Case Interviews in a Second Language 2026: Nine Offices

Nine offices, nine published requirements, one firm

The densest document on this subject in our whole corpus is a recruiting brochure. OC&C's 2024 global graduate brochure runs fifteen pages and its own navigation names nine office sections. Every one of the nine carries a line headed "Language skills".

We counted the column by hand. Three offices ask for English alone. Five put a second language at the same rank as English. One asks for English and calls the local language preferable. Three plus five plus one is nine, which matches the nine sections the brochure's own navigation lists.

That distribution is worth a moment. A candidate who reads "consulting is conducted in English" has read something true about London, Sydney and New York and false about the other six. Within one firm, on one document, in one year.

The single office that names a level

Now the finding that the table hides. Of those nine requirements, exactly one attaches a measurable standard: "business fluency in German (at least at C1 level) & English", for Munich and Düsseldorf. The other eight write "business fluency" and stop.

C1 is a band of the Common European Framework. It is testable, it appears on certificates, and a candidate can find out in an afternoon whether they hold it. "Business fluency" has no scale, no test and no threshold, and the same brochure uses the phrase eight times without defining it once.

This matters more than it looks. One office in nine gives you something you can check against yourself. In the other eight you are guessing at a bar the firm has not drawn, which means the only honest move is to ask. A recruiter can tell you in one sentence what their own page will not.

The same brochure shows what a firm looks like when it does want to be precise. London's qualifications line names a tool by name: "Experience with Python and/or R for the Associate Consultant - Analytics positions". The firm is perfectly capable of writing a testable requirement. On language, in eight of nine offices, it chose not to.

Two firms name where the local language is mandatory

Two firms publish the rule as a rule rather than as an office note, and they draw it differently.

ZS answers the question on its own interview process page. The question is printed as "Do I need to speak the local language to be considered for an office?" and the answer is "All ZSers must speak English, though a second language is preferred in certain offices. ZSers in Germany and Japan are required to speak the local language in addition to English." Two countries, named. Everywhere else the second language is preferred and the page does not say which offices those are.

Roland Berger draws it wider and softer at once: "A natural communicator, your spoken and written English is excellent. You also have a good command of the language local to the office you are applying to." No country list, no exceptions, no level. Apply to Munich and the local language is German. Apply to Paris and it is French. The rule follows the office you pick.

Put the two side by side and the shape of the problem appears. One firm names two countries and calls the rest a preference. The other requires the local language everywhere and defines it nowhere. Neither rule can be generalized into an industry norm, and a page that tried would be wrong about at least one of them.

McKinsey carries the requirement somewhere else again: inside the job posting. Its Business Analyst posting asks for "Strong communication skills, both verbal and written, in English and local office language(s), with the ability to adjust your style to suit different perspectives and seniority levels." The plural in "language(s)" is the firm's own. A candidate reading the students pages would never find it.

What language the case interview runs in, and who says

Here is the count that reframes the whole subject. Ten firms and issuers in our reading publish something about language: OC&C, Roland Berger, ZS, Oliver Wyman, IQVIA, IBM, Capgemini, McKinsey, Gupta Strategists and Deloitte in Spain. Two of the ten say what language an interview is conducted in.

Oliver Wyman's Latin America page is the clearest and the strangest. It publishes a full process, and one line reads "A brief, 15-minute conversation conducted in English". That is a region whose working languages are Spanish and Portuguese, and the firm names English for one stage. The same page describes four case interviews and says nothing at all about what language those run in.

IQVIA is the other. Its interview FAQ says "Most interviews are conducted in English unless otherwise specified. However, some questions may be asked in local language." Both "most" and "some" are the firm's own words, and neither is a number.

Then look at what a detailed FAQ leaves out. Analysis Group publishes that its employees "come from more than 40 countries, speak more than 40 languages, and enjoy transfer opportunities among our 15 global offices", and on the same page it tells you the first round is behavior based and conducted virtually, and that doctoral candidates present their research. Composition, format, stage, all published. Language of assessment, absent.

McKinsey publishes a language rule for the file rather than the room: attachments for Austria and Germany may be in English or German, and "Please submit only one version in your application." Its Solve assessment is offered in four languages. Neither statement covers an interview.

Two firms score your English, and neither publishes the bar

Two issuers do more than require a language. They test it, and they announce the test.

IBM describes a screening assessment in plain terms: "Respond to multiple-choice questions in a text-based format to help us understand your level of English proficiency." Its sister assessment FAQ lists the families it draws from and includes "English language assessments" among them. So a candidate can be filtered on English before any human reads the file.

Capgemini publishes a conditional version. At the assessment stage a candidate "may be required to complete proctored online technical and language assessments, where applicable." Which language, which level and when it applies are all left open.

Neither publishes a threshold, a scoring method, a duration or a test publisher. The pattern is familiar from every other part of consulting recruiting: the stage is announced and the bar is withheld.

That is the whole published record. One firm scoring English in multiple choice, one firm reserving the right to, two firms naming an interview language, six firms naming a hiring requirement without naming an assessment. Anything beyond that, on any firm, was estimated by somebody.

What the case interview format demands, whatever the language

Since eight of the ten publish nothing about the room, the honest way to prepare is to reason from what the exercise mechanically demands. That reasoning is ours, and it is labeled as ours.

