HookHow the Elizabeth line lost three and a half years
Crossrail was Europe's largest construction project, and in the summer of 2018 its official opening date was still December 2018 — until August 2018, when the announcement came that it would not open on time. The Elizabeth line finally carried its first passengers on 24 May 2022, three and a half years late, with the bill up from a budget of roughly £14.8bn to around £18.9bn. The delay did not come from tunnelling — the tunnels finished years earlier — but from the network's least glamorous tasks: integrating and testing three separate signalling systems. Those tasks sat on the project's critical path, so every week they slipped, the whole railway slipped with them.
B3.3 is the toolkit that exists to stop this happening: forecast demand before you commit (quantitative sales forecasting), test whether the numbers justify the investment (payback, average rate of return, discounted cash flow), put values on uncertain outcomes (decision trees), and schedule interdependent tasks so you know which ones cannot be allowed to slip (Critical Path Analysis). Each technique has a calculation the exam will pay you to execute cleanly — and an assumption buried inside it that the top-band marks come from dragging into the light. The number is Level 2; what the number cannot see is Level 4.
ModelQuantitative sales forecasting — smoothing the noise to see the trend
Sales data jumps around — weather, promotions, luck. A moving average smooths the jumps by averaging each period with its neighbours, so the underlying trend shows through. Once you have a trend, extrapolation extends it forward: if smoothed sales have grown about 5% a quarter for two years, pencil in 5% next quarter. Correlation is the third tool: plot sales against a possible driver (advertising spend, temperature, footfall) on a scatter diagram; a tight upward pattern suggests a positive relationship you can forecast with, a loose cloud says the driver explains little.
All three share the same buried assumption: the future will behave like the past. That is usually true and occasionally catastrophically false. Every UK retail forecast built on 2015–2019 data was destroyed by March 2020; extrapolating Aldi's growth assumed the Big Four wouldn't respond, until Tesco's price-match campaign changed the trend line. And correlation is not causation — ice cream sales correlate with sunglasses sales because summer drives both. State the calculation, then state the world in which it breaks: that pairing is the full answer.
A station coffee kiosk sells 120, 150, 180 and 210 units across January to April. Three-month moving averages: Jan–Mar = (120 + 150 + 180) ÷ 3 = 450 ÷ 3 = 150, centred on February. Feb–Apr = (150 + 180 + 210) ÷ 3 = 540 ÷ 3 = 180, centred on March. The trend is +30 units per month, so extrapolation forecasts roughly 240 units for May (the April-centred average would be 210, plus 30). One sentence of caution completes it: the forecast holds only if the drivers of that trend — commuter footfall, no new rival kiosk — hold too.
ModelInvestment appraisal — three rulers for the same decision
Should the firm spend £X now for a stream of returns later? Edexcel gives you three rulers. Payback period: how quickly do cumulative net cash inflows repay the outlay? It measures speed and liquidity risk, and cash-poor firms live by it — but it ignores everything after payback. Average rate of return (ARR): average annual profit as a percentage of the initial outlay, comparable directly with interest rates or a firm's target return — but it treats a pound in year four as worth a pound today. Net present value (NPV) fixes exactly that flaw: multiply each year's cash flow by a discount factor reflecting the time value of money, sum them, subtract the outlay. A positive NPV means the project beats the discount rate; a negative one means the money works harder elsewhere.
The judgement layer: the three rulers can disagree, and which should win depends on the firm. A start-up with tight cash weights payback; a stable firm choosing between projects weights NPV. And every appraisal is only as good as its forecast cash flows — which come from the forecasting techniques above, with all their fragility. Appraising Crossrail's benefits meant forecasting London commuting decades ahead; the pandemic rewrote those numbers before the line even opened.
A bakery weighs a £500,000 production line returning £200,000 a year in net cash for four years. Payback: £500,000 ÷ £200,000 = 2.5 years. ARR: total inflows £800,000, so total profit = £800,000 − £500,000 = £300,000; average annual profit = £300,000 ÷ 4 = £75,000; ARR = 75,000 ÷ 500,000 × 100 = 15% — comfortably above borrowing costs. NPV at 10% (discount factors 0.91, 0.83, 0.75, 0.68): £200k × 0.91 = £182k; × 0.83 = £166k; × 0.75 = £150k; × 0.68 = £136k. Sum = £634k; NPV = £634k − £500k = +£134,000. All three rulers say yes — but note how the discounting shaved £166,000 off the raw £800,000 of inflows. That is the cost of time made visible.
