Business Paper 3 — Business Decisions and Strategy

A Star Gap Analysis

#WBS13 — SIGNAL: A* Gap Analysis (Unit 3: Business Decision-Making)

Provenance (2026-09-29). This is a synthesis/index document, not new research — no primary PDF was opened for this pass. It is built from six documents already in this technique-hub, each read in full first: forge-level-threshold-guide.md (the verbatim level-descriptor language and every real threshold this paper's mark schemes draw, with graded and illustrative anchors), forge-error- pattern-tracker.md (the paper-wide technique/command-word and topic-confusion codes), forge-exam- technique-guide.md §4 (the danger-phrases index) and §3 (its own self-mark checklist, checked specifically so this document does not restate it), signal-frequency-and-boundaries.md §3 (checked for grade-boundary data — see §0 below for why none appears here), forge-judgements-guide.md (two real, freshly-verified 12-mark Assess questions — Spotify/Audible and Peloton/Precor — used in §2.4 and §5 below) and forge-evaluations-guide.md (the Center Parcs decision-tree question, used in §2.5 below for a distinct AO4-specific ceiling). signal-examiner-intelligence.md was also read in full and checked against every claim below for overlap; where it already states a finding, this document cites it rather than re-deriving it. Every claim below carries the same citation it carries in its source document; nothing here is re-verified against a fresh primary source, and nothing here invents a diagnostic category, a percentage, or a boundary the source documents don't already state. Short quotes (under ~15 words) are reused with their original attribution; everything else is paraphrase. Not affiliated with or endorsed by Pearson Edexcel.

What this is. A self-diagnostic tool: yes/no questions to run against your own last practice answer, each one built from a real, already-cited finding about what specifically kept a real graded or mark-scheme-anchored response short of the top band, or let it reach it. Where forge-error- pattern-tracker.md gives you a log to record that a mistake happened, this document exists to answer the question that log doesn't: why does this specific category cap the mark, and what real evidence says so.

What this isn't. New research, a new taxonomy, or a grade-boundary document. Every diagnostic question below maps to a category, trap or boundary that already exists in one of the six source documents named above — see the citation after each one.

#0. What "A*" means in this document, and what it deliberately doesn't

signal-frequency-and-boundaries.md §3 checked this course's full WBS13 sources specifically for a grade-boundaries document and found none: "No grade-boundary data exists anywhere in this course's WBS13 sources… Pearson's own published grade-boundary documents were never part of the material mined for this paper." That finding stands, and this document does not repeat the search or invent a figure to fill the gap it leaves.

What this document means by "A*," instead, is the thing the mark scheme itself actually measures: the ceiling of each level-of-response ladder — Level 3 (6–8) on the 8-mark Discuss, Level 4 (9–12) on the 12-mark Assess, Level 4 (15–20) on the 20-mark Evaluate — plus, on the discrete-mark questions (Define, Calculate, Explain), clearing the specific real threshold that otherwise caps an accurate answer one mark short of full. "A good pass" here means accurate, on-topic, Level 2/mid-Level 3 work with no factual errors; "the gap" is the specific, named thing that answer is still missing, not a vague call for "more detail." Nothing below states or implies what raw-mark total across a whole paper this corresponds to — that conversion doesn't exist in anything reviewed for this paper, and this document doesn't manufacture one.

#1. The one mechanism behind almost every gap on this paper

forge-level-threshold-guide.md §5 states three findings that recur often enough, across independent series and independent topics, to be the load-bearing pattern behind nearly every threshold this document's own §2 works through:

One-sidedness caps a response below the top band on every levels-based tariff, not only Discuss. Discuss's Level 3 needs a second, developed side with no conclusion; Assess's and Evaluate's Level 4 need a second side and a judgement that weighs it. In every real anchor cited in that document, the single side offered is accurate and well-developed — the ceiling is about coverage of the argument, not the quality of any one part of it.
Copying or narrating the source material, rather than using it as evidence, caps a response low regardless of tariff. Confirmed independently on a 12-mark question (Oct 2021, British Airways: "many candidates simply copied large chunks of the information from the Extracts") and a 20-mark one (Oct 2023, Disney: "just copied out the extract and described events"). This is distinct from one-sidedness — a narrated extract can technically cover both sides of an issue and still stay low, because nothing has been done with the material.
Breadth of coverage is not the same lever as depth of argument. Summer 2024's Bosch ER states this explicitly: "it was not necessary to cover all four aspects of SWOT to achieve level 4." A response naming every mechanism a topic offers does not automatically outscore one that develops fewer points further.

