Level Threshold Guide

Provenance (checked 2026-09-28): built on research/veridian/WBS12-verified-facts.md, content/veridian/curriculum/wbs12/paper-anatomy.ts and content/veridian/technique-hub/wbs12/forge-exam-technique-guide.md — all three read in full first, so this document assumes and does not re-derive their tariff structure, AO splits, command-word technique or the level-descriptor tables themselves, which are already reproduced in full in the facts bank and in paper-anatomy.ts. content/veridian/question-bank/wbs12.ts was also read in full; its two real, fully-worked level-exemplar ladders (Sew Sew, Jan 2021 Q2(d); Żabka, Jun 2023 Q1(d)) are referenced by pointer below, not reproduced. New primary-source work for this specific pass: twelve WBS12 examiner reports were opened directly from the local archive and searched specifically for numerically-marked worked exemplars on the 8-mark Discuss, 10-mark Assess and 20-mark Evaluate tariffs — a question the facts bank's own pass answered at the level-descriptor and common-error layer but did not extract exemplar-by-exemplar. Reports checked: January 2020, January 2021, January 2022, June 2022, June 2023, October 2019, October 2020, October 2021, October 2022, January 2024 (from the pre-extracted local text archive), plus January 2025 and June 2025 (re-extracted this pass with pdftotext -layout from wbs12-01-pef-20250306.pdf and wbs12-01-pef-20250814.pdf, the same primary PDFs forge-exam-technique-guide.md used). October 2022, October 2023, October 2024 and January 2025's Section C entries were also checked and confirmed to carry no numerically-marked exemplar box in the format the other eight reports use — a real, checked absence, not an unexamined gap. Every direct quotation below is kept under roughly 15 words with series and question number cited; nothing longer is reproduced, and the full text of any candidate response is never quoted, only the examiner's own explanatory sentence, paraphrased where it runs long.

#WBS12 Level/Threshold Guide — What Actually Separates One Mark Band From the Next

forge-exam-technique-guide.md already establishes how levels-based marking works in general — best-fit, not a checklist — and what each command word demands. This document goes one layer deeper and stays there: for the paper's three levels-based tariffs (Discuss, 8 marks; Assess, 10 marks; Evaluate, 20 marks), it works through every real level boundary in turn, using the verified descriptor wording plus real, mark-numbered examiner-report exemplars to show what specifically separates a response that just misses a level from one that just clears it — and, where the evidence supports it, what separates the bottom of a level from its top. Define, Calculate/Explain/Construct and Analyse are AO-marked, not levels-based, and so have no mark-band boundary of this kind to describe; they are out of scope here by design, not oversight (forge-exam-technique-guide.md covers their technique in full).

This is not a restatement of the level-descriptor tables — check those directly in the facts bank or paper-anatomy.ts. What follows only exists because of new work this pass: twelve examiner reports searched specifically for numerically-marked exemplars, twenty-three found in total across the three tariffs, and one genuinely new, cross-series-confirmed finding about how a mark is actually placed inside a level once it's chosen, which is the natural place to start.

---

#1. How a mark actually gets placed inside a level

forge-exam-technique-guide.md already states the general mechanism: examiners find the best-fit level for a whole response, then place the mark toward that level's bottom or top depending on how strong the match is. Three examiner reports — June 2022, June 2023 and June 2025, all describing Section C's 20-mark Evaluate — narrate that mechanism with more precision than has previously been written up anywhere in this course's own WBS12 material: having decided a response is a level-4 match, examiners "started at the mid-point within the level of 18 marks" (Jun 2023 and Jun 2025 phrase it identically; Jun 2022 phrases the same idea as starting "at the mid-point of 18 marks"), then move the mark up or down from that anchor for a stated reason. The three real outcomes:

• Jun 2022 (bank loans/peer-to-peer funding): moved up from 18 to 19/20 — no reason to move down, and it moved up because the response was judged "fairly strong in each element."
• Jun 2023 (retained profit/sale of assets): stayed at the anchor, 18/20 — no reason to move down, but it didn't move up either, because the response "did stray from the question to some extent."
• Jun 2025 (inflation vs interest rates): moved down from 18 to 17/20 — it didn't move up because it lacked "a full awareness of the validity and significance" of competing arguments, and it moved down because effective use of business context "was not throughout the response."

