Guided Cases: Prompts, Fifty-Seven Cases, One Clock

What the fifty-seven cases are made of

Case banks are usually sold by size. Size tells you very little. What decides whether a library helps is whether it covers the type you keep losing on, so here is the composition, counted from the catalog file the library reads.

Those six add to 57, which is the whole library. By running mode the split is 23 guided and 34 voice. By difficulty it is 32 medium, 19 hard and 6 easy. By the house style each case imitates it is 19 McKinsey, 16 Bain, 15 BCG, 4 independent, and one each for Kearney, L.E.K. and Deloitte. Fourteen industries appear, with consumer and retail the largest at 15 cases and food and beverage next at 8.

Two of those attributes deserve a second look. Difficulty is skewed toward the middle, with 32 of 57 cases at medium, so a candidate who needs a hard case every day will exhaust the 19 hard ones inside three weeks. And the house style is a label on the problem's shape rather than a claim about a firm's real question bank, since no firm publishes one. Treat it as a way to vary the pattern you rehearse.

Read the pricing row as a gap rather than an oversight. Two pricing cases in 57 means the library trains pricing thinly, and a candidate interviewing somewhere that leans on pricing work should know that before subscribing. You can browse the whole list, filtered by all five of those attributes, at the case library.

What a prompt changes in case interview practice

Reading a worked case is pleasant and close to useless. Telling a good structure from a bad one is a skill most candidates already have before they start preparing. Building one at speed, out loud, with somebody waiting, is the skill preparation has to add. A transcript exercises the first. A prompt exercises the second.

The mechanism is simple. At each step the case asks for your output before it shows you anything, so you commit. Then it shows a strong version of the same step. Your own attempt is still in front of you, which is what makes the comparison informative: you can see which branch you missed and, more usefully, why you missed it. Model answers read before you attempt anything teach almost nothing, because agreement feels identical to competence.

There is a second effect, and it shows up in how long a session lasts. A blank case ends early for most people, because the first hard step produces a stall and a stall produces a closed tab. A prompted case survives the stall, since the next instruction arrives whether or not you solved the last one. Volume of completed cases matters more in the first two weeks than the quality of any single attempt, and prompting buys that volume.

The cost of prompting is real and worth naming. A prompted case is easier than a live one. Somebody has already told you that this step calls for a structure and that step calls for arithmetic, and in a room nobody will. That is the gap the spoken mode exists to close, and it is described on voice mode.

The weights a guided case is training

The structure you announce in step one is the most expensive sentence in a case. Our rubric for it puts almost half its points on one dimension. These weights are ours, read from the grading code, and no firm publishes a scoring sheet of its own.

Two complete cases a firm publishes free

Before paying anyone for practice cases, work the two that Bain puts on its own site. They are the strongest free rehearsal available, because the firm doing the hiring wrote them.

1. The coffee shop case gives every input and prints its own equation. A town of 100,000 people, one cup a day, three pounds a cup, one pound of variable cost, 245,610 pounds of setup and 163,740 pounds of annual operating cost. It resolves to 204,675 cups in year one. You can check that by hand, which almost no published case allows, and our page on case math recomputes it and finds the second break-even hiding inside it.
2. The retail case gives a declining women's fashion business and two named options: a rewards program at 25 percent participation of 10 million customers with a one-off 50 dollar activation and a permanent 20 percent discount, or intermittent sales at 20 percent off for three months a year. The published answer takes the second option and says it is "projected to generate $50M more than the awards program (Option A)".
3. The stronger version of that answer, which the firm also prints, adds what the recommendation could get wrong. It says option A could win over a longer horizon if year two enrollment rises, names styling and technology as unaddressed risks, and proposes a consumer survey alongside the chief executive. That is the difference between an answer and a recommendation.

One caution on the retail case. Its calculation tables are published as images, so we cite the conclusion the firm states and we do not reconstruct the arithmetic behind the 50 million dollar gap. Rebuilding a number from a figure you cannot read is how pages end up wrong.

Prompts also appear on the employer's side of the table. Korn Ferry runs four free tests, with one warning attached, and a guided case is poor rehearsal for a timed one.

When to stop guided case practice

Guided mode teaches the shape of a case, and the mistake is staying in it after it has finished teaching. Two signals say you are done. You start predicting the prompt before it appears, and your structures stop changing between cases. At that point the scaffolding is carrying you.

A third signal is quieter and more reliable. You stop being surprised by the strong version at the end. While that comparison still stings, the mode is doing work. Once your own structure and the model one differ only in wording, the remaining gap is delivery under pressure, and delivery is trained somewhere else.

The order most candidates need is guided cases to learn the shape, drills to fix one dimension at a time, assessment simulators to rehearse a screen under its real clock, and spoken practice to remove the scaffolding. Our six simulators run free and without an account, and each one keeps the timing and the forward-only progression of the format it copies.

What is open right now and what needs a plan is worth being blunt about. The drills, the calculators, the builders and the six simulators need no account at all. Running a case from the library goes through checkout, and every plan includes the same core: unlimited cases, uncapped drills, learning mode, the behavioral simulator and the spoken mode. Bain calls a case interview "a structured conversation where you work through a real business problem", and adds that it is "not looking for a single right answer". A guided case is a rehearsal of that conversation with the pauses written in. What it is scored on is set out at how an answer is assessed, and the weights there are read from the same code that grades your attempt.