This is the single question worth asking of any answer before running the tariff-specific checklists in §2: is there a second, developed thing here — an opposing case, a weighed condition — or is this one side, or one narrated extract, however accurate? Every checklist item below is a more specific version of that same question, pinned to a real anchor.

#2. Run this on your own answer: a tariff-by-tariff self-diagnostic

#2.1 Calculate (4 marks) — point-marked, not levels-based; the discrete-mark equivalent of the same gap

forge-level-threshold-guide.md §1 excludes Calculate from its own boundary analysis because it is point-marked, not levels-based — there is no "level" to climb. But forge-error-pattern-tracker.md Part 1 documents a real, specific way an otherwise fully correct calculation still loses a mark, which is this tariff's own version of "good pass, not quite full marks":

Does my percentage answer carry its own % sign? Confirmed in two non-adjacent series: Oct 2022 and Jan 2025 (CALC-1). The underlying arithmetic can be exactly right and this alone still caps the mark one below full.
Have I actually multiplied by 100 in the formula, not just written the right final figure by coincidence? Confirmed the same way, Oct 2022 (CALC-2) — the Knowledge mark specifically attaches to the ×100 step, not only to the final number.
Is my answer rounded to exactly the number of decimal places the question specifies — no more, no fewer? Confirmed across three series: Jan 2023, June 2023, Jan 2025 (CALC-3).
Have I used ARR's own formula (average profit per year ÷ cost of investment × 100), not simple payback's years-and-months answer? Confirmed on Jan 2022's Bramwell Brown question, where real candidates "confused ARR with simple payback" (CALC-5).
On a YED calculation, is quantity demanded over income — not income over quantity? Confirmed on Jan 2025's Spotify question, where candidates "reversed the YED formula" (CALC-6).
Do I actually know the gearing-ratio formula, rather than approximating it? Confirmed independently in two non-adjacent series — Oct 2022 and Jan 2025 — both reporting candidates who "did not know the gearing ratio formula at all" (GEAR-1).

#2.2 Explain (4 marks) — the same discrete-mark shape

Does my answer open with the way, reason, impact or aim the mark is actually for — not a bare definition of the term or stakeholder first? Confirmed across four of the six mined series carrying an Explain question — June 2023 (Lush), Oct 2022, June 2022 and Jan 2025 all repeat this warning near-verbatim (EXP-1). This is a structural cap, not partial credit for a correct but misdirected opening sentence.

#2.3 Discuss (8 marks) — ceiling is Level 3 (6–8); no Level 4 exists on this tariff

Have I named a second, developed side, evidenced from the extract — not a single well-applied case? The confirmed real anchor: Oct 2021's IAG/Air Europa question, where "many candidates ignored the command word 'Discuss' and only gave a one-sided response." The exact shape of the fix is also real and evidenced: marks were given for recognising IAG's takeover disadvantages "might be short lived" given its own prior integration experience — the second side has to be built from evidence already in the extract, not invented [forge-level-threshold-guide.md §2.3].
Have I resisted adding an unneeded conclusion? Appendix 6 fixes Discuss's own ceiling shape before any script is read: "a brief assessment… no conclusion required" [forge-level-threshold- guide.md §2]. A conclusion at this tariff is not rewarded, and time spent writing one is time this paper's tight Section A budget cannot spare (DISC-1).
Have I moved from naming a mechanism to actually attaching it to this business's own situation? The confirmed Level 1→2 boundary: Jan 2023's Brompton Bikes question saw weaker candidates "stating or listing different types" of economy of scale rather than explaining why each one actually lowers Brompton's own unit costs [forge-level-threshold-guide.md §2.2].

