Reading Case Exhibits: The So What Before the Numbers

The criteria one firm prints under a case interview exhibit

The Dairy Farm case sits at oliverwyman.com, labeled a practice example at medium difficulty. A private equity client is choosing whether to invest in one of two dairy farms in an Illinois county.

The firm prints evaluation criteria beneath each question, and the lists change as the case moves. At the opening it wants the key question identified, the relevant information selected, a clear structure, and that structure communicated succinctly. On the following question, where data arrives, it wants three different things: several factors produced, adaptation as the data comes in, and the data arranged so that computation is fast.

Take the third of those literally, because it is the one nobody teaches. Laying the data out is a physical act on paper before it is an analytical one. A table copied in the order the exhibit presents it is a table you will compute slowly. A table rewritten with the two columns you need next to each other is one you will compute quickly.

Bain's case page supports the same habit from the practical side. Its in-person list asks you to bring pen and paper, write legibly, and be ready to show your notes. A page the interviewer may read is a page worth arranging.

The firm's published model answer for this case has three branches: farm profitability, local dairy market dynamics, and fit with the client's portfolio. The third is the one candidates skip, and it is the one that only exists because the buyer is a fund.

Four moves to read a case interview exhibit, on a channel table

Here is an exhibit with stated figures, and a read that takes about ninety seconds.

Every figure recomputed by hand on 30 September 2026. 260 times 0.34 is 88.4. 180 times 0.19 is 34.2. 60 times 0.46 is 27.6. Revenue sums to 500 and margin to 150.2. Blended margin is 150.2 over 500, which is 30.04 percent, so 30.0 percent as shown.

Move one: say what the exhibit is, in one sentence. Revenue and gross margin by channel, one year, three channels totaling 500 million dollars.

Move two: find the row that is out of line. Online carries a 46 percent margin against a 30 percent blended rate. Wholesale carries 19 percent. The spread between the best and worst channel is 27 points.

Move three: compute the two shares and compare them. Online is 60 over 500, so 12 percent of revenue, and 27.6 over 150.2, so 18.4 percent of gross margin, rounded from 18.38. Wholesale is 180 over 500, so 36 percent of revenue, and 34.2 over 150.2, so 22.8 percent of margin, rounded from 22.77.

Move four: say the so what and name the next test. The channel mix is working against the margin: wholesale supplies 36 percent of revenue and only 22.8 percent of the margin, while online does the reverse. The next thing I would want is whether the 19 percent wholesale rate comes from price concession or from a different cost to serve, because those have different fixes.

The comparison in move three is the engine of the whole read. A share of revenue against a share of profit tells you which part of the business is carrying the other, and it is available from any table that has a rate column.

The detail in the prose that is not in the chart

Some of the exhibit is not in the exhibit. One firm publishes both the rule and a worked example.

Accenture's careers blog at accenture.com states: "If the interviewer says something, it probably means something. Don't dismiss seemingly extraneous details." Its own illustration: a retailer has been acquired, and the previous owner was fond of the children's brand. That detail tells you that selling the children's business is off the table, which removes a branch from your structure without appearing in any table.

The same page adds why the remark is there at all: "The interviewer is your biggest asset in the room. They have the information you need to "solve the case" successfully."

Three kinds of out-of-chart detail are worth listening for.

1. A constraint. An owner who will not sell, a regulator who will not approve, a plant that cannot be closed this year. Each one deletes a branch.
2. A basis. Figures are annual, or quarterly, or per store. An exhibit whose basis is stated aloud rather than labeled is the most common source of unit errors.
3. A time window. Three years of data spoken as last three years before the acquisition is a different exhibit than three years to date.

Deloitte's practice tool makes the same point structurally. Among its three usage steps is "Access context at any time", with the note that the business situation stays available in the corner of the screen. The context sits beside the question because the question is not self-contained.

A published case interview exhibit whose answer lands on its cost

One published case question resolves so neatly that it is worth working in full.

Deloitte's Recreation Unlimited case at usrecruiting.deloitte.com asks: "If Recreation Unlimited chooses to improve their website it will cost them $10M (one-time cost), but they expect that each visitor will spend an additional $1. The profit margin on any products purchased on the website is 50%. How would you evaluate the potential opportunity of improving the website? The client has also provided us with the data point they have 20M people visit the site annually."

The firm also separates two calculations in its own notes: "Calculation A: Contribution is the case influx generated by the additional visitors and spending. It does not account for the profit margin on products sold. Calculation B: Includes profit margin and is the actual profit generated."

The answer table on that page is an image, so the arithmetic below is ours, derived entirely from the published figures and recomputed by hand on 30 September 2026.

• Calculation A: 20 million visitors times 1 dollar is 20 million dollars of incremental revenue.
• Calculation B: 20 million dollars times 50 percent is 10 million dollars of incremental profit.
• Against a one-time cost of 10 million dollars, the project returns its cost in one year exactly.

That result is the reason this question is well designed. The number does not decide anything on its own, so the candidate has to say what would. A one-year payback on a one-time cost is attractive if the dollar of extra spend persists, and worthless if it is a launch effect. Naming that is the so what, and the number alone is not.

Watch the trap in the middle. A candidate who stops at calculation A reports 20 million dollars against a 10 million dollar cost and concludes the project doubles its money. The firm separates the two calculations because that error is the one it expects.

The parade of facts, and the firm that named it

Accenture's case workbook names three case types, and one of them is built from surplus detail: "The Great Unknown", "The Parade of Facts", and "The Back of the Envelope".

The Parade of Facts is defined as a case with a great deal of detail, some of it not useful, which tests whether the candidate can sort. The workbook also states that these cases "are relatively rare because of the amount of time necessary to present the case description", and that "The "Great Unknown" and "Back of the Envelope" cases are the most common types of cases."

So the exhibit-heavy case exists, has a name given by a firm, and is the least common of the three by that firm's own account. The document carries a 2018 copyright and announces a headcount of 435,000 against 791,000 in the firm's September 2026 postings, so it is dated and we say so.

The sorting discipline it calls for has a simple test. For each figure in the exhibit, ask which branch of your structure it belongs to. A figure that belongs to no branch is either a distractor or a sign your structure is missing a branch, and deciding which takes four seconds.

The workbook also publishes a worked case of exactly this type, on fresh prepared meals, where the published method is unit margin, then fixed cost, then a division, then market share at breakeven. That sequence is worth memorizing as a shape, because it converts a pile of facts into three operations.

So what carries 30 of 100

These weights belong to us. They are read from the grading code in src/lib/grading, and no consulting firm publishes a marking scheme of its own anywhere in our corpus.

Drilling case interview exhibit reads

Exhibits are the most drillable part of a case, because a single table is a complete exercise.

1. Set ninety seconds per exhibit and run the four moves. What it is, the row out of line, the two shares compared, the so what with a next test.
2. Rewrite the table before you compute. Put the two columns you need side by side. Oliver Wyman's own criterion asks for exactly this.
3. Always compute a share of revenue against a share of profit. It is one division per row and it produces the insight more often than anything else.
4. Make yourself say the basis out loud. Annual, per store, per visitor. Then check the exhibit agrees.
5. Practice the two-calculation trap. Take any revenue uplift and compute both the revenue and the margin version. Deloitte's published question separates them because candidates do not.
6. Sort every figure into a branch. Anything left over is a distractor or a missing branch.

Our chart drill scores So what, Observation, Implication and Next test as separate lines and names the first correction. Our math drill times the arithmetic an exhibit provokes. Both are free and neither needs an account. Neither one can tell you which of two defensible so whats a partner would have wanted, which is the judgment being hired for.

For the arithmetic habits behind the computation, see case interview math tips. For the hypothesis the exhibit is meant to settle, see hypothesis driven cases. For the firm's six published cases, see the Oliver Wyman case interview guide.