Deep Dive: Forecasting and Investment Appraisal

#CODEX Deep-Dive: Forecasting and Investment Appraisal (WBS13, 3.3.3.1–3.3.3.2)

Provenance — written 2026-09-28. Read and grepped research/veridian/WBS13-verified-facts.md (2,057 lines) for every 3.3.3.1/3.3.3.2 entry, the Quantitative Formulas table, and the Universal calculation-error traps list; read content/veridian/curriculum/wbs13/forecasting-and-investment- appraisal.ts in full (1,007 lines — its own provenance header, all worked chains, the trap-taxonomy and the beyond-spec block); read content/veridian/technique-hub/wbs13/codex-definitions-and- application-bank.md (§3.3.3.1–3.3.3.2), forge-exam-technique-guide.md, and signal-frequency-and- boundaries.md in full; grepped content/veridian/question-bank/wbs13.ts (213 lines, currently zero entries for this topic). No new primary-source fetching was done for this pass — every citation below is inherited from the verified-facts file's own already-checked provenance, not independently re-verified against a fresh PDF pull.

What this document adds. The lesson file already contains the full pedagogical derivation of every mechanism here (worked chains, chain-drills, a mechanism block for both discounting and moving-average noise-cancellation) — this document does not re-run those. The CODEX definitions bank already glosses every term with its formula and the same one or two real anchoring examples. What neither existing document does is (a) collect the complete real exam history of this specific topic — every confirmed series, question, tariff and what was actually rewarded or penalised, cross- referenced against the spec's own five-part structure — in one place, and (b) name plainly which parts of the spec have the thinnest real exam history in the twelve series this course has read, stated as a floor on confirmed sightings rather than a ceiling on what the specification can still examine. Its micro-examples use original numbers, distinct from the lesson file's Amberview/Larkspur/ Bramwell Brown figures, computed independently with python3 and re-stated in full below so every step is checkable.

Update, 2026-09-29. Re-checked this document against a correction the sibling signal-topic-brief-forecasting-and-investment-appraisal.md had already flagged against §2's table — that Oct 2022 Q1(b)'s 3-year moving average is a real, separately-tariffed Calculate question, not indicative content borrowed from elsewhere. Verified directly, not taken on the brief's word: read Master_Consolidated_Docs/Business_Paper3_Mark_Schemes.md, ..._Past_Papers.md and ..._Examiner_Reports.md at /Users/neosteinhoff/Documents/SOL/Papers/IAL_Paper3_Complete/ Master_Consolidated_Docs/ — the same locally-cached, genuine-Pearson-archive transcription WBS13-verified-facts.md itself already cross-checks other WBS13 citations against (its Jan 2022 and Jan 2023 provenance rows). That check confirmed the brief's correction and surfaced two further, previously-missed facts checked the same way: (1) the simple payback formula, verbatim, from the Oct 2021 mark scheme, with Airbus's real $366m/$30.5m figures; and (2) a real, confirmed NPV question — Oct 2023 Q1(b), Five Guys' milkshake machine — that the facts bank's own provenance table already records as fully read for this course's research, but, per its own note, mined only for financial-rewards/ROCE content (3.3.5) and never cross-referenced back into investment appraisal (3.3.3.2). All three corrections are applied throughout §1–§3 below, each citing the specific archive file and question checked. This is a check against an already-established, previously-cited local archive, not a fresh qualifications.pearson.com fetch or independent verification of that archive's own fidelity to the original Pearson PDFs.

---

#1. The real theory, compressed

#1.1 Quantitative sales forecasting (3.3.3.1)

The technique rests on one model: any period's reported figure is treated as an underlying trend (the slow-moving true direction) plus a variation component (short-term noise or a regular seasonal/cyclical pattern) layered on top. Three named tools operate on that model:

(a) Moving averages smooth a run of past figures into a single trend line by averaging a fixed number of consecutive periods. Two genuinely distinct sub-cases exist, and the spec names both by number: a 3-period (or 3-year) moving average, and a 4-quarter moving average.

