Oliver Wyman Case Interview 2026: What the Firm Marks You On

Three firms, three framings of the same exercise

The wording a firm chooses tells you what posture it expects. The two comparison rows below come from our own verified pages for those firms, linked so you can check them, and the reading in the third column is ours rather than anyone's policy.

Read the middle column first and the right one second, because the right one is interpretation and the middle one is the record. Each row still points at a different way to lose the interview. The McKinsey failure is walking your own branches uninvited when nobody asked you to. The Bain failure is following the conversation down one branch and never returning to the other three. The Oliver Wyman failure is the one this page is about: performing self-sufficiency in front of someone who has told you, in writing, that they will help.

What the client framing changes in an Oliver Wyman case interview

If the interviewer is a client, three things that feel risky become correct.

Asking for data is not weakness. A consultant in front of a client asks for the numbers they need. They do not guess in order to look independent. The firm says to ask clarifying questions and to summarize the situation in your own words, which is exactly what a first client meeting sounds like.

Being stuck out loud is better than being stuck silently. Silence reads as lost. Naming the blockage reads as diagnosis: "I can see two ways to split this and I'm not sure which is more useful here. What I'd want to know is whether the decline is concentrated in one region." A client answers that question. So does an interviewer.

Thinking on paper is expected. The firm says to put your initial thoughts down, diagram your ideas, and structure your thinking. That means the notes are part of the work rather than something to hide. Turn the page around. Point at the branch you are in.

The posture has a floor. A client who ends up doing the structuring themselves does not rehire you, and an interviewer who ends up doing it does not pass you. What the framing licenses is a professional asking well-formed questions and then using the answers in front of the person who gave them.

The format, the test, and the year a fit failure costs

The regional page for entry-level hiring in Brazil and Mexico is the most detailed thing the firm publishes about its process, and it describes four stages rather than two.

The test is described, and the description is easy to walk past. The firm names its four question types: critical reasoning, problem solving, data sufficiency, and integrated reasoning. Those four names belong to the GMAT family of questions rather than to the puzzle-and-game format some firms use. They are question types. We have not read an official page listing what a current GMAT-family test is made of, so we will not tell you how that test is built today. What survives the uncertainty is practical: data sufficiency punishes a candidate who meets it for the first time on the day, because the task is to judge whether a question can be answered at all, and working the rules out live costs minutes you do not have.

Three more lines from the same page change how you should treat the process. Every stage is evaluative and eliminatory, so there is no round you can coast. The entire process runs virtually, assessments through final interviews. And if your application fails at the fit interview stage, you must wait a year before reapplying, which makes the fifteen-minute conversation the most expensive quarter hour in the process.

Treat the counts as regional, because the page is. Your invitation names your own format and it is the authority. Plan for the stamina anyway: four cases across two rounds is closer to a final-round profile than to a single-case process, and the fourth case is where structure degrades.

The six free cases, and the criteria printed inside

Oliver Wyman publishes six complete cases on its own careers site, under their own names, each with a difficulty label and a full sample answer. This is the fullest free set we have found published by a firm itself. It is free, it is complete, and we have no way to measure how many candidates work through it before paying for cases elsewhere.

Start with Urban Grocers, because it is the only one the firm labels easy, and a fleet decision is clean unit economics. Note the spread after that: an investment screen, a retail add-on, a growth question, a product decision and a sizing estimate. Five different arithmetic shapes. Working all six tells you which shape you are slowest at, which is a diagnosis six profitability cases cannot give you.

The cases print what is being marked, question by question. This is the part to exploit. Under each question the firm lists evaluation criteria in its own words: on the dairy farm opening, it names identifying the key question, structuring clearly, and communicating that structure succinctly. On the next question it names generating multiple factors, adapting as data arrives, and setting data up so you can calculate quickly. That is a marking scheme, published, per question. Mark your own answer against it before reading the sample.

Work the prompt with the solution still closed. Take the opening, build a structure under a clock, say it aloud, run the arithmetic, and commit to a recommendation. Then read the write-up looking for what you missed rather than for whether you agreed.

A pricing case where one cent needs a fifth more volume

Five moves, with a number carried through.