18 is the arithmetic midpoint of Level 4's own 15–20 range, so this isn't a special rule for Section C — it's the general best-fit mechanism made explicit, for once, in the reports' own words. The reports reviewed for this pass don't narrate the identical arithmetic for the 8-mark Discuss or 10-mark Assess tariffs in these same words, but the real mark distribution for those two tariffs (§2 and §3 below) is consistent with the same logic: Level 3's 6–8 band has a midpoint of 7, and of eleven real Discuss exemplars found across the reports checked, every near-miss lands at 7 and every full match lands at 8, none at 6 or below within that top band. Level 4 Assess's 8–10 band has a midpoint of 9, and of nine numerically-marked Assess exemplars found, the same pattern holds — treat that as a genuine, evidence-consistent inference this pass drew from real data, not a second Pearson quotation to the same effect.

One honest pattern worth naming before the tariff-by-tariff sections below: every single numerically-marked worked exemplar found across all twelve reports checked for this pass — twenty-three in total across Discuss, Assess and Evaluate — sits in that tariff's own top level. Not one report reviewed included a marked exemplar sitting at Level 1 or Level 2 for any of the three levels-based tariffs, and for Assess and Evaluate specifically, none sat at Level 3 either. Examiner reports consistently choose a strong, often near-full-marks response to model in detail, and describe weaker responses only in prose, without a matching worked exemplar attached. That's a real, useful thing to know about the source material itself: the lower-boundary illustrations below lean on real descriptor language and real (but unmarked) examiner commentary about what weak responses do, not on a matching marked exemplar Pearson never published — and each place that's true is flagged as it comes up, rather than papered over with an invented mark.

---

#2. Discuss — 8 marks, 3 levels (1–2 / 3–5 / 6–8)

#2.1 Level 1 → Level 2

The descriptor's own language draws this line sharply: Level 1 is "recall based" with "weak or no relevant application," while Level 2 requires accuracy "applied accurately to the business and its context" and at least an attempted chain of reasoning, even if that chain "may be assertions or incomplete." No report checked for this pass carried a marked Level 1 or bottom-Level-2 Discuss exemplar, but the prose commentary is real and repeats the same failure mode across series: a response that engages only one side of the topic, or stays generic enough to fit almost any business, cannot climb past this boundary regardless of accuracy. January 2022's Etsy business-failure question records exactly this: some candidates, instead of weighing internal causes against the evidence on both sides, ended up "discussing only external causes of business failure" — a real, one-sided answer that the report treats as a genuine ceiling on its own, separate from any accuracy problem.

Illustrative pair (VERIDIAN-original, not a reproduction of any real question or candidate response — Calder Outdoors, a small outdoor-equipment retailer):

• Just misses (Level 1): "Legislation could affect Calder Outdoors because all businesses have to follow the law." — accurate as a generic statement, but names no specific law, no specific fact about Calder Outdoors, and nothing a different business's answer wouldn't also say.
• Just clears (Level 2): "Because Calder Outdoors sells climbing rope, it must meet consumer protection standards on product safety — if a rope failed and injured a customer, the business could be sued and lose its safety reputation." One real chain of reasoning, genuinely tied to this business, even though only one side of the argument is present yet.