• With an odd window (3 periods), the average lands exactly on the middle period — no further step is needed. Micro-example, a bakery's weekly sales (£000), Weeks 1–5 = 40, 48, 44, 52, 50: the 3-week moving average centred on Week 2 = (40+48+44)÷3 = 44.0; centred on Week 3 = (48+44+52)÷3 = 48.0; centred on Week 4 = (44+52+50)÷3 = 48.67.
• With an even window (4 quarters), each raw average falls at the midpoint between two quarters, not on one — a second step, centring (averaging two consecutive uncentred averages), is required to produce a trend value that lines up with an actual quarter. Micro-example, five quarters' sales (£000) = 30, 50, 38, 42, 34: the first 4-quarter moving total (Q1–Q4) = 160, average = 40.0; the second (Q2–Q5) = 164, average = 41.0; centring these two — (40.0+41.0)÷2 = 40.5 — produces the trend value for Q3, the quarter sitting in the middle of the two spans averaged. This is the exact mechanism the spec's "4-quarter" wording is testing, distinct from the 3-period case in needing this extra step at all.
• Both sub-cases share the same cost: a window of n periods loses periods at both ends of the series (a full window on each side is needed before a trend value can be produced at all), and a longer window loses more data than a shorter one for a smoother result — the direct trade-off the spec's own "limitations" sub-point (3.3.3.1c) is built to test.

(b) Cyclical/seasonal variation, and using it to forecast one specific future period (the wording the lesson file's own provenance note traces to spec 3.3.3.1b via Pearson's Getting Started Guide), is the genuine second half of forecasting that a trend line alone cannot do: a moving-average trend only ever says "sales are rising," not how one specific upcoming quarter compares to that rise. The method: for every period with a computed trend value, take variation = actual − trend; group those variations by their position in the cycle (all the Q1 variations together, all the Q3 variations together, and so on) and average each group; then forecast a specific future period by extrapolating the trend forward to that period and adding back its own average variation. Compressed micro-example (trend values taken as already computed, to isolate this step specifically): a firm's Q1 trend value was £52k in Year 1 (actual £45k, variation −£7k) and £58k in Year 2 (actual £50k, variation −£8k) — average Q1 variation = (−£7k + −£8k)÷2 = −£7.5k. If Year 3's Q1 trend is extrapolated to £64k, the Q1-specific forecast is £64k + (−£7.5k) = £56.5k, not the bare extrapolated £64k — omitting the variation step overstates this firm's Q1 by exactly that £7.5k gap, the same kind of quantified cost of skipping the step that the lesson file's own worked chain demonstrates on its own (larger) dataset. The averaging step is the identical noise-cancelling logic moving averages use: any one year's Q1 gap is partly a one-off shock, and averaging several years' worth of same-quarter gaps together gives a more stable read of how that quarter structurally sits relative to trend.

(c) Scatter graphs and the line of best fit extend the same forecasting logic to two paired variables (e.g. advertising spend and sales revenue) rather than one variable over time: points are plotted with no joining line, then a single straight line is drawn by eye, as close as possible to as many points as possible. Reading a value inside the plotted range is interpolation; reading a value beyond the last real point is extrapolation — the same forward-looking assumption (that the relationship keeps holding beyond where it was actually observed) that makes moving-average extrapolation a genuine forecast rather than a description of the past. No formula for the line itself is required on this paper.

(d) Limitations (3.3.3.1c). Genuinely two-sided, and confirmed from two different real mark schemes on two different real companies: a real Pets at Home mark scheme cautions that past growth "does not mean this will continue" (Oct 2022 MS Q1c), while a real Spotify mark scheme credits the opposite case — a forecast that actually held, off by "2 million more than the forecast number" in the business's favour (Jan 2025 MS Q1c). Both are indicative content for the same 8-mark Discuss question-type, on two different real series — the balance itself, not either side alone, is what the mark scheme is built to reward.

#1.2 Investment appraisal (3.3.3.2)

All three methods take the same two inputs — the initial investment and a set of forecast net cash inflows, one per year — and ask a genuinely different question of them.

(a) Simple payback asks only how long the cash takes to come back: run a cumulative total of net cash inflow and find where it first reaches the initial investment. When the crossing happens partway through a year (the normal case), the fraction of that year still needed is (investment remaining at the start of that year) ÷ (that year's full inflow), converted to months by ×12. Micro-example: a £54,000 investment with net cash inflows of £18,000, £22,000, £24,000 across Years 1–3. Cumulative cash = £18,000 (Y1), £40,000 (Y2), £64,000 (Y3) — the crossing happens inside Year 3. Amount still needed at the start of Y3 = £54,000 − £40,000 = £14,000; £14,000 ÷ £24,000 = 0.583 of the year; 0.583 × 12 = 7.0 months exactly. Payback = 2 years, 7 months.