ModelDecision trees — putting numbers on maybes
A decision tree maps a choice under uncertainty: squares for decisions, circles for chance events, probabilities on each branch, payoffs at the ends. The arithmetic is expected value: multiply each outcome by its probability, add them up, subtract the cost of the choice to get net gain, and pick the branch with the highest. It forces disciplined thinking — you must state your probabilities and payoffs out loud, where they can be challenged — and it strips emotion from ‘gut feel’ decisions.
Its weaknesses are exactly those inputs. The probabilities are usually management estimates dressed as science: nobody knows a launch has a 0.6 chance of success. Expected value is an average across many imaginary repeats, but a firm launches once — an EV of +£1.4m is cold comfort if the 40% failure branch would sink the company. And trees ignore risk appetite and qualitative factors: brand damage, staff morale, strategic fit. Best exam line: a decision tree does not make the decision; it makes the assumptions visible enough to argue about.
A snack brand can launch a new range costing £2m. Market research suggests a 0.6 probability of success (payoff £5m) and 0.4 of failure (payoff £1m). Expected value = (0.6 × £5m) + (0.4 × £1m) = £3.0m + £0.4m = £3.4m. Net gain = £3.4m − £2m = +£1.4m, versus £0 for not launching — so launch. Now stress it: if the success probability is really 0.4, EV = (0.4 × £5m) + (0.6 × £1m) = £2.6m, net +£0.6m — still positive, but the margin for error has more than halved on one changed estimate. Showing that sensitivity IS the evaluation.
ModelCritical Path Analysis — finding the tasks that cannot slip
A project is a web of dependent tasks. Critical Path Analysis draws the network, works out the earliest start time (EST) of each task moving forward and the latest finish time (LFT) moving backward, and identifies the critical path — the longest chain through the network, which sets the minimum project duration. Tasks on it have zero float: any delay delays everything. Tasks off it have float — spare time — so managers know where slack lives and where it doesn't.
The payoffs: management attention lands on the truly critical tasks; resources can be shifted from high-float tasks to critical ones; and materials can be ordered just in time against ESTs, cutting stockholding costs. The limits: the network is only as honest as its duration estimates, and unexpected events re-route the critical path mid-project. Crossrail is the cautionary tale — tunnelling finished with fanfare years before opening, but the critical path ran through signalling integration and testing, and every month that testing slipped moved the opening day one-for-one. A Gantt chart of finished tunnels is no consolation when the zero-float task is the one on fire.
A café refit has three tasks: A, strip the interior (3 days) and B, order and await new fittings (5 days), which can happen simultaneously; then C, installation (4 days), which needs both complete. Paths: A→C = 3 + 4 = 7 days; B→C = 5 + 4 = 9 days. The critical path is B→C and the minimum duration is 9 days. Task A has float = 5 − 3 = 2 days: it can slip two days without delaying reopening, so the manager should watch the fittings order, not hurry the strip-out.
VocabularyKey terms the mark scheme pays for
TrapsMisconceptions that cost marks
ExamWhat examiners want
This is the most quantitative section of Theme 3, and Edexcel's quantitative marks reward ritual: state the formula, substitute the numbers, compute, then interpret the result in one sentence tied to the decision. The interpretation line — ‘payback of 2.5 years is within the firm's 3-year ceiling, so proceed’ — is a mark candidates throw away constantly. Show working even when the answer is wrong; method marks survive arithmetic slips. Keep units and rounding disciplined: £000s stated, percentages to one decimal place unless told otherwise.
On ‘assess’ and ‘evaluate’ questions, the technique itself is the evaluation target. Interrogate the inputs — whose probability estimates, how old is the sales data, what discount rate and why — and weigh the technique against the context: payback for a cash-strapped start-up, NPV for a utility with patient capital, CPA only as good as its duration estimates. The strongest candidates run the calculation AND write the sentence ‘this number inherits every weakness of the forecast behind it’. That sentence, applied to the case, is routinely the difference between Level 3 and Level 4.