#2.2 Level 2 → Level 3

This is the boundary with the richest real evidence on file, because two full, real-question level-exemplar ladders already exist for it in content/veridian/question-bank/wbs12.ts — Sew Sew (Jan 2021 Q2(d), health and safety and consumer protection legislation) and Żabka (Jun 2023 Q1(d), the taxation cut) — and are not reproduced again here; read those for a complete worked climb through all three levels on real questions. What's added here is the sharper boundary evidence a third real series supplies. October 2020's Q2(d) (personal savings as a source of finance) scored 6/8 — the very bottom of Level 3 — and the report's own reason is precise: the answer "has a two-sided argument which uses the extract but could benefit from developing the chains of reasoning further." That's the whole boundary in one sentence: a second, opposing point is present (which is what gets the response into Level 3's "awareness of competing arguments" at all, where Level 2's own descriptor explicitly says that awareness is "unlikely" to appear) — but the second side is thin, so the mark lands at the level's own floor rather than higher inside it.

Illustrative pair (VERIDIAN-original, Calder Outdoors):

• Just misses (top of Level 2): "...Calder Outdoors must meet consumer protection standards on product safety, or it risks being sued. However, following extra safety standards costs money." — a second point exists, but it's asserted, not developed: no reason is given for why the cost matters here, or how much.
• Just clears (bottom of Level 3): "...However, meeting stricter safety standards on climbing gear means buying more expensive certified rope, which raises Calder Outdoors' costs at a time it is already a small, low-margin retailer — a real trade-off against the reputational benefit." The same second point, now carried one step further with a specific, applied consequence.

#2.3 Inside Level 3: what separates 7 from 8

Because Discuss has no Level 4, this ceiling matters as much as any other boundary — it's the difference between a strong answer and full marks. Eleven numerically-marked real exemplars were found across the reports checked: one at 6/8, three at 7/8, seven at 8/8, and in every case the stated reason for stopping short of 8 is specific, not generic:

The pattern across all three 7/8 cases is a response that has genuinely reached two-sidedness — the precondition for Level 3 at all — but has one specific, nameable shortfall on top of that: an imbalanced counter-argument, a drift away from the exact question asked, or a final point left as an assertion rather than fully reasoned through. None of the 8/8 cases are described as flawless; they're described as complete against the descriptor's own three components (accurate and applied knowledge, a logical chain of reasoning, and a genuinely balanced awareness of competing arguments) with no single component left short.

---

#3. Assess — 10 marks, 4 levels (1–2 / 3–4 / 5–7 / 8–10)

#3.1 Level 1 → Level 2 and 3.2 Level 2 → Level 3

As with Discuss, no report checked for this pass carried a marked exemplar below Level 4 for Assess — every one of the nine numerically-marked exemplars found sits in the top band (§3.3 below). The lower boundaries have to be drawn from descriptor language plus real, unmarked prose evidence, which is genuinely there and specific. January 2024's Q2(e) (interest rates and Kajak Kanu Klub) gives an unusually direct, real contrast between the two sides of the Level 1/2 line and the Level 2/3 line in the same paragraph: a bare, unsupported assertion — that the business "had a loan and therefore must pay back a lot of interest" — was "less likely to progress through the levels," while a more developed version considered whether the business actually held a loan, its size, and that "the rate of interest had only increased a small amount." The first is Level 1/2 territory: a plausible-sounding but unexamined claim. The second reaches toward Level 3: the same underlying idea, now conditioned on real facts rather than asserted outright — which is exactly the shift the Level 3 descriptor's own "analytical perspectives" language calls for, over Level 2's "elements of knowledge and understanding... applied" without much development.

October 2019's Q1(e) (liquidity, acid test and current ratio) names the same move for a full Level 4 response, which by extension marks where Level 3 stops short: examiners wanted "developed arguments stating both why there should be concern... but also why the finance manager may not have needed to worry" — i.e. two developed, opposing analytical strands, not one strand stated twice or one strand left undeveloped.