1. Play back the objective. "A regional fuel retailer wants to know whether to change its pump pricing, and success is higher total profit rather than higher volume. Over what horizon?"
2. Ask two or three things that change the structure. What does the competitor across the road charge, and how quickly do they match? What share of profit comes from fuel against the shop? Those two answers reshape everything.
3. Structure from the mechanism. "Pricing turns on willingness to pay, elasticity and competitor response. Here the product is close to a commodity, so elasticity is high and response is fast. I'd look at whether fuel is the profit or the footfall driver before touching price."
4. Run the number aloud. Say 40 sites selling 4 million gallons a year each is 160 million gallons. At 6 cents of margin per gallon that is $9.6m. Cut a cent and the margin per gallon falls to 5 cents, so holding $9.6m needs 192 million gallons: a 20% volume rise. Say the trap out loud as well, because it is the one that catches people here. The margin you gave away is 1.6 of 9.6, about a sixth, and it is tempting to call that the volume you need back. It is not: you have to recover the lost margin on a thinner unit, which is why the answer is a fifth rather than a sixth.
5. Recommend, first sentence. "Hold pump price and move the shop range instead, because a 1 cent cut needs a 20% volume rise to break even and the competitor matches within a day. Evidence: 6 cents on 160 million gallons is $9.6m, and at 5 cents you need 192 million gallons to stand still. The risk is that footfall depends on being price-competitive at the pump, so I'd check shop spend per fuel customer before deciding."

Notice where the client framing shows. Move two asks for competitor data rather than assuming it. Move five names the thing that would change the answer, which is what you would tell a client rather than what you would hide from an examiner.

The entry title, and the trap of reading a word across firms

The regional page speaks of consultant positions and says they require availability to travel. That is the whole of what the pages we read publish about the title, and it is less than most guides imply.

The rest of this section is our reading, and we mark it as ours. The same job title sits at different heights at different firms, and we have not surveyed that mapping, so we will not tell you what a given word is worth as a rung. The posting is the evidence: the degree it expects, the start date it names, the band it quotes. The sound of the title is decoration.

What is assessed is what any entry-level consulting track assesses: a structure that holds under time, arithmetic run without being asked, and a recommendation defended under challenge. Whatever the title promises about year three, the interview bar is set by those three things.

We have also not measured Oliver Wyman's sector mix, and nothing on the pages we read describes one, so you will not find a claim about it here. If a particular practice is your reason for applying, say so in your own words and attach something you have already done to it. A real sector reason reads as a reason at any firm. One assembled for the application reads as a guess at any firm.

One family of cases is worth naming separately, because its arithmetic is unlike the rest. Asset-based businesses earn a percentage of what they hold, spread businesses earn the gap between two rates, transaction businesses earn per event. Establishing which of the three you are in is the first analytical move, and it comes before any number. Financial services cases covers each, and pricing cases covers the mechanism this page's worked example turns on.

The six things the Oliver Wyman case interview assesses

The interview preparation page gives two lists of six. The first is what the case is for, written as six questions the firm asks itself about you, each with its own reason attached.

Two of those six are about asking and explaining rather than analyzing, which is the same message as the client framing arriving by another route. The fourth is the one candidates most often fail without noticing: a structurally correct answer that no operator could act on is a theoretical construct, and the firm has said in advance that it does not want one.

The six-step method, and the structure you rejected

The second list of six is the method it recommends: pinpoint the issue, break it down, pick an approach, analyze the issues, use numbers, make a recommendation. Under "pick an approach" it adds an instruction almost nobody follows, which is to show why your approach is better than the alternatives. Naming the structure you rejected takes one sentence and it is the difference between a framework and a choice.

Below are the dimensions our own grader measures, with the weight each carries out of a hundred.

Where Oliver Wyman candidates lose it

1. Performing self-sufficiency. The firm says to work with the interviewer. Guessing at data you could have asked for is the specific failure this format invites.
2. Silent stalling. Being stuck is survivable. Being stuck quietly is not, because it reads as lost rather than as thinking.
3. Treating the fit interview as the warm-up. Fifteen minutes with no case, and a three-minute story eats a fifth of it. Of every stage, this is the one where a rambling answer costs the most, and it is the stage candidates rehearse least.
4. Fading by the fourth case. Four across two rounds is a long day, and a rehearsal schedule built one case at a time never puts a fourth structure under fatigue.
5. Ignoring the six free cases while paying for cases elsewhere, and ignoring the marking criteria printed inside them.
6. Meeting data sufficiency for the first time on the test. It is one of the four named question types, and it punishes an unfamiliar candidate hardest, because the task is to judge whether you can answer rather than to answer.
7. A recommendation that does not answer the question asked. Answer first plus Evidence is 60 of the synthesis mark, and the firm names this instruction explicitly.