#2.4 Assess (12 marks) — ceiling is Level 4 (9–12); this paper's Q1(d)/(e) never carry the lighter

10-mark Units-1/2 version

Am I answering the question actually asked, or the analysis of the tool itself? The confirmed real trap: Summer 2024's Bosch SWOT question saw candidates who produced "a SWOT of Bosch itself, without discussing the benefits of using the tool for developing Bosch's strategy," reported as "typically placed in the lower levels" — a correct, applied SWOT that still answers the wrong question [forge-level-threshold-guide.md §3.2].
On an HR/reward question naming a specific outcome (labour turnover), have I actually argued that outcome — not the adjacent productivity/motivation mechanism? Confirmed independently on two separate real questions, both examiner reports naming the same trap in near-identical language: Lush's ESO question (June 2023) and Five Guys' financial-rewards question (Oct 2023), both reporting candidates held back for focusing "on mainly productivity or motivation rather than labour turnover" [forge-level-threshold-guide.md §3.3; signal-examiner-intelligence.md §4].
Does my closing sentence take a position and weigh it, or just stop once both sides are stated? This is the single most consequential threshold on this tariff. Confirmed across three non-adjacent series — Oct 2021 (BA), Oct 2023 (Five Guys), Jan 2024 (Patagonia) — each examiner report stating some variant of "a conclusion/judgement is required… but is not often seen" [forge-level-threshold- guide.md §3.4; signal-examiner-intelligence.md §2].
Is my closing judgement a genuine decision with a stated condition, or a repetition of the balance already argued above it? A sharper, distinct failure one step past "no conclusion at all," freshly verified against the primary June 2022 ER for Peloton's takeover of Precor: a real candidate conclusion was marked down specifically because it was "just a repetition of earlier points and did not come to a judgement" [forge-judgements-guide.md §3]. A fluent closing paragraph that restates "there are advantages and disadvantages" in different words satisfies the instruction to write a conclusion without satisfying what Level 4 actually asks for.
If my judgement genuinely can't be more certain than the evidence allows, have I said so — or have I invented false confidence? The same Peloton question's own indicative content states, verbatim, that a merger or takeover of this kind "is thought to lead to improvements in only 50% of all mergers and takeovers" [forge-judgements-guide.md §3] — Pearson's own stated base rate, not this document's estimate. A judgement built on this business's own extract-specific detail (adjacency, an already-skilled workforce) and honest about that 50% ceiling on confidence is the credited shape; one that ignores the base rate to sound more certain is not.
On a question with a stated dependency in its own indicative content, have I actually restated that dependency as my condition, rather than inventing an unrelated one? Freshly verified against the primary January 2025 mark scheme for Spotify vs Audible: the indicative content's own final bullet frames the entire answer as conditional — "the outcome will depend on Spotify's ability to effectively navigate these challenges" — not a verdict on whether Spotify wins or loses [forge-judgements-guide.md §1]. A candidate who writes some version of "this depends on whether Spotify's audiobook entry succeeds where its own podcast investment already failed" is reconstructing the mark scheme's own last move, not applying an external technique to it.
Have I reached for a taught framework the question didn't actually ask for? Confirmed on the same Spotify/Audible question: candidates who defaulted to Ansoff's Matrix had their response "steered" toward the familiar framework rather than Spotify's own stated intentions (FRAME-1).
Have I kept a technique's evaluation tied to the specific, named company — not turned it into a generic "this technique is useful" essay? Confirmed on Summer 2024's Burger King CPA question: a genuine 18/20 exemplar lost marks specifically on "an underdeveloped conclusion," not on the arithmetic (TECH-1).

#2.5 Evaluate (20 marks) — ceiling is Level 4 (15–20)

Am I using the extract's own evidence to support each point, or narrating what it already says? The confirmed real anchor: Oct 2023's Disney question, where many candidates were "just paraphrasing the extracts and describing the events," against responses that "really used the extracts to contextualise the potential issues" — the difference the report names is not which side was argued, but whether the extract content did analytical work or was simply restated [forge-level-threshold-guide.md §4.2].
Are the figures or qualitative details I've quoted actually doing evidential work in my sentence, or just present? The confirmed real anchor, a genuine 18/20 script: June 2023's Apple Watch question ties each of three forces to a real figure (Apple's 33.5% smartwatch share; the 2022 supply-chain disruption) rather than narrating that the forces exist [forge-level-threshold-guide.md §4.3].
Does my conclusion propose something concrete — a solution or recommendation — or only a judgement between the sides already argued? This is the clause Evaluate's Level 4 adds beyond Assess's own "supported judgement." The confirmed real anchor: Summer 2024's Burger King CPA script, correct on the calculation and the analysis, still held below full marks specifically by "an underdeveloped conclusion" [forge-level-threshold-guide.md §4.4]. June 2023's Sosyo/Hajoori franchising question supplies the positive version of the same threshold — a genuine 17/20 script credited for proposing a genuine alternative, but only after fully evaluating franchising itself, not displacing it [forge-level-threshold-guide.md §4.4].
On a decision-tree or CPA question, have I gone past the number to say what the comparison is a trade-off between? A distinct AO4-specific gap, verified against October 2024's Center Parcs decision-tree question: a candidate can compute both EMVs correctly (€80m vs €33m) and still sit at Level 3, not Level 4, by stopping at "€80m beats €33m, so build the new village." Level 4 is reached by naming why the higher-EMV option is also the more volatile one — the mark scheme itself cautions that the larger payoff "may not" be the right basis for the decision alone, given its own failure probability [forge-evaluations-guide.md §2]. This is a calculation success and an evaluation shortfall at the same time — a different failure from a wrong EMV figure, and one that survives even once the arithmetic is entirely correct.

