Evaluations Guide

Provenance (written 2026-09-29). Built from research/veridian/WBS13-verified-facts.md (the course's provenance-checked WBS13 facts bank — 9 located series, Oct 2021–Oct 2025, 7 fully mined plus two partial passes) and cross-read against every file already in content/veridian/technique-hub/wbs13/ before writing a word, specifically to avoid restating forge-level-threshold-guide.md, which already works through all eleven Level-boundary crossings on this paper's three levels-based tariffs (Discuss 8, Assess 12, Evaluate 20) with real anchors and illustrative pairs at each one. This document does not repeat that boundary-by-boundary analysis. Its two additive contributions are: (1) a short, direct answer to a narrower question the threshold guide doesn't ask — which command words on this paper carry AO4 marks at all, and how much of the paper's total mark allocation AO4 actually is — and (2) one single, continuous worked build of a real 20-mark Evaluate answer, developing the same argument through all four levels in one unbroken pass, rather than the threshold guide's paired before/after boundary snapshots. The worked question — Center Parcs' growth-strategy decision tree, October 2024 Q2 — is named in the threshold guide's own corroboration list but not built out there; it is also used for calculation-method teaching in decision-trees-cpa-and-contribution.ts and referenced for its real figures in codex-deep-dive-decision-trees-cpa-and-contribution.md, but neither document maps it against the AO4 level-descriptor language the way this one does. A check specific to this document type's own history: the brief for this file flagged that a sibling document (WBS12's forge-evaluations- guide.md) once contained an entirely fabricated "trigger phrase" mechanism for Level 4, since corrected. ~/Downloads/FINSTO/ was checked directly for a WBS13 equivalent of that raw draft to verify or debunk here — none exists; the folder holds evaluation-guide drafts for WBS12 and for the Economics units (WEC11–13) only. There is therefore no WBS13-specific fabricated claim on record to correct, and no verification ledger is included below for that reason — not because the check was skipped, but because it turned up nothing to report. content/veridian/technique-hub/wbs13/*.md was also grepped directly for "trigger phrase" and "AO4" before writing: zero and one hits respectively, confirming the AO4-specific angle below is genuinely unclaimed ground in this folder, not a restatement under a new title.

#WBS13 — Evaluations: What AO4 Credit Actually Looks Like on This Paper

#1. Where AO4 lives on this paper, by command word

AO4 (evaluation) is not evenly spread across WBS13's command words. Appendix 6 (spec p.56) defines each command word's own requirement, and only three of the paper's seven command words ask for anything evaluative at all:

Two things follow directly from that table, both worth stating plainly rather than left implicit. First, on Define/Calculate/Construct/Explain — half this paper's command-word vocabulary — writing an evaluative "however" earns nothing; Appendix 6's own wording for each of these four asks for recall, a working, or a stated cause, never a weighed argument. Second, Analyse is the one case where this is not just a structural absence but a named exclusion: the spec's own wording rules evaluation out of a 6-mark Analyse answer specifically, which means an Analyse question and a Discuss question sitting on the same extract can look superficially similar in what they ask a candidate to write about, while one of them structurally cannot reward the very move the other requires.

At the whole-unit level, AO4 is worth 6.9% of Unit 3's overall assessment weighting (spec p.39) — smaller than AO3's 7.5% but larger than AO1's 5%, meaning this paper rewards demonstrating a weighed judgement slightly more than it rewards recalling a fact, even before the 20-mark Evaluate essay's own, much heavier concentration of that weighting is accounted for. This figure describes the whole paper's assessment design, not any single question's mark allocation.

#2. One real question, built through all four levels in a single pass

Everything above is structural. What follows is one real, mark-scheme-sourced Evaluate question, built as a single developing answer — the same underlying argument, extended level by level — rather than the isolated before/after pairs used elsewhere in this course's material, so the cumulative shape of AO4 credit is visible in one place.

The question, real, October 2024 Q2 (Publications Code WBS13_01_2410_MS), 20 marks: using decision-tree-analysis calculations and Extracts E and F, evaluate how far DTA will help Center Parcs decide its growth strategy. Extract F gives two options: build a new holiday village (cost €520m; 0.8 probability of €780m success, 0.2 probability of a €120m loss) against taking over an existing one (cost €100m; 0.7 probability of €220m success, 0.3 probability of a €70m loss). The mark scheme's own worked figures — independently recomputed here and confirmed correct — give the new village an expected monetary value of €80m against the takeover's €33m, and recommend the new village on that basis. A genuine, flagged limit on this example before using it: only the mark scheme and question paper for this series have been read for this course; no examiner's report for October 2024 currently exists in this course's sources, unlike every other worked question this technique-hub draws on. That means Levels 1–3 below cannot be anchored to a real graded script or ER commentary the way the threshold guide's other examples are — they are constructed illustratively, in that document's own labelling convention, to show the shape of the climb. Only the Level 4 close below is genuinely sourced throughout, from the mark scheme's own indicative content.

