |
| 8 | **The confidence key was internally impossible**: `High` was defined as "printed verbatim on a slide", yet ten of twelve transcript rows were graded High | Resolved by separating the two things the column was conflating. **Type** already carries authority, so **Confidence** now rates *extraction* confidence. A transcript row can be High confidence while remaining A5 authority. Every transcript row was then re-graded against the new definition, not just the key |
| 9 | **A literal `<h1>` inside a print-CSS comment** inflated tag counts and could trip a naive validator | Reworded, the same fix previously applied to a literal `<details>` in a JS comment |

### 17.2 Marking and marker-behaviour language — second and third waves

The earlier pass caught the *"where the marks sit"* family. This pass used a different phrase list and found **45 further instances** in constructions the earlier regexes missed: *"are the marks"*, *"that column is the marks"*, *"Why it loses marks"*, *"cost you the whole question"*, *"lose a quarter of the question"*, *"loses a third of the marks"*, *"scores poorly"*, *"Why it costs you"*, *"what an examiner is looking for"*, *"the marker cannot tell"*, *"leaving the comparison to the marker"*, *"Markers do not carry marks"*, *"caps out early"*, *"carries as many marks as the thing itself"*, and their Chinese equivalents (得分点、丢分、失分、分数所在、分值相当、评卷人…).

Two of these were **inside the answer-spine SVG** (*"trim from the middle when marks are low"*, *"Most lost marks are here"*). Because that SVG is byte-identical across EN and CN by design, both labels were replaced with the same new English strings in both files, preserving SVG parity.

One was a genuine **EN/CN divergence**: the Chinese "阶段名称不给分" sentence had been neutralised in the earlier pass, but its English counterpart — *"Markers do not carry marks for naming stages. They carry marks for showing…"* — had not. Fixed.

**What deliberately remains, and why.** Wording that is *attributed* rather than asserted is kept:

- The **tutor's own description of how he marks** ("bullet points… might make things clearer"; "it's more like *we mark* based on the points like the criteria") is retained in the sprint guide, labelled **Tutor explanation** with a timestamp, and immediately followed by the note that Week 10 slide 15 still says **no marking criteria will be provided to students**. It is presented as guidance on presenting an answer, not as a rubric.
- Statements of **published fact** — "no marking criteria released", "how marks are distributed inside a sub-part" appearing in the *what the slides do not say* list, the 1–5 Impact/Importance scoring from Week 3 — are untouched.
- The **mark values themselves** (Q3 is 55 marks; the 5/8/10/12/15 sub-part split) are printed on slide 23 and are stated plainly throughout.

### 17.3 How the twelve revision areas are now described

The sprint guides keep **twelve sections**; only the attribution changed. Corrected in the `<meta name="description">`, the top-bar branding, the sidebar group heading, the section heading, the area table's footnote, the process-map lead, and both mock-exam preambles — in both languages:

> **12 exam-relevant revision areas.** Eleven are explicitly listed on Week 10 slide 12. Quantitative and qualitative data is included as an additional revision area because Week 10 slide 30 revises it directly and Practice Question 3 explicitly examines it. Slide 12 prints eleven entries rather than twelve because *"Workshop / interview planning"* is a single combined entry.

Chinese: **12 个与考试相关的复习领域** — 其中 11 项由 Week 10 slide 12 明确列出，另加入定量与定性数据作为额外复习领域。

### 17.4 How transcript certainty is now represented

Re-reading `lecture transcript\wk10.txt` separated three claims that had been presented at the same strength:

| Claim | Source wording | Treatment |
|---|---|---|
| **Two drawn sub-parts, both inside Question 3** | "in regard to your drawing questions… it will be um **within the question three**. So… **two out of these 12345 questions are drawing questions**" — unhedged, declarative | **Kept as a strong spoken clarification**, confidence **High**, clearly separated from printed slide content |
| **Four sheets of paper** | "So you'll be given, **I think** each one of you will be given 4 pieces of paper" — hedged on the number | **Downgraded.** Split into its own Source Map row (15.1b) at confidence **Medium**, and reworded in both guides as what students were *expected* to receive rather than a confirmed fact |
| **Theoretical questions scenario-framed** | "there is going to be a **combination of** theoretical questions. So imagine if you were… this business" — one illustrative framing, and "combination" implies variety | **Softened** from *"are scenario-framed"* to *"may be scenario-framed"*, and row 15.4 regraded from High to **Medium** |

Separately, *"two sub-parts cannot be answered in prose"* was too absolute — a drawn sub-part still needs labels and annotation. It now reads **"require a hand-drawn artefact and cannot be answered with prose alone"**, in all four guides.