The one real, confirmed payback Calculate question on this paper takes a genuinely simpler shape than the crossing-point case above, and its formula is now verified rather than left as the un-sourced line this document previously carried. Oct 2021 Q1(b) (4 marks, Airbus A350-1000 aircraft) supplies a single average yearly net cash flow rather than a year-by-year series, so payback divides out cleanly with no crossing point or month-fraction step at all. Verified directly against the real mark scheme: simple payback = cost of initial investment ÷ average yearly net cash flow; applied to Extract B's own figures (purchase price $366m, average yearly net cash flow $30.5m, a 27-year, 380-seat aircraft), $366m ÷ $30.5m = 12 years exactly. This closes what the facts bank itself flags as an open gap — no verbatim payback formula previously captured — but it also means the crossing-point/month-fraction mechanic above, while a correct and spec-required technique for uneven year-by-year cash flows, has not itself been the substance of the one real Calculate question found for this sub-point; Airbus's flat average sidesteps that step entirely.

(b) Average rate of return (ARR) expresses profitability as a single average annual percentage: average annual profit ÷ cost of initial investment × 100. The genuinely distinct sub-case here is what form the question hands over. The real anchoring example (Bramwell Brown, Jan 2022 MS Q1a) supplies annual profit figures directly, needing no adjustment before the formula is applied: six years totalling £287,550, average £47,925, ARR = £47,925 ÷ £150,000 × 100 = 31.95%. When a question instead supplies net cash inflow (the same figures payback uses), ARR's standard simplification applies first — total profit = total net cash inflow − initial investment — before the formula can be run at all. Two original micro-examples on the same £80,000 investment make the distinction concrete:

• Profit given directly: four years of profit £10,000/£14,000/£18,000/£22,000, total £64,000, average £16,000; ARR = £16,000 ÷ £80,000 × 100 = 20%.
• Cash inflow given instead: four years of net cash inflow £26,000/£30,000/£34,000/£38,000, total £128,000; total profit = £128,000 − £80,000 = £48,000, average £12,000; ARR = £12,000 ÷ £80,000 × 100 = 15%.

Same formula, same size of investment, a materially different answer (20% vs 15%) purely because of which raw figure — profit or cash inflow — the question started from. A real Jan 2022 examiner report on this exact question confirms this is a genuine, observed failure point, not a hypothetical one: some candidates scored only the method mark for correctly placing £150,000 as the denominator, having not known the ARR formula at all; others gave an answer "in years and months," confusing ARR's percentage output with payback's; a further group calculated ARR correctly but lost a mark for dropping the % sign.

(c) Net present value (NPV), the only discounted-cash-flow method this paper examines (the spec explicitly scopes this to NPV, not IRR), converts every future cash flow into today's money before comparing it to the investment. The discount factor for year n at rate r is 1÷(1+r)ⁿ; multiplying a future cash flow by its year's factor converts it to a present value; summing every year's present value and subtracting the initial investment gives NPV. Micro-example: a £60,000 investment with net cash inflows of £25,000 in each of Years 1–3, discounted at 10%. Discount factors (3dp): Year 1 = 0.909, Year 2 = 0.826, Year 3 = 0.751. Present values: £25,000×0.909 = £22,725; £25,000×0.826 = £20,650; £25,000×0.751 = £18,775. Total present value = £62,150. NPV = £62,150 − £60,000 = +£2,150 — positive, so the project is worth accepting: it returns more, once every pound is priced at what it is actually worth today, than the £60,000 it costs today.

A real, confirmed NPV question now exists on this paper — Oct 2023 Q1(b), Five Guys' £10,000 milkshake machine over five years — using exactly this discount-factor-supplied shape (5% factors to 3dp given directly in the extract, not derived from 1÷(1+r)ⁿ). This confirms, rather than merely infers, the exam format §3 previously flagged as unverified; see §2 for the full figures and the examiner report's own reaction to it.

(d) Calculation and interpretation, together. A single set of cash flows can rank differently under each method, because timing changes the answer for NPV but not for payback's yes/no crossing point and not at all for ARR — ARR only ever sums total profit and divides by years, never asking when inside those years the cash actually arrived. This is why the spec pairs "calculation" with "interpretation" as one combined sub-point rather than two separate ones: a number without a stated decision rule (accept if payback is inside the target period; accept if ARR clears the target percentage; accept if NPV is positive) is only half the answer this spec point is built to test.

(e) Limitations. Confirmed directly from a real Jan 2023 examiner report on Brompton Bikes' new factory investment (Q1e, 12-mark Assess): the strongest real answers linked method choice to the nature of the business — a fast-changing, technology-driven market favouring payback's speed-of- return focus over a profitability-only lens — while the standard counter-argument, confirmed the same report, is that payback "ignored the time value of money or the overall profitability of the investment." That one sentence names two separate weaknesses at once: payback treats every pound as equal regardless of when it arrives, and it stops looking at the project the moment cash is recovered, even if the most profitable years are still ahead. ARR shares the first weakness (blind to timing) but not the second (it does use every year's profit) — and NPV is the one method built specifically to fix the timing weakness, at the cost of needing a discount rate estimate and a longer forecast that is itself only ever an assumption.

