Analysis Group Case Interview 2026: Bring Your Own Project

The two Analysis Group interview rounds the firm actually publishes

The Analysis Group careers FAQ and the FAQ block inside the analyst role page describe the same process from two angles. Put together, they give you this.

Two details in that table are easy to skim past.

The first is that the analyst and the associate are not walking into the same day. The associate track carries more meetings and an explicit note that the format shifts with the degree. A doctoral candidate presents research. An undergraduate does not.

The second is the phrase "formal and conversational". It appears in the firm's wording, and it is a warning. Some of the meetings in that building are not scored the way you expect. A coffee conversation with a consultant is still part of a hiring day, and treating it as a break is a quiet way to lose.

Which office that is deserves a line, because the second round happens inside one. Analysis Group gives its headquarters as Boston, Massachusetts, and states that it has 15 offices across North America, Europe and Asia, read 8 October 2026. The cover letter the posting asks for is where you name which of them you want.

The analyst page adds the word the FAQ does not: analytical. Interviews there cover "a mix of behavioral, teamwork, and analytical topics". The firm names that third bucket and never describes it. We will not invent a description for it either, but the next two sections narrow what it can reasonably contain.

Why a strategy case is the wrong thing to rehearse

Analysis Group writes its own boundary, which saves us guessing. Its economic consulting page sets out the difference directly.

Read the right-hand column as a specification for the interview and the shape changes.

A profitability tree answers "where is the money leaking and what should we do." A damages question answers "did this specific act cause this specific loss, and how much was it." The second needs a counterfactual: what the world would have looked like if the act had not happened. No standard business framework contains that box.

The firm's published mock matter makes the point concrete. In the Pear TV damages litigation, a streaming platform loses a channel group for two weeks after a carriage negotiation fails. The firm poses the questions in its own words: did the blackout cause subscribers to leave, if so how many, and at what cost.

Notice what is missing. There is no ask for a recommendation. Nobody wants to know whether Pear TV should enter a new market. The deliverable is a defensible number and a method that survives an opponent trying to break it.

The same page says the finished report must be understood by non-economists, including judges and juries. So the skill being hired is estimation you can explain to someone who will never see your code. Another economic consultancy puts that same skill under a clock, and the Bates White case interview runs 25 to 40 minutes in three published parts with a worked case on price fixing beside them.

The Analysis Group interview turns on the project you bring

Here is the single most actionable line the firm publishes, and it is in the FAQ. Analyst candidates are told to be prepared to talk through their own experience on a research or quantitative project. The firm gives three examples: a project that involved analyzing quantitative data, an internship task that involved building a model, or a research paper written for a college course.

That is a prompt you can prepare exactly, because you already own the material.

Pick one project. Then be able to answer these, out loud, without notes.

1. What question was it answering? One sentence, no jargon. If you cannot say it plainly, the interviewer learns something about your writing before you reach the analysis.
2. Where did the data come from, and what was wrong with it? Sample size, missing values, the observations you dropped and why. Dropping rows is a judgment call, and judgment calls are what a second reader probes.
3. Why that method and not the obvious alternative? If you ran a regression, say what you would have seen from a simple mean comparison and why it was not enough.
4. What would have changed your conclusion? Name the assumption carrying the most weight. The firm's own case page describes stress-testing a methodology by trying alternate growth rate assumptions, so this habit is the job.
5. What did you get wrong the first time? A project with no wrong turn either was trivial or is being remembered generously.

Two failure modes show up here. One is narrating the pipeline step by step and never reaching a finding. The other is stating the finding with total confidence and having no idea how fragile it is. The second is worse in this setting, because the whole trade is built on surviving an opposing expert.

Our synthesis drill scores the answer-first habit that fixes the first failure. The teamwork and leadership questions sit beside all of this, and the fit practice covers the probe that breaks a thin story.

A damages question, worked in dollars

Analysis Group publishes the Pear TV matter with no figures at all. So the numbers below are ours, built in the shape its page describes, to show what the arithmetic feels like when the question is causal rather than strategic.

The setup. A streaming platform had 4,000,000 subscribers when a two-week blackout began. Over the prior year it lost 2.0% of its base per month. In the blackout month it lost 116,000. Average revenue per subscriber is $60 a month at a 40% gross margin. A lost subscriber would have stayed 20 more months.

Move one: do not use the raw jump. Raw excess looks like 116,000 minus 80,000, so 36,000. That number assumes nothing else moved that month, and something did. A comparable market with no blackout saw churn rise from 2.0% to 2.3% in the same weeks.

Move two: build the counterfactual. Apply the untouched market's rate to this base. That is 2.3% of 4,000,000, so 92,000 expected cancellations. Excess attributable to the blackout is 116,000 minus 92,000, which is 24,000 subscribers.

Move three: price one subscriber. Contribution is $60 times 40%, so $24 a month. Over 20 months that is $480. Gross harm is 24,000 times $480, which is $11,520,000.

Move four: take off what the other side will take off anyway. Suppose a quarter of those subscribers return within a year. That leaves 18,000 permanently lost, so $8,640,000. And the platform paid no carriage fee during the blackout: $0.90 per subscriber per month across 4,000,000 subscribers for half a month is $1,800,000 saved.

Move five: land it. Net harm is $8,640,000 minus $1,800,000, so about $6.8 million. Say it in one breath, then name the weak joint yourself: the whole figure moves with the 2.3% counterfactual rate. If the comparison market is a poor match, the number is wrong, and that is the first thing an opposing expert will attack.

Sanity check before you speak: 24,000 is 0.6% of the base, lost over two weeks. That is large but not absurd for a blackout. A figure of 600,000 would have been absurd, and catching that is worth more than the decimal places.

How a quantitative answer gets graded here

Analysis Group publishes no marking scheme. It publishes a GPA floor and a salary estimate, and nothing about how an interview answer is judged. The hundred points below are ours, written by us. They are the dimensions our own grader measures on quantitative work, and the firm has never seen them.

The Analysis Group analyst salary estimate and its 3.3 GPA floor

The guides we read describe the interview and skip the posting, which is where the firm is most specific. These come from a live 2027 generalist analyst posting on the firm's own applicant portal, read on 29 September 2026