A case interview asks you to do four things out loud at once: hold a structure in memory, speak it in order, run arithmetic while speaking, and listen for a correction. In a second language, three of those four compete for the same working memory that the arithmetic needs. The fourth, listening, gets harder when the interviewer speaks quickly.

The consequence is specific rather than general. It is arithmetic aloud that degrades first, because reciting a multiplication in a second language costs attention that the multiplication itself was using. Structure degrades second, because a long sentence in a second language is where speakers lose their own thread. Vocabulary breadth degrades not at all, because a case needs perhaps two hundred words and you have them.

So the preparation that pays is narrow. Numbers spoken aloud in the interview language, to the point of boredom. Short declarative sentences, which read as confidence in any language. Explicit signposting, because saying "first revenue, then cost" tells the interviewer where you are and tells you too. None of that is a concession to a weaker speaker. It is what a strong answer sounds like when a native speaker gives it.

Five steps for the weeks before a case interview

In order, and the first one is the only one that removes uncertainty rather than managing it.

1. Ask the recruiter what language each stage runs in. Eight of the ten firms above do not publish it, and the person scheduling you knows. Oliver Wyman shows the question is answerable: it published the answer for one stage of one region.
2. Ask what the language requirement means for your office. One of nine OC&C offices names a level. If yours does not, the standard exists in someone's head and a recruiter can describe it.
3. Drill numbers aloud in the interview language. Do it out loud rather than on paper. Percentages, division by a two-digit number, scaling by thousands and millions, and round-and-say. Silent arithmetic is a different skill from the one being tested.
4. Cut your sentence length deliberately. Aim under twenty words. Long sentences are where the thread breaks, and short ones are where a second-language speaker sounds most in control.
5. Script your three signposts. One to open a structure, one to move between branches, one to hand back a number. Three phrases, rehearsed, remove most of the load that a second language adds.

If a firm publishes a language assessment, as IBM does, add a sixth step and take a free multiple-choice English practice test before the screening rather than after it.

The dimensions our own scoring applies

Nothing above is scored by a firm, because no firm publishes a rubric. The weights below are ours, set by us and read from our own code.

Read down the right-hand column and the reassuring part is arithmetic. Seventy of the hundred points on our structuring rubric sit on decisions you make before you open your mouth. The ten points most exposed to language are the smallest block on the scale.

Our math rubric weights Answer 40, Working 25, Units 15 and Interpretation 20, and Working is the one that suffers when you narrate a calculation in a second language, because that is the part you say out loud. The structure drill and the math drill both run in the browser without an account, and how an answer is assessed sets out each weight and why it carries what it carries. Neither drill teaches vocabulary, and nothing here reproduces any firm's interview.

What none of the ten publishes

Five gaps, and they are the same five everywhere we read.

Nobody defines "business fluency". Eight of nine OC&C offices use the phrase, and the ninth is the only place in the document where a level appears. No firm read here points to a framework, a certificate or a test.

Nobody publishes an accommodation. No extra time. No right to have a question repeated. No option to take the exercise in another language. The only language accommodation published anywhere in our corpus is that McKinsey's Solve assessment exists in four languages, and that is a test rather than an interview.

Nobody says whether language is scored. No firm publishes a rubric of any kind, which is a constant of this industry rather than a fact about language. Our own weights are ours.

Nobody publishes a threshold for the two assessments that do measure English. IBM and Capgemini announce the stage and withhold the bar.

And four firms we wanted to read refused our connections. bain.com, kearney.com, horvath-partners.com and alvarezandmarsal.com returned 403 to a browser user agent and to the text proxy, and the proxy itself returned 403 on every URL tested on 1 October 2026. That is a refusal of access rather than an absence of information, so no language rule is attributed to any of the four, in either direction.

For the requirement written into a European application file, our page on the Dutch consulting market carries the office-level rules, and consulting in Germany carries the German-language ones. For the exercise itself, independent of language, the consulting interview process holds the published stage lists.

What changed on this language page, and when

1 October 2026. This replaces a 236-word page that asserted what interviewers assess about a candidate's English and cited nobody for it.

That claim is gone from the page. No firm in our reading publishes what it assesses about a candidate's language, and the old version was stating a courtesy nobody has promised in writing. The published record turns out to be three different things: a hiring requirement, sometimes a scored assessment, and almost never an interview language. Collapsing the three is what produced the old sentence.

Two other claims were removed with it. The old page said that arithmetic aloud is the highest-return thing to drill, which is our reasoning and now appears labeled as ours rather than as a finding. And it said that vocabulary breadth is not scored, which nobody can know, because no firm publishes a rubric of any kind. The replacement says that a case needs a small working vocabulary and that no firm publishes what it measures.

Added in this version: the nine OC&C office requirements quoted verbatim with the count done by hand, the single C1 level and what it implies about the other eight, the two countries ZS names against the firm that names none, the McKinsey job posting that carries the requirement where a student would never look, the two firms out of ten that publish an interview language, the two scored English assessments and their missing thresholds, a five-step plan whose first step is the only one that removes uncertainty, our own four structuring weights with the exposure of each, and the four firms whose careers pages returned 403 to every method we tried.