Types of Case Interviews: The Only Firm-Named Taxonomy

Three types of case interviews one firm names, with a frequency claim

The definitions come from Accenture's case workbook at accenture.com, and they classify by how much information you are given rather than by industry.

The firm then ranks them: "The "Great Unknown" and "Back of the Envelope" cases are the most common types of cases." And it explains why the third is rare: "These cases are relatively rare because of the amount of time necessary to present the case description".

Its own practice set confirms the ranking. The workbook lists 21 cases, of which 13 are Great Unknown and 8 are Back of the Envelope. Checked by addition: 13 plus 8 is 21. That is 13 over 21, which is 61.9 percent, against 8 over 21, which is 38.1 percent. Zero of its 21 practice cases are Parade of Facts, which matches the firm's own claim about rarity.

Recognizing which one you have takes one question at the start. How much did the prompt tell you?

1. Two or three sentences with no numbers. Great Unknown. Your first move is a structure and a request for data.
2. A paragraph with six numbers, two of which look irrelevant. Parade of Facts. Your first move is sorting each figure into a branch.
3. One question with a number attached to it. Back of the Envelope. Your first move is an estimation chain.

The document carries a 2018 copyright and reports 435,000 people against 791,000 in the firm's September 2026 postings. It is dated, and we say so rather than presenting it as current.

The 43 published case interviews, cyber and valuation included

The firm-named taxonomy sorts by information. Candidates usually want the sort by subject, so here is the census of what the firms actually published, counted from their own pages.

Twenty-one plus ten plus six plus four plus two is 43, verified by addition on 30 September 2026. Kearney announces one further case on promotional planning whose content is absent from the archived snapshot, so 44 are announced and 43 readable.

Read that census against the usual five-category list and two things stand out.

Market entry and launch appear, but not as a majority. McKinsey's electrolyte launch and Oliver Wyman's water park and oil and gas cases sit there. Bain's coffee shop is an opening decision. That is four of 43.

And a whole family appears that no standard taxonomy contains. Deloitte's ransomware case, its data exfiltration case, its capital projects case and its warehouse valuation case belong to Cyber, Forensic and Valuation practices. A candidate preparing only profitability and market entry has nothing to bring to any of them.

The only difficulty labels anyone publishes

Oliver Wyman is alone in telling you how hard its cases are. Its preparation page at oliverwyman.com links six cases, and each carries a label.

Urban Grocers, at oliverwyman.com, is labeled easy, and it asks whether to change the truck fleet used to move product. The other five are labeled medium: Supermarket Pharmacy, Dairy Farm, Autism Digital Device, Oil Gas Price Strategy, Poseidon Water Park.

So one of six is easy, which is 16.7 percent, and five of six are medium, which is 83.3 percent. None is labeled hard.

That distribution is worth taking seriously as information about calibration. A firm publishing practice material has no reason to withhold its hardest example, and it stopped at medium. The likely reading, and we label it as ours, is that medium is the working level of a real first-round case rather than a warm-up.

Deloitte publishes a completion estimate instead of a difficulty label. Its practice tool gives "Estimated completion time: 15-20 minutes" per case, and states: "This is not an assessment and we do not track your responses or results." A 15 to 20 minute written exercise is shorter than the 30 to 45 minute spoken case portion Accenture publishes, so the tool is a component drill rather than a full rehearsal.

Deloitte sorts case interview types by practice rather than problem

The ten cases at usrecruiting.deloitte.com are filed by the practice that would run them, and the filing is the most useful signal on the page.

Three sit under Strategy: a federal health agency finance strategy, a market expansion for a bus manufacturer, and talent management for a federal bureau. One sits under Enterprise Tech Strategy. One sits under Functional and Operations Strategy. One sits under Cyber. One under Forensic. Two under Valuation. One under Forensic Analytics.

The page also separates undergraduate from advanced degree students, and separates consulting practice cases from risk and financial advisory scenarios.

Three titles appear in the filter menus with no openable page behind them, and the corresponding URLs return 404, so we count ten rather than thirteen.

What this filing tells a candidate is which vocabulary to bring. A Valuation case wants discount rates and multiples. A Cyber case wants incident response, exposure and continuity. A Strategy case wants the profit equation. Preparing one shape for all three is how a well-prepared candidate still has nothing to say.

Interviewer and candidate led, case interview types nobody publishes

The most repeated split in this subject divides cases into interviewer led and candidate led. Neither phrase appears on any firm page in our reading, and we read seven of them. What the firms publish instead is a numbered sequence, and a sequence tells you who is expected to move next.