Illustrative pair (VERIDIAN-original, Calder Outdoors — Assess the likely impact on Calder Outdoors of a rise in interest rates):

• Just misses (Level 2): "Higher interest rates mean loan repayments cost more, so Calder Outdoors would have less profit." One accurate, generic mechanism, asserted rather than developed against anything specific to the business.
• Just clears (bottom of Level 3): "Calder Outdoors took out a loan to open its second store, so a rate rise directly increases its repayments — but because most of its stock is imported and paid for in the supplier's currency, not borrowed locally, the scale of the hit depends on how much of that loan is actually still outstanding." A genuine analytical perspective, tied to a specific, applied feature of this business, though not yet weighed into an overall assessment.

#3.3 Level 3 → Level 4 — the supported-judgement line

The descriptor's own language marks this boundary more sharply than any other in the paper: Level 3's own top line is an assessment that's merely "unlikely to show the significance of competing arguments," while Level 4 requires that same balanced awareness "leading to a supported judgement." No report checked for this pass included a marked exemplar sitting inside Level 3 for Assess to show the crossing directly, but the nine real Level 4 exemplars found are unusually informative about where that line sits, because several of them describe a response that reaches Level 4 with real deficiencies in exactly the feature that defines it:

• Jan 2022, DFS environmental legislation (Q2e), 8/10: reaches Level 4 but "fails to reach the top of the level" specifically for "the lack of a supported judgement."
• Oct 2020, Soul Boul business failure (Q2e), 9/10: "only missing a final judgement" — one mark short of full, for the same missing component.
• Jan 2021, Sew Sew financial forecasts (Q2e), 10/10 and Oct 2021, share capital (Q1e), 10/10: both full-marks responses are singled out for successfully "linking the points" into an assessment, or being "supported by a valid conclusion."

Read together, this says something real and worth stating plainly: a response can be assessed as a Level 4 match — balanced, developed, well-applied — while still being marked down within Level 4 for an underdeveloped or missing supported judgement, rather than being pushed back down into Level 3 altogether. The boundary itself is crossed by attempting a stated, reasoned preference at all, even an imperfect one; how fully that judgement is developed and supported is then what separates the bottom of Level 4 from its top (§3.4).

Illustrative pair (VERIDIAN-original, Calder Outdoors):

• Just misses (top of Level 3): "...So a rate rise creates a real cost pressure on Calder Outdoors' loan repayments, though the exchange-rate exposure on its imported stock is arguably the bigger risk. Both factors matter for the business's costs." Two developed strands, genuinely weighed — but the sentence stops at naming that both matter, without saying which one the business should actually worry about more, or why.
• Just clears (bottom of Level 4): "...On balance, the loan-repayment pressure is likely to matter more in the short term, since it is a direct, certain cost, while the exchange-rate exposure depends on currency movements Calder Outdoors cannot control or predict." The same two strands, now resolved into an actual stated preference with a reason — even though, per the pattern above, a fuller, more evidenced version of exactly this sentence is what would move the mark from the bottom of Level 4 toward the top.

#3.4 Inside Level 4: 8 vs 9 vs 10

Nine real, numerically-marked Assess exemplars were found across the reports checked — five at 8/10, two at 9/10, two at 10/10 — and the stated reasons cluster tightly around three components, each independently capable of holding a response below full marks even once Level 4 itself is reached:

• Analysis/reasoning not fully developed: Jan 2021 Greggs (8/10) and Oct 2020 Lotus Garments (8/10) use the identical phrase independently — "not as strong as would be preferred from a full mark response" — and Jan 2022 Etsy (8/10) and Oct 2021 cash-flow forecasts (8/10) each cite chains of reasoning that are "not fully developed" or of "limited development."
• The judgement itself under-supported: Jan 2022 DFS (8/10, "lack of a supported judgement") and Oct 2020 Soul Boul (9/10, "only missing a final judgement") — the closer a response gets to full marks, the more specifically this narrows to just the closing judgement rather than the whole response.
• Balance: Jun 2022 R B Rustico (9/10) — "the balance could be better, preventing the full ten marks."

