Question Recurrence Report

How questions repeat across the publicly observable PMQ question universe.

Method: exact (normalised-text equality) → near (≥0.97 similarity) → structural (≥0.88 + shared syllabus) → conceptual (≥0.70 + shared syllabus + shared family + keyword overlap). Raw frequency and independent-source frequency are reported separately (per spec §24).

Headline numbers

148 raw occurrences → 129 unique questions → 18 duplicate/variant groups with >1 independent source.
14 exact/near groups, 2 structural, 2 conceptual.
Every group retains all source references (see 06_NORMALISED_CORPUS/duplicate_groups.json).

Highest-signal recurrence groups

Group	Type	Sources	Underlying question
DG_0033	exact	SRC_0001, SRC_0002 (official SP1+SP2)	"…HR platform resource planning — which consideration is not critical?"
DG_0035	near	SRC_0001, SRC_0010 (official + Wellingtone)	Wind-farm stakeholder-expectations long question
DG_0080–0095	exact/near/structural	SRC_0004 + SRC_0005 (BoK7 official + QA)	All 16 old-format questions (QA answered the official paper)
DG_0081	conceptual	SRC_0004, SRC_0005, SRC_0006 (+3 indep)	Functional vs matrix organisational structure
DG_0107	structural	SRC_0011, SRC_0015 (Parallel + PMW)	Advantages of effective configuration management

Commercial-bank copying chains (Tier 5, unverified)

Verbatim-identical questions across storefronts indicate a single circulating practice bank, not independent evidence:

FreeCram Q2 ↔ Pass4Success Q1 (finance software / T&M contract)
Marks4sure Q2 ↔ DumpsTool Q6 (organisational practices/norms analysis technique)
FreeCram Q1 ↔ ValidExamDumps Q4 (eco-office sustainability life-cycle point)
DumpsPedia Q2 ↔ ValidExamDumps Q2 (banking product / iterative life cycle)
DumpsTool Q12 ↔ Pass4Success Q3 (defence / linear life cycle)

⇒ For these, independent_source_count = 1, regardless of how many storefronts host them.

Recurring old-format "banker" topics (candidate/trainer-consensus)

Quoted or implied by multiple independent old-format sources (PMW, Parallel, QA, BoK7):

project management plan components; configuration management advantages; risk management process steps; projects vs BAU; functional vs matrix; sponsor responsibilities; team development stages; EVM advantages/disadvantages; business case content; procurement strategy contents.

Current-format within-exam evidence

Only one exact within-exam repeat is observable (SP1 Q33 = SP2 Q14) — both official papers are independent constructions, so this is the strongest public evidence that question reuse (not just topic reuse) occurs between sittings/versions. All other current-format recurrence is conceptual (e.g. re-estimation: SP1 Q18 ↔ SP2 Q12; sustainability reporting: SP1 Q4 ↔ SP2 Q6).

Cautions

Recurrence in this corpus measures public availability, not APM exam probability. Use independent_source_count for any probability modelling, and never treat a Tier-5 question's multiple storefronts as multiple occurrences.
Old-format recurrence (BoK7/QA/2021-answers) documents the historical exam, not the current one.

Second pass update (2026-08-27)

20 multi-source duplicate groups (was 18). All manually re-checked; no mis-groupings.
New verified groups: DG_0120 (2017 Q13 ↔ Projex Q1 — programme benefits, structural, 2 indep.); DG_0136 (Parallel 2023 Q1 ↔ PA tips Q1 — governance principles, exact, 2 indep.); DG_0143 (Parallel 2023 Q2 ↔ PMW Q2 — configuration management, structural, 2 indep.).
Recurrence evidence strengthened: the governance-principles question family now spans 3 independent old-format sources (Parallel, PA tips, PA 2023) plus Projex Q9 (six principles variant) — the strongest old-format 'banker' family in the corpus.
The programme-vs-projects benefits question family now spans 4 independent sources (BoK7 Q15a, QA Q15a, 2017 Q13, Projex Q1).
Candidate topic-level recall (SRC_0063/64) independently names procurement/BATNA-ZOPA, estimating methods, Belbin/Maslow, NPV and IRR as appearing in old-format sittings — consistent with the corpus's old-format frequency rankings (procurement 7 indep. sources; schedule/estimating 7).