#3. Your own error-log, read as a top-band diagnosis

forge-error-pattern-tracker.md Part 1 and Part 2 give you a log to record which named category a mistake falls into. This table adds the piece that log doesn't state on its own: what each category actually costs you against the ceiling, using the same citations as §2 above.

Code (`forge-error-pattern-tracker.md`)	What it caps, specifically	Where this is shown
`CALC-1`–`CALC-6`, `GEAR-1`	Structurally short of full marks on an otherwise correct calculation — not partial credit for the right method	§2.1 above
`EXP-1`	Structurally short of full Explain marks — a correct definition first, not the way/reason/impact/aim	§2.2 above
`DISC-1`	Caps Discuss at Level 2 — the ceiling this whole tariff is named for	§2.3 above; IAG anchor
`HOL-1`	Caps any levels-based tariff below the top level even when evaluation is attempted, because marking is holistic, not a checklist of Assessment Objectives — confirmed in 6 of 6 mined examiner reports	`forge-level-threshold-guide.md` §1
`APP-1`	Reads as narrated knowledge, not evidence for an argument — caps Discuss/Assess/Evaluate below their own top level regardless of how balanced the coverage otherwise is	§1 above; BA and Disney anchors
`ASS-1`	Caps Assess at Level 3 — a real, repeated gap across four separate series	§2.4 above
`FRAME-1`	Caps Assess where the taught framework displaces the question's actual, named ask	§2.4 above; Spotify/Audible anchor
`TECH-1`	Caps Assess/Evaluate where a correct calculation is not matched by an evaluation anchored to the named company	§2.4–2.5 above; Burger King anchor
`331-1`–`331-4`, `332-1`, `334-1`–`334-5`, `336-1`–`336-4`	Topic-specific misreadings that keep a response answering the wrong (adjacent, more familiar) question — caps the level regardless of how accurate the content actually written is	`forge-error-pattern-tracker.md` Part 2; `signal-examiner-intelligence.md` §3

Categories DISC-1, HOL-1, APP-1, ASS-1, FRAME-1 and TECH-1 specifically cap a levels-of-response answer below its ceiling even when the underlying content is accurate — the actual "good pass vs top band" gap this document is named for. CALC-1–CALC-6, GEAR-1 and EXP-1 cap a discrete-mark answer in a way that is closer to binary (the mark scheme either finds the required element or it doesn't) — a real gap, but a structurally different one. The Part 2 topic codes sit underneath whichever tariff they were tested at; they are the content reason a candidate reaches for the wrong lever, where ASS-1/FRAME-1/TECH-1 name the technique consequence of doing so.

#4. Danger-phrase quick check — which named traps are specifically top-band ceilings

forge-exam-technique-guide.md §4 indexes this paper's traps across four categories (A–D). Most of category C in particular are knowledge-level confusions (a misunderstood term, a mixed-up mechanism) that stop an answer scoring at all, not gaps between a good pass and the top band. The subset below is specifically the second kind — confirmed cases where the content was substantively right but a specific, nameable habit still kept the answer off the ceiling:

One-sided coverage, on any of the three levels-based tariffs — the paper's single highest-frequency ceiling trap (§1, §2.3–2.5 above); confirmed across Discuss, Assess and Evaluate independently.
Extracts paraphrased or copied rather than used as evidence — confirmed on BA (Oct 2021) and Disney (Oct 2023), non-adjacent series and topics [forge-exam-technique-guide.md §4D.1].
A closing paragraph that repeats the balance already argued, rather than stating what it depends on — the Peloton/Precor finding, genuinely distinct from "no conclusion at all" (§2.4 above).
Reaching for a taught framework the question didn't ask for — confirmed on Spotify/Audible [forge-exam-technique-guide.md §4D.2].
Evaluating a technique generically rather than tied to the named company — confirmed on Burger King's CPA question, where a correct calculation and a real 18/20 script were still held back by this specific habit [forge-exam-technique-guide.md §4D.3].
A calculation-correct decision-tree answer that never weighs what the comparison trades off — the Center Parcs EMV finding (§2.5 above); not yet a named item anywhere else in this technique-hub.