The ladder (20-mark tariff, paraphrased from the Jan 2025 mark scheme and cross-corroborated elsewhere in this course's sources; see forge-level-threshold-guide.md §4 for the full descriptor text and its own citation trail):

Level 1. "Center Parcs could build a new holiday village or take over an existing one." A choice is named from the extract; nothing is calculated, connected, or weighed. This clears the floor into Level 1 by naming a relevant, extract-based fact — nothing more.

Level 2. "Building a new holiday village costs €520m but could bring in €780m if it succeeds, which is much more than the €220m the takeover option could bring in. This suggests building a new village is the better choice." A real figure is now used, and a preference is stated — but it compares raw best-case payoffs, not the decision-tree-weighted outcome the question specifically asks for, and never mentions either option's failure branch at all. This is a judgement "attempted," in the ladder's own phrase, but not one built from the technique the question names.

Level 3. "Weighting each outcome by its stated probability, the new village's expected monetary value is €80m (0.8 × €780m, less the 0.2 × €120m loss, less the €520m cost), against the takeover's €33m using the same method. On this basis the new village is the stronger option, since €80m is the higher net gain once both branches of each decision are properly accounted for." The DTA calculation is now actually doing the evaluative work — connecting cause (probability-weighted outcome) to consequence (which option nets more) — and a conclusion is reached. What is still missing, on the ladder's own wording, is a full awareness of the significance of the competing arguments: the €80m-vs-€33m comparison is complete, but nothing yet engages with what makes the new village the riskier of the two options, only with which number is larger.

Level 4. The same figures, extended: "However, the new village's higher EMV rests on a 20% chance of a €120m loss against a smaller €520m-to-€780m spread, while the takeover's outcomes are less extreme in both directions — so the choice is not simply between a bigger and a smaller number, but between a higher-EMV option that is also the more volatile one. Given Center Parcs' scale, the firm can plausibly bear a one-in-five chance of a €120m loss without existential risk, which is precisely why EMV, not the best-case figure alone, is the right basis for this specific decision — but a firm less able to absorb that downside would reasonably weigh the takeover's smaller risk profile more heavily even at a lower expected return." This is genuinely sourced: the mark scheme itself credits exactly this caution, stating in substance that the new village's largest possible payoff "may not" be the basis for the decision alone, given the accompanying failure probability. The arithmetic is identical to the Level 3 version above; what closes the distance is naming why the higher-EMV option is also a live judgement call, not just a bigger number, which is the entire, specific difference between "may lead to a conclusion" and "an effective conclusion" on this ladder.

The distinct trap this example isolates, worth stating on its own. A candidate can get every figure in this question numerically correct — both EMVs computed and compared without error — and still sit at Level 3, not Level 4, if the write-up stops at "€80m beats €33m, so build the new village." That is a calculation success and an evaluation shortfall at the same time: the arithmetic in decision-trees-cpa-and-contribution.ts already warns against a different, purely mechanical error (comparing raw expected value before subtracting cost); this is the AO4-specific failure that survives even once the arithmetic is entirely correct — stopping at the comparison instead of weighing what the comparison is actually a trade-off between.

#3. What this document cannot tell you

Named directly, as this course's other technique-hub files do: this is the only WBS13 decision-tree question with a named real company found across nine series checked for this sub-topic, so nothing above should be read as describing a recurring pattern for decision-tree questions generally — it is one confirmed instance, not a trend. No examiner's report has been read for the October 2024 series, so Levels 1–3 above are this document's own construction against the general descriptor wording, not quotations of a real script or real ER commentary — flagged plainly rather than left to look otherwise. This paper's sources give no raw-mark-to-grade conversion at any series (see signal-frequency-and-boundaries.md §3); nothing above states or implies one. And for the full, independently-anchored account of every Level boundary on all three of this paper's levels-based tariffs — with real graded exemplars where they exist — forge-level-threshold-guide.md remains the fuller document; this one is deliberately narrower.