### 17.5 Suggested issues not applied, and why

| Suggestion | Decision | Evidence |
|---|---|---|
| Remove language matching the supplied marks phrase list | **Applied selectively, after reading each hit in context.** Kept: the attributed tutor explanation of his own marking; the published fact that no criteria are released; "how marks are distributed inside a sub-part" in the *what the slides do not say* list; the Week 3 1–5 scoring scales; and the published mark values | The review itself says not to delete every occurrence. Deleting the attributed tutor passage would remove the only direct evidence the pack has on how answers are approached, and deleting the published mark values would make time planning impossible |
| Reduce transcript claims to Medium across the board (the alternative offered for the confidence key) | **Not taken.** Separated authority from extraction confidence instead | Grading a flat, unhedged, literally-verified statement as Medium purely because it was spoken would lose the distinction between that and a hedged "I think". The Type column already records that all of them are A5 |
| Verify the Blueprint claim is no longer in the Source Map | **Verified, and it was already clean** — 0 strikethroughs, 0 "withdrawn" rows. The defect was in the *audit's description* of it, which is what was fixed | Full-text check of `01_SOURCE_MAP.md` |
| Re-check the mock-exam corrections | **Verified intact, not re-edited.** Mock 1 still locates the failure in the support process; Mock 2 still names the four Week 7 biases and still states that moderator influence and sampling are *not* among them | The new transcripts contain nothing bearing on either, so there was no evidence to justify changing them |
| Compress the sprint guides toward 20–30 pages | **Not done.** They print at 43 and 40 | The review says to prefer readability over the page target, and the content is the content the reviews asked for |

### 17.6 Final measured state

**HTML structure** — full guides (EN and CN identical): **21** sections · **83** tables · **52** figures · **52** figcaptions · **36** images · **33** details · **16** SVGs · **578** `tr` · **217** `th` · **1543** `td` · **21** nav links. Sprint guides (EN and CN identical): **19** sections · **22** tables · **4** figures · **0** images · **11** details · **4** SVGs · **105** `tr` · **67** `th` · **220** `td` · **19** nav links.

**Print** — full guide **EN 175 pp / CN 185 pp**; sprint guide **EN 43 pp / CN 40 pp**. No chrome leaks; collapsed panels expand; **no section kicker or heading is stranded** (0 headings in the bottom 12% of a page in either sprint guide).

**Parity** — exact on all 33 measured element and class counts for the full guides and all 15 for the sprint guides, plus identical section ids, image paths and SVG geometry, and every Chinese `alt` (36/36) and SVG `aria-label` (16/16) verified as actually translated.

**Technical validation** — all four files opened and driven in a browser engine: **34/34** each. Valid parsing; no duplicate ids; no broken internal anchors; correct `lang` (`en` / `zh-CN`); companion links present and resolving; every local asset resolving with non-empty alt text; no remote CSS, JS, fonts or images; **0 embedded U+FEFF**; the two-regex search fix (`testRe` non-global / `markRe` global) intact and verified exhaustive; print CSS loading; and **no horizontal page scroll** — `maxScrollX` is 0 at both 1400 px and phone width in all four guides.


---

## 18. Final marking-language and transcript-confidence cleanup

Each item was checked against the source before being changed. Counts were re-measured from the built files after editing, and all four guides were reopened and driven in a browser afterwards.

