A Case Answer, Marked Line by Line

The case, and what the interviewer is writing down

Interviewers do not mark recall of a framework. They mark whether a short argument holds: whether the structure fits the question, whether the numbers are right, whether the conclusion follows from them and whether the candidate communicated it so that a client could act. The weighting differs by firm, and the firms do not publish it, so what follows is our reading of how most interviewers describe it.

The case below is invented. Northway Coaches, a regional coach operator, saw operating profit fall from £6.0 million to £3.6 million over two years. The client asks why, and what to do. You are given the figures in the first table and told that mileage rose from 6.00 million miles to 6.25 million.

Mark one: the opening and the structure

A strong opening restates the question in one sentence, asks one clarifying question that changes the analysis, and states an objective. Here the candidate says: “So the goal is to understand why operating profit fell by £2.4 million and what would restore it. Is the business one price list across all services, or a mix of contract and charter work?” The second question matters because it decides whether price is a lever at all.

The structure is then a tree, not a list. Profit is revenue less costs. Revenue is miles multiplied by price per mile. Costs split into fuel, drivers, vehicles and overheads. The marks go to the fact that each branch can be tested with a number from the exhibit. The weak version is a memorised list of headings that the candidate then works through in order without noticing which one the data points to.

Spoken, it sounds like this: “I would like to split profit into revenue and cost. On revenue, I would look at miles and price per mile. On cost, I would look at fuel, drivers, vehicles and overheads, and test each per mile so that growth in volume does not hide a change in price. Can I start with the exhibit?” It takes twenty seconds, commits to nothing the data cannot support and tells the interviewer the route before the walk begins.

Mark two: the analysis, and the move that finds the answer

The decisive step is to convert totals into units. Revenue rose £2.0 million and costs rose £4.4 million, which explains the fall, but not why. Dividing by miles does. Revenue per mile is £8.00 in both years, so price has not moved. Fuel per mile rose from £1.60 to £1.92 and driver cost per mile from £2.80 to £3.12. Vehicles fell slightly, from £1.40 to £1.36, and overheads per mile fell from £1.20 to £1.02 because the same fixed costs were spread over more miles.

Mark three: the check that earns the most

Profit per mile fell by £0.42. The changes on the cost lines are +0.32, +0.32, -0.04 and -0.18, which sum to +0.42. A candidate who says that aloud has proved the bridge closes, and the interviewer can stop worrying about the arithmetic. This is the small habit that distinguishes a clean answer from a lucky one.

The common error is also visible here. Fuel spend rose 25 per cent, from £9.6 million to £12.0 million. A candidate who stops there reports a 25 per cent fuel price rise. Mileage grew 4.2 per cent, so spend per mile rose by a factor of 1.25 divided by 1.042, which is 1.20, or 20 per cent. The first figure confuses volume with price and, carried into a recommendation, overstates the problem by a quarter. How the slip is handled matters more than the slip: one caught by the candidate costs little, and one that reaches the conclusion unnoticed costs a great deal.

Mark four: the answer first, then the recommendation

The synthesis should arrive in the first sentence, not the last. A model delivery runs: “Profit fell because fuel and driver costs per mile each rose by 32 pence while revenue per mile stayed flat. Extra mileage helped overheads, but not enough. To restore £1.00 of profit per mile at today’s costs, the price per mile would need to rise from £8.00 to about £8.42, a 5 per cent increase, which is about £52.6 million of revenue on 6.25 million miles.”

The recommendation then names its own risks, because a client will ask. A 5 per cent price rise is possible on charter work and slower on contracts that renew annually. Some of the rise could be recovered by fuel-efficiency measures or driver scheduling rather than price. The next step is to split the revenue and cost lines by contract and charter, since the answer may differ sharply between them. A recommendation with no risk, or no next step, reads as unfinished.

What happens when the maths goes wrong

Suppose the candidate divides £12.0 million by £9.6 million and says 1.15. An interviewer will usually say nothing, because stopping a candidate mid-calculation interrupts the rhythm and removes the chance to see whether they catch it. A strong candidate does a quick reasonableness check aloud: “9.6 plus a quarter of 9.6 is 12.0, so that is 1.25, not 1.15.” The slip has become evidence of a good habit.

The same slip with no check is the expensive version. The wrong figure feeds the per-mile calculation, the bridge no longer closes by a few pence, and the recommendation rests on a number that does not tie. At that point the interviewer is no longer assessing the answer but the process that produced it. Our judgement is that most firms treat an unnoticed carried error as a more serious signal than a slow start.

Marking your own practice

The easiest way to improve is to record yourself working a case aloud and then mark the recording against four questions. Did the first sentence of the conclusion contain the answer? Did every number I said get checked against something else? Did the bridge close? Did the recommendation name a risk and a next step?

Four yeses on a case means the structure is holding. A no usually points to a single fixable habit, and fixing one habit per week moves faster than rehearsing new frameworks. If you cannot find a partner to interrupt you, ask the recording to do it: pause it after each number and say what you would check.

Where the marks are usually lost

Three patterns account for most of what separates a good answer from an adequate one, and none of them is technical. The first is delivering the conclusion last, so the interviewer has to wait for the point. The second is arithmetic done in silence, which means an error cannot be caught or credited. The third is a recommendation without a risk or a next step.

All three are fixable with deliberate practice. Work cases aloud, against a clock, with someone who will interrupt. Say the structure before you use it. Say each number as you compute it. End every case with the answer, the main risk and what you would do next. The guides on the case interview and on case maths set out the formats and the drills in more detail.