Answer Validation Report
Rules applied: never treat marketing claims as fact; never manufacture consensus; preserve disagreements; a provider-supplied answer is "probable", not "verified", unless independently corroborated; no APM-published answer key exists publicly for any sample paper, so no answer in this corpus is labelled verified against an official key.
Answer status counts (148 raw occurrences)
- probable — 62. Provider-supplied model answers aligned with the official syllabus (QA APMP-5, Wellingtone, Parallel, APM 2021 Sample Answer Document) + analyst-inferred keys with confidence ≥ 0.8.
- disputed — 3. Genuinely ambiguous without the official key:
- SP1 Q10 — PESTLE element for "new government regulations": Political vs Legal (both defensible in APM-style material; classic P→government policy/regulation, but Legal also plausible). Analyst pick: Political, confidence 0.65.
- SP2 Q3 — life cycle for high deliverable uncertainty: iterative vs evolutionary, and the benefit clause (priorities vs business case). Analyst pick: iterative / clearly identify priorities, confidence 0.5.
- SP2 Q27 — benefits-identification information source: project sponsor vs current supplier. Analyst pick: sponsor, confidence 0.55.
- unsupported — 83. No supplied answer, or analyst confidence < 0.8, or commercial claims with no corroboration.
What "probable" rests on
- QA APMP-5 (16 questions): provider model answers + "relevant content" bullet lists mapped to Assessment Criteria. Marked probable (80) — credible provider, aligned with APM BoK7; explicitly "partial examples", not official keys.
- APM 2021 Sample Answer Document (10 questions): APM-authored strong/weak examples; marked probable (80). Explicitly "example answers, not model answers".
- Wellingtone (2): trainer-authored worked answers with published mark splits; probable (80).
- Parallel (3): delegate answers with tutor feedback; probable (75–80).
- Analyst-inferred keys on official sample papers (current format): high-confidence items (≥0.8) marked probable, e.g.:
- SP1 Q9 earned value = EV metric for "costs committed and work achieved" (0.9)
- SP1 Q17 team cohesion = open communication + collaboration (0.9)
- SP1 Q20 Thomas-Kilmann = avoidance / competition (0.9) — also corroborated by BoK7/QA Q12b and the current Handbook LO 11
- SP2 Q29 risk-process order (0.9) — corroborated by BoK7/QA Q1a
- SP1 Q26 RACI = OBS + WBS (0.9); SP2 Q17 change response = assess impact (0.95)
- SP2 Q37 iterative resource commitment = "available for the entirety" NOT needed (0.7→unsupported bucket boundary; kept unsupported)
Independently corroborated answers (strongest evidence)
| Question family | Corroboration |
|---|
| Thomas-Kilmann modes | SP1 Q20 (current) + BoK7/QA Q12b + LO 11 syllabus |
| Risk management process order | SP2 Q29 (current) + BoK7/QA Q1a + LO 23 |
| Resource smoothing vs levelling | BoK7/QA Q7b + SP1 Q5 + LO 21d |
| Benefits management 5 steps | SP1 Q27 + SP2 Q27(b) + LO 9a |
| Configuration management steps | SP1 Q15 + BoK7/QA Q11a + LO 24 |
| Projects vs BAU differences | BoK7/QA Q8a + Project Accelerator Q4 + LO 2 |
Long-response model marking points
answer_key.csv carries, for long-response questions with any answer treatment, a model marking-point summary (e.g. SP1 Q35 wind farm: 2 generic reasons + 3 scenario-specific reasons; SP1 Q7 business case: per-section contents). These are analyst/trainer-derived marking points, not APM examiner keys; each entry states its basis.
Explicitly NOT answered / flagged
- All select-from-list (SFL) items in official papers lack official keys; analyst picks are marked with confidence and ambiguity notes (SP1 Q4, Q12, Q20, Q28, Q37; SP2 Q3, Q11, Q19, Q27, Q35).
- Commercial dump answers (FreeCram/Marks4sure/ValidExamDumps/Pass4Success/DumpsTool) are recorded as unverified claims; none is used as corroboration.
Disagreements preserved
Where independent sources could disagree, no consensus was manufactured. The three disputed items above are the only places where the corpus found material ambiguity; if a downstream pass finds conflicting provider answers (e.g. a Scribd "model answers" document vs QA's), the disagreement should be added to 07_ANSWERS/disputed_answers.jsonl with both sides preserved.
Second pass update (2026-08-27)
- 77 probable / 3 disputed / 106 unsupported (was 62/3/83). The +15 probable come from newly extracted provider-supplied answers: Projex mark schemes (9), Parallel Starter-for-Ten delegate answers (2), PA tips (2), APM Annotated Q6 (2).
- Strongest corroboration added: the APM Annotated Sample Answer Q6 (SRC_0008) provides marker-ratified model answers for (a) EVM interpretation (SPI<1 → behind plan; CPI>1 → cost-efficient; EAC £96k vs £120k BAC; justified responses incl. re-visiting the business case) and (b) information management (collection/storage/dissemination). SP1 Q9 (earned-value metric) upgraded to 0.95 with this corroboration.
- Disputes re-checked against all new material — all three remain unresolved (SP1 Q10 PESTLE Political/Legal; SP2 Q3 iterative/evolutionary; SP2 Q27 benefits source). No official key exists; preserved as disputed per the brief.
- New content-quality red flag: CertEmpire (SRC_0060) current-format MCQs are explained with PMBOK references — evidence that some Tier-5 'PMQ practice' content is recycled PMP material. Answers from such sources are recorded as unverified and excluded from corroboration.
- Projex mark-scheme evidence (10 marks per complete explained point; 3 marks for a listed item + 7 for description; 5 marks per stated principle, 2 if merely listed) corroborates the 'one point ≈ one mark' answer-structure guidance used throughout the curriculum.