---

#2. Every real exam angle this topic has actually been tested from

Cross-referencing the verified-facts file's own citations against the signal-frequency document's count (twelve WBS13 series read in total, of uneven depth):

That is the complete set now confirmed for 3.3.3.1–3.3.3.2 specifically, across all twelve series this course has read — two rows added on this pass (Oct 2022 Q1(b), Oct 2023 Q1(b); see the provenance note above for exactly which archive files were checked to add them). Three things follow directly from the corrected table, stated plainly rather than left implicit in a bare count:

• NPV has now been confirmed once, not zero times. Oct 2023 Q1(b), 4 marks: Five Guys' milkshake machine, discounting £5,000 a year for five years at 5% against a £10,000 initial cost. Discounted cash flows £4,760/£4,530/£4,315/£4,110/£3,920, total £21,635, NPV = £21,635 − £10,000 = +£11,635 — independently re-summed and confirmed correct against the extract's own figures. The examiner report states plainly that this was a genuine first — "the first time this question has been asked" — and that it was answered poorly, with many candidates not understanding the concept at all: real, quantified evidence that this sub-point is examined and that students are currently weak on it, not evidence it is rare content to under-prepare for. This corrects both this document's own earlier claim and the signal-topic-brief's "not yet confirmed at all" framing for NPV — both were wrong in the same direction, because the Oct 2023 series had already been fully read for this course's research but, per the facts bank's own note, mined only for its financial-rewards/ROCE content (3.3.5), never cross-referenced back into investment appraisal (3.3.3.2).
• The 3-year moving average has a real, standalone Calculate appearance, and this document previously understated it. Oct 2022 Q1(b) is a genuine Knowledge-1/Application-2/Analysis-1, 4-mark Calculate question in its own right — not, as an earlier draft of this document stated, mechanics "confirmed... only via... a mark scheme's indicative content." Forecasting's limitations sub-point (3.3.3.1c) has still only ever been tested as an 8-mark Discuss (Pets at Home, Spotify) — that part of the earlier analysis stands — but the moving-average calculation mechanic itself is confirmed, separately-tariffed Calculate content, tested in the same series and off the same extract as one of those two Discuss questions.
• Investment appraisal has now been tested as a direct multi-mark Calculate question three times (Airbus payback, Bramwell Brown ARR, Five Guys NPV), plus once as a 12-mark Assess (Brompton Bikes) — every one of the paper's three appraisal methods now has a real, confirmed Calculate hit, which was not true before this pass located the Five Guys question.

No likelihood claim follows from any of this — a 1-of-12 sighting (NPV) is stated as a floor on confirmed sightings, not a ceiling on what the specification can still examine, consistent with the signal-frequency document's own standing rule. No grade-boundary data exists anywhere in this course's WBS13 sources for either sub-topic, and none is stated here.

---

#3. What remains genuinely unverified for this specific topic

• RESOLVED this pass — the simple payback formula. Previously flagged as never captured verbatim; now confirmed directly against the Oct 2021 mark scheme (Q1(b)): simple payback = cost of initial investment ÷ average yearly net cash flow, applied to Airbus's own $366m/$30.5m for 12 years exactly (see §1.2a and §2). What remains genuinely unverified is narrower than before: whether the crossing-point/month-fraction mechanic — needed whenever cash flows are uneven year to year rather than a flat average, and taught in the lesson file and this document's own §1.2(a) micro-example — has itself ever been the substance of a real Calculate question. Airbus's flat average sidesteps that step entirely, and no other confirmed payback Calculate hit exists on this paper to check against.
• RESOLVED this pass — discount factor tables. Previously stated as an inference from format alone, since no real NPV question had been located. Now confirmed directly: Five Guys' real Oct 2023 question supplies 5% discount factors to 3dp in the extract itself (0.952, 0.906, 0.863, 0.822, 0.784 for Years 1–5), exactly the shape this document's own NPV micro-example in §1.2(c) follows.
• Multiplicative seasonal variation (variation as a ratio rather than an additive actual-minus- trend figure) does not appear anywhere in the verified-facts file or the spec wording it transcribes for this paper; nothing here should be read as implying it is required, and this document does not teach it. Nothing found on this pass touches this item — it stays open.