The two 10/10 cases (Sew Sew, Jan 2021; share capital, Oct 2021) are each described as resolving all three at once: developed reasoning, extract genuinely used throughout, and a real, evidenced judgement or conclusion — not a flawless answer by some higher, unstated standard, but one with no specific gap left for the report to name.

---

#4. Evaluate — 20 marks, 4 levels (1–4 / 5–8 / 9–14 / 15–20)

#4.1 Level 1 → Level 2 and 4.2 Level 2 → Level 3

The clearest real evidence for Evaluate's lower boundaries is a specific, cross-series-confirmed failure mode: a response that lists advantages and disadvantages of the right topic but never evaluates them against the actual question asked. October 2019's Q3 (loans vs leasing) names this directly — weak responses gave "a generic list of the advantages and disadvantages of loans and leasing" rather than weighing them against each other. January 2024's Q3 (Freewing, short product lead-in times and quality management) and January 2025's Q3 (Coles, the business cycle) each independently confirm the identical pattern on unrelated topics, two further series apart: both reports use the near-identical opening "failed to evaluate these," then name what was skipped — Freewing's "competitive advantage" in the first, Coles's "extent to which it is likely to affect" the business in the second. Seen in three of the reports checked, on three unrelated topics, this is a genuinely paper-wide failure mode, not a one-off: knowing the content, and even applying it, without the response ever turning back to answer the specific question, is what caps an Evaluate response around Level 1/2 regardless of how accurate the content itself is — matching the Level 2 descriptor's own language that a "comparison or judgement may be attempted" but "will not successfully show an awareness of the key features" of the actual business situation asked about.

Illustrative pair (VERIDIAN-original, Calder Outdoors — Evaluate the extent to which a rise in interest rates is the most significant external influence on Calder Outdoors' profitability):

• Just misses (Level 2): "A rise in interest rates increases loan repayments, which is bad for a business's profit. However, exchange rates could also affect a business that imports stock, and inflation can raise costs too." Accurate content about three real external influences — but never actually compares them against each other for this business, or returns to the question's own "most significant" framing.
• Just clears (bottom of Level 3): "Because Calder Outdoors' loan is fixed-rate for its remaining term, a rate rise has no immediate effect on its existing repayments — whereas its imported stock is exposed to exchange-rate movements on every single order, making that channel the more live risk for this specific business right now." The same underlying knowledge, now turned directly onto the comparison the question actually asks for.

#4.3 Level 3 → Level 4 — the required-conclusion line

The descriptor text draws a genuinely sharp, previously under-stated distinction here: Level 3's own top line says quantitative and qualitative information is used to support judgements and "may lead to a conclusion" — a possibility, not a requirement — while Level 4 requires "an effective conclusion that proposes a solution and/or recommendations." That single word, "may," is doing real work: a Level 3 response can be complete, developed and well-evidenced throughout and still, by the descriptor's own wording, never actually conclude anything. No report checked for this pass included a marked exemplar sitting inside Level 3 to show a response stopped at exactly this line, but every one of the seven real Level 4 Evaluate exemplars found is explicit that the presence of some conclusion — however underdeveloped — is what the response is being credited for reaching Level 4 at all, with its quality then deciding where inside the level the mark lands (§4.4).

Illustrative pair (VERIDIAN-original, Calder Outdoors):

• Just misses (top of Level 3): "...So the interest-rate channel is limited by the fixed-rate loan, while the exchange-rate channel is a live, ongoing exposure. Both external influences are relevant to Calder Outdoors' profitability in different ways." Two fully developed, evidenced strands, genuinely weighed against each other — but the response ends there, without a stated recommendation.
• Just clears (bottom of Level 4): "...On balance, exchange-rate movements are the more significant influence on Calder Outdoors' profitability at present, since the interest-rate channel is temporarily locked in by the fixed-rate loan; the business should prioritise hedging or renegotiating its supplier contracts in a stronger currency before its next major stock order." The same two developed strands, now closed with an actual stated recommendation.