### 18.1 The four-sheet claim, separated from the Q3 drawing claim

The Source Map had been carrying the same four-sheet statement at two different confidence levels: row 15.1 asserted it at **High** alongside the drawing claim, while row 15.1b acknowledged at **Medium** that the lecturer hedged. That is not a defensible pair of rows, and it has been resolved:

| Row | Claim | Confidence | Why |
|---|---|---|---|
| **15.1** | **Two sub-parts require a hand-drawn answer, and both are inside Question 3** | **High** | Stated flatly: *"in regard to your drawing questions… it will be um **within the question three**. So… **two out of these 12345 questions are drawing questions**"*. No hedge. The four-sheet clause has been removed from this row |
| **15.1b** | **Students were expected to receive four sheets of paper** — stated tentatively | **Medium** | *"So you'll be given, **I think** each one of you will be given 4 pieces of paper"*. The hedge attaches to the number, not to the existence of the drawn sub-parts |

**No source in this folder confirms the sheet count**, so the guides never assert it. The dashboard panels in both full guides carry an inline note that the number is hedged while the location is not; both sprint guides say *"expect four sheets rather than count on it"* / *「四张纸」是预计，不是定数*. The audit's own §15.2 summary was reworded to match.

### 18.2 Marking-language cleanup — third and final wave

Nineteen further phrases were removed from the guides. All were **guide-written predictions** about internal mark allocation or marker behaviour, none traceable to any source:

*"a cheap, distinctive mark"* · *"Content that scores"* · *"It is free and…"* · *"that is where the 8 marks are"* · *"every mark on this topic is for interpretation"* · *"The marker does not need your whole project"* · *"do not make the marker infer it"* · *"A 15-mark sub-part is where a drawn artefact most often appears"* · *"the single most predictable content in the course"* — and the Chinese equivalents 低成本又有辨识度的一分 · 白送的分 · 这 8 分就在那里 · 每一分都是给解读的 · 评分人不需要 · 不要让评分人自己去推断 · 最常出现在 15 分档 · 最可预测的内容.

Two were **EN/CN divergences** where one language had already been neutralised and the other had not: *"Content that scores:"* (CN was already *「一份完整答案该包含的内容」*), and the *"marker does not need / do not make the marker infer"* pair.

Each became neutral answer-quality language — *a concise, distinctive sign of understanding*; *A complete response should include*; *It takes only one line, and it is the thing tutors corrected most*; *that is the core comparison the question requires*; *focus on interpreting the data and explaining its implications*; *Keep the project context brief and focus on applying the concept*; *state the connection explicitly rather than leaving it to be inferred*.

**What deliberately remains.** The pack still contains marking-related wording in three forms, and this is intentional:

1. **Published fact** — the mark values on Week 10 slide 23 (Q3 = 55, sub-parts 10/5/10/15/15), and the statement that *no marking criteria will be provided*.
2. **Attributed tutor commentary** — the Week 10 tutor's own description of how he approaches marking (*"the bull reports might make things clearer… it's more like we mark based on the points like the criteria"*). It appears **7 times across the pack**, always under a **Tutor explanation / 导师口头说明** label with a timestamp, and always adjacent to the reminder that Week 10 slide 15 says no criteria are released to students. It is presented as guidance on presenting an answer, never as a rubric.
3. **Course content** — Week 3's 1–5 Impact/Importance scoring and feasibility scoring.

It would be inaccurate to say all marking-related language was removed. What was removed is every *unsupported prediction*; what remains is published, attributed or course-taught.

### 18.3 The 15-mark drawing prediction

The sprint guides claimed a drawn artefact *"most often appears"* in a 15-mark sub-part. **Nothing supports this.** The evidence establishes only that Question 3 has five sub-parts worth 10, 5, 10, 15 and 15, and that two of them are drawing questions — the transcript never links drawing to a mark value. Replaced in both languages with:

> A drawn artefact may appear in one of Question 3's sub-parts, but the transcript does not identify which mark values those sub-parts carry — so do not assume it will be a 15. Where drawing is required, use the artefact as the main application block and put the reasoning into its labels and annotations.

The surrounding answer skeletons were checked for the same implication; none remained.