The first three are specific to how this firm frames and runs its interviews, and they are avoidable once you have read the firm's own pages properly. The rest recur everywhere, which is why a drill diagnosis is worth more than firm-specific study once the format is known.

One pattern sits underneath several of them. A candidate who treats the case as an examination performs independence, hoards uncertainty, and delivers a summary. A candidate who treats it as client work asks, flags, and answers the question. The firm has told you which of the two it wants.

A preparation order that fits the Oliver Wyman case interview

The constraint here is stamina as much as technique, because four cases is a long day. Sequence accordingly.

1. Diagnose, in twenty minutes. One short drill of each type: math, structuring, brainstorming, exhibits, synthesis, sizing. Read the dimension each rubric names as the first correction. That list is the plan.
2. Work the losing dimension until it changes. The structuring drill and the math drill are open without an account.
3. Work all six published cases, starting with the easy one, and mark each answer against the criteria the firm prints under that question before reading its sample.
4. Rehearse asking. Take a case and force yourself to ask three questions you would normally assume the answer to. It feels wrong and it is the posture the firm describes.
5. Time your fit stories at ninety seconds. Fifteen minutes, no case, and a year's wait before you may reapply if it goes wrong. Over-prepare this stage.
6. Sit one timed run of GMAT-family questions, data sufficiency included, so the question types are familiar on the day.
7. Run four cases in one sitting, once. The point of the sitting is the discovery that your fourth structure is worse than your first, and the only way to find that out is to build four.

The limits of what is published here

Whether the online test applies outside Brazil and Mexico. The page that describes it is regional. The test exists and is documented there. Whether your process includes it is in your invitation.

What a current GMAT-family test is made of. The firm names four question types and stops. We have not read an official page listing the sections of that test today, so we describe the four as question types and leave the rest alone.

Whether four interviews is the format elsewhere. One regional page is one region, and we will not generalize a count from it.

Any threshold or scoring weight the firm uses. Nothing published on the test or the cases, and we will not invent one. The weights in the rubric above are ours.

What the fit interview asks. The regional page gives its length, its language and the fact that it carries no case. The content is not described, so prepare it as you would any fit conversation: evidence of judgment you have already exercised, timed. This is the gap in the published record that costs the most.

Whether the six published cases reflect what you will be given. They come from the firm, which makes them the closest thing available, and they carry its own difficulty labels and marking criteria. They are also teaching material, and teaching material is tidier than the real thing.

What the Marsh rebrand changes for candidates. The careers pages now carry a notice that Oliver Wyman is part of the expanded Marsh brand. Nothing on the pages we read says whether that touches the hiring process, the entity you would be employed by, or the application route. If it matters to you, ask the recruiter rather than reading a guide.

Three kinds of statement sit on this page, and they are meant to stay distinguishable. Quotations and close paraphrases of an Oliver Wyman page, each linked. Absences we name as absences, gathered above. And our own judgments, which are marked as ours where they appear, including every weight in the rubric. If your own experience of this process differs, write to us with what you saw and when, and we will read it back against the firm's pages and date whatever we change.

The case is one part of a process the firm describes elsewhere in minutes and months: fifteen minutes for the first screen, one year before you may reapply.

What changed on this page, and when

We correct this page in the open and date every change, so a reader can tell what they are reading. Corrections live here rather than in the body.

29 September 2026. Three corrections after an external review of the page.

An earlier version told readers we had found no published online assessment at this firm. That was wrong, and the description was sitting on the regional page we were already citing: 26 questions, 50 minutes, four named question types. Telling a reader that something is undocumented when it is documented on our own cited source is worse than leaving it out.

An earlier version said the interview preparation page gives four tips. It gives two lists of six, and it also publishes six evaluation questions that the earlier version ignored. Those questions were the strongest material on the page and they now carry a table of their own.

An earlier version listed a sixth case named Wumbleworld. No case by that name is published by Oliver Wyman. The sixth case is Urban Grocers, and all six now link to the firm's own pages.

Before that. The worked pricing case once stated the wrong break-even volume for a one cent cut. The figure is a 20% rise, and the step that produces it is now written out in full so you can check it against your own arithmetic.