#4.4 Inside Level 4: the midpoint mechanism, worked three times

§1 above already gives the clearest real evidence in the whole archive for this specific boundary — the three June-series Section C exemplars, each starting from the same 18-mark midpoint of the 15–20 band and moving in a different, specifically-explained direction. The other four real Level 4 exemplars found (Jan 2021, 16/20; Oct 2020, 16/20, identical wording to Jan 2021 — "doesn't flow very well and is not nuanced"; Jan 2022, 17/20; Oct 2021, 17/20, identical wording to Jan 2022 — "could make better use of quantitative and qualitative information") confirm the same pattern from the other direction: two independent series each reusing the same reasoning, once for a low-Level-4 mark and once for a mid-Level-4 mark, which suggests these aren't one-off comments on one candidate's individual answer so much as standard descriptions of what a 16/20 or a 17/20 Evaluate response generically looks like on this paper. Taken together, the seven real cases span the entire top band: 16 twice ("doesn't flow... not nuanced" — present but weak use of the required elements), 17 three times (present, contextualised, but under-using quantitative/qualitative evidence or, in the June 2025 case, not sustaining business context throughout), 18 once (a fully competent response that doesn't stray but doesn't especially excel either — the true midpoint, unmoved), and 19 once (strong in every element, no flaw specific enough for the report to name one). None of the seven reach 20/20 in the reports checked — a real, honest limit of the evidence available, not a claim that 20/20 is unreachable.

---

#5. What this document deliberately does not cover, and why

• The level-descriptor tables themselves. Already verified and reproduced in full, with citations, in research/veridian/WBS12-verified-facts.md and content/veridian/curriculum/wbs12/paper-anatomy.ts — repeating them here in full would be exactly the duplication this document's own brief rules out. Only short, attributed fragments are quoted above, at the specific point each one is doing real work.
• Full worked level-exemplar ladders for real past questions. Two already exist in content/veridian/question-bank/wbs12.ts (Sew Sew, Żabka) and are pointed to, not reproduced, in §2.2. This document's own illustrative pairs are deliberately short — one or two sentences per level, not a full essay — because their job is to isolate the discriminating feature at each boundary, not to model a complete answer a second time.
• A marked, worked low-level (Level 1 or 2) exemplar for Assess or Evaluate. Flagged honestly at the point it matters (§1, §3.1–3.2, §4.1–4.2): no report checked for this pass published one. This document uses real descriptor language and real, unmarked examiner-report prose to characterise those boundaries instead of inventing a marked exemplar Pearson never produced.
• Grade boundaries (A–E, out of 80). A completely different kind of boundary — the whole paper's raw-mark cut-offs, not a single question's internal mark bands — already compiled with real, sourced figures in content/veridian/technique-hub/wbs12/signal-frequency-and-boundaries.md. Not duplicated here.

---

#Sources

research/veridian/WBS12-verified-facts.md (verified level-descriptor tables and tariff structure); content/veridian/curriculum/wbs12/paper-anatomy.ts and content/veridian/technique-hub/wbs12/forge-exam-technique-guide.md (levels mechanism, tariff technique — both read in full and cross-referenced rather than repeated); content/veridian/question-bank/wbs12.ts (two real level-exemplar ladders, referenced by pointer). Primary-source examiner reports opened and searched directly for this pass (all from the local archive also cited throughout this course's WBS12 material): January 2020, January 2021, January 2022, June 2022, June 2023, October 2019, October 2020, October 2021, October 2022, January 2024, January 2025 and June 2025 — twelve reports, twenty-three real numerically-marked worked exemplars found and cited above by series and question number. Not affiliated with or endorsed by Pearson Edexcel; every quotation above was checked against the real examiner report it is attributed to during this pass, kept under roughly 15 words per quotation, and the "Calder Outdoors" illustrations throughout are explicitly VERIDIAN-original teaching material, not reproductions of any real Pearson question or candidate response.