#5. A worked self-diagnostic pass, using a real anchor already cited elsewhere

This section does not introduce a new exam question. It shows how §2.4's checklist, run honestly, would have caught the exact gap the real June 2022 examiner's report already names for Peloton's takeover of Precor — so nothing here is invented.

The real situation. A response develops a genuine, balanced case for Peloton's takeover of Precor — speed to market, distribution reach, and skilled staff against culture-clash and diseconomies-of-scale risk — then closes with a paragraph that restates both sides again in slightly different words.
Run §2.4's fourth question against it: "Does my closing judgement take a position and weigh it, or just stop once both sides are stated?" Then run the fifth: "Is my closing judgement a genuine decision with a stated condition, or a repetition of the balance already argued above it?" Answer to both: no — this is exactly the shape the real examiner's report describes as "just a repetition of earlier points and did not come to a judgement" [forge-judgements-guide.md §3].
What the fix looks like, built only from the same real, mark-scheme-verified content already cited above: naming that the mark scheme's own indicative content states a real base rate — this class of deal "is thought to lead to improvements in only 50% of all mergers and takeovers" — then using the extract's own detail (Precor's adjacency to Peloton's market, its already-skilled workforce) as the specific, checkable reason to expect this deal sits on the successful side of that split, while stating openly that the 50% figure is the honest limit on how confident that judgement can be [forge-judgements-guide.md §3].

The point of walking through it this way is that a student running §2.4's checklist against their own draft, before ever seeing an examiner's report, would land on the identical diagnosis the real report gives after the fact — that is what makes this a genuine diagnostic tool rather than a retrospective commentary.

#6. What this document deliberately leaves out

No grade boundary, raw-mark total, or likelihood figure appears anywhere above — see §0. signal-frequency-and-boundaries.md §3 is the authority on this gap, and this document adds nothing to it.
The 0→1 threshold on every levels-based tariff has no real cited diagnostic. forge-level- threshold-guide.md §§2.1, 3.1 and 4.1 flag that no genuine Level 0/1 borderline script survives in the mined examiner reports on this paper — unsurprising, since a Level 0 response has nothing creditable in it for an examiner to quote. This document cannot offer a self-check for that specific boundary any more than its source could, and doesn't invent one.
Analyse (6 marks) has no self-diagnostic here at all. forge-level-threshold-guide.md §1 names this as a genuine, flagged gap: no WBS13 series read for this course contains a real Analyse-tariff question with a level ladder or exemplar attached, and forge-evaluations-guide.md §1 separately confirms Analyse is the one command word where evaluation is explicitly, structurally excluded by Appendix 6's own wording ("does not include evaluation"). A top-band diagnostic presupposes a real graded answer exists to diagnose against; where none does, there is nothing here to synthesise.
This document gives no worked account of how a mark is chosen within a level's range (why a Level 3 Assess answer gets 6 rather than 7). forge-level-threshold-guide.md §1 names this as a genuine gap in what this course's sources can support, and this document does not guess at it.
The Spotify/Audible and Center Parcs findings used in §§2.4–2.5 rest on a narrower evidence base than the rest of this document. forge-judgements-guide.md §7 notes no examiner's report for January 2025 was obtained locally for that question, so the Spotify material is anchored to the mark scheme only, not to any graded script; forge-evaluations-guide.md §3 notes the same for October 2024's Center Parcs question — no ER exists in this course's sources, and Levels 1–3 of that worked example are this course's own construction against the general descriptor wording, not a real graded script. Both are flagged as such at their source and again here, not treated as more certain than they are.
This document does not cover the paper's own genuine content gaps (Analyse's missing exemplar; organic growth methods, 0 of 12 series; the finer allocation within a level) — those are named in full in forge-level-threshold-guide.md §6 and signal-frequency-and-boundaries.md §2, and this document doesn't repeat or extend that list.