### 18.4 Two other wording corrections

- **"the single most predictable content in the course"** (UX vs Service Design) → *"a strongly supported revision priority — it appears in both the Week 8 P&P task and Week 10 Practice Question 1"*. The topic's prominence is real and evidenced; the prediction about the exam was not.
- **Inventory, tutor feedback** — *"it tells you what a marker actually rewarded"* implied that group-assignment feedback reveals the final-exam scheme. Now: *"The highest-value student-adjacent source, because it records what the tutor praised, criticised and asked the group to improve. It is feedback on a group assignment, not evidence of how the final exam is marked."* The **A5** classification is unchanged.

### 18.5 Historical counts

One historical figure remains in §17.1 — *"§10 claimed 81 tables; the current full guides contain 83"*. It is retained in that form because it names the section that made the claim **and** states the current figure in the same sentence, which is less ambiguous than a bare "at that revision stage" prefix. Sections 1–13 contain no stale count: every structural number there was re-measured and matches the built files.

### 18.6 Final measured state

**Structure** — full guides (EN and CN identical): **21** sections · **83** tables · **52** figures · **52** figcaptions · **36** images · **33** details · **16** SVGs · **21** nav links · 578 `tr` · 217 `th` · 1543 `td`. Sprint guides (EN and CN identical): **19** sections · **22** tables · **4** figures · **0** images · **11** details · **4** SVGs · **19** nav links · 105 `tr` · 67 `th` · 220 `td`.

**Print** — engine: **Microsoft Edge 151.0.4129.59 (Chromium), headless `--print-to-pdf`, default A4 print viewport**. Full guide **EN 175 pp / CN 185 pp**; sprint guide **EN 43 pp / CN 40 pp**. Page counts will differ on other engines and paper sizes. No chrome leaks into the print; collapsed answer plans expand; both mock exams and all three drawing sheets are present in the printed output; **0** headings stranded in the bottom 12% of a page and **0** section kickers separated from their headings.

**EN/CN parity** — exact on all 33 measured element and class counts for the full guides and all 15 for the sprint guides, plus identical section ids, image paths and SVG geometry. The attributed tutor commentary appears **5 + 2 times in English and 5 + 2 in Chinese**. No claim, uncertainty level, source label, example, warning or mock-exam reasoning differs between languages.

**Technical validation** — all four files reopened and driven in a browser engine: **34/34 each**. Valid parsing; no duplicate ids; no broken internal anchors; correct `lang` (`en` / `zh-CN`); companion links present and resolving; every local image resolving with non-empty alt text; no remote CSS, fonts, scripts or images; **0 embedded U+FEFF**; JavaScript executes without error; search works, with the **non-global `testRe`** used for the TreeWalker filter and the **global `markRe`** used for marking, verified exhaustive against an independent DOM count; **no body-level horizontal overflow** (`maxScrollX` = 0 at 1400 px and at phone width in all four guides).

### 18.7 Remaining limitations

Unchanged from §15.4 and §16.2, and none introduced by this pass:

- **ASR quality.** Week 1 renders *intuitive* as "intrusive" throughout; Week 2 has no speaker labels and is a single 88,330-character line, so every attribution from it is a labelled inference; Week 4 never renders the second viewing pattern as "Z". Quotations reproduce the error and gloss the intended word in brackets.
- **Timestamps** exist only in `WEEK10\tut transcript raw.txt`; the seven lecture transcripts are cited by speaker plus anchor phrase.
- **Audio not re-transcribed** — `WEEK10\新录音 11.m4a` (109 MB) and the three earlier recordings; the supplied `.txt` files were used.
- **The lecturer declines to name which two artefacts are drawn**, so all three drawable artefacts stay prepared, and the mark values of the drawn sub-parts remain unknown.
- **No marking criteria exist** for this exam, so every answer structure in the pack is revision guidance.
- **The sprint guides exceed the aspirational 20–30 page target** at 43 and 40 pages; type sizes were left readable rather than compressed.

---

*Companion files: `00_FILE_INVENTORY.md`, `01_SOURCE_MAP.md`, `02_INFS3700_Final_Exam_Revision_EN.html`, `03_INFS3700_Final_Exam_Revision_CN.html`, `05_INFS3700_Exam_Sprint_Guide_EN.html`, `06_INFS3700_Exam_Sprint_Guide_CN.html`.*