Four firms, four sequences, and all four hand the candidate the structuring move. None of the pages describes an interview where questions arrive one at a time and the candidate answers them in the order given, which is what the interviewer led label is supposed to name.

So the practical reading is this. The labels describe a real difference that candidates experience, and no firm has written it down, which means you cannot prepare against the label. You can prepare against the sequence, and the sequences agree on the first three moves: understand the problem, ask what you need, structure before computing. Those three transfer to either format.

One behavior does change with the format, and it is worth a sentence. In a case where the interviewer is handing you questions, the structuring move still has to happen out loud, early, and in under ninety seconds, because nobody is going to wait while you build it silently.

The problem shapes underneath, and the equation each needs

Sort the 43 cases by the equation they require and four shapes cover them. This grouping is ours rather than any firm's.

1. Profit movement. Profit equals revenue minus cost. Open revenue into volume times price and cost into fixed plus variable. Bain's fashion case at bain.com is one, running five years of revenue decline against two costed options, and its published answer prefers the intermittent discount over the loyalty programme.
2. Threshold or entry. Breakeven volume equals fixed cost over unit contribution, then compared to a bar. The bar is a convention rather than a fact, and the firms publishing entry cases use different ones.
3. Sizing. A chain of assumptions from a population down to a unit count, then a sanity check. This is the Back of the Envelope type.
4. Investment or valuation. Payback equals capital over annual cash, or value equals cash flow over a discount rate. Deloitte's two Valuation cases live here.

Recognizing the shape does more for you than recognizing the industry, because the equation is what you will actually write down. A dairy farm and a supermarket pharmacy are different industries and the same shape.

One caution about the sizing shape. McKinsey's electrolyte case asks a launch question and hands you a threshold problem, and Oliver Wyman's water park case asks for daily revenue, which is a sizing problem. The prompt's subject does not reliably predict the shape.

The Back of the Envelope case interview type, worked once

The type Accenture calls most common alongside the Great Unknown deserves one full worked example, with stated assumptions.

Question. How many coffee shops does a city of 1.8 million people support?

Chain.

• Buyers: 45 percent of 1.8 million buy coffee out, so 810,000 people.
• Frequency: 3 cups a week each, so 810,000 times 3 is 2,430,000 cups a week.
• Annual: 2,430,000 times 52 is 126,360,000 cups.
• Shop capacity: 600 cups a day over 360 operating days is 216,000 cups a year.
• Shops: 126,360,000 over 216,000 is 585 shops.

Recomputed by hand on 30 September 2026. 1.8 million times 0.45 is 810,000. 810,000 times 3 is 2,430,000. Times 52 is 126,360,000. 600 times 360 is 216,000. And 216,000 times 585 is exactly 126,360,000, so the division is clean.

Sanity check, which is the part that earns the points. 585 shops for 1.8 million residents is 1,800,000 over 585, which is one shop per 3,077 people. That is a number you can argue about, and arguing about it is the exercise. State it and say which assumption you would test first.

The assumption to test first here is frequency. Three cups a week is the input with the widest plausible range, and doubling it to six would give 1,170 shops. Naming your most fragile assumption before you are asked is the move that separates this from a guess.

Which type your rubric exposure changes

Our weights, read from the grading code in src/lib/grading, differ by what the case demands. No consulting firm publishes a marking scheme.

Choosing which types of case to practice

Five decisions, using the census above.

1. Practice the two common types first. Great Unknown and Back of the Envelope, on Accenture's own account, and its own 21 cases contain nothing else.
2. Practice the four equation shapes rather than five industries. Profit movement, threshold, sizing, investment.
3. Add the practice-specific vocabulary for the firm you are applying to. Deloitte's filing tells you whether you need valuation or cyber language.
4. Use the labeled cases to calibrate. One easy and five medium is the only published difficulty scale available.
5. Do at least one Parade of Facts even though it is rare. Sorting is a separate skill from digging, and you will meet it inside larger cases.

Our structure drill handles the Great Unknown shape, the market sizing drill handles the Back of the Envelope, and the chart drill handles the sorting problem. All three are free and none needs an account. None of them will tell you which shape a prompt is, which is the judgment the first thirty seconds of a real case demands.

For the exercise in general, see what is a case interview. For the cases themselves with their numbers, see case interview examples. For the questions that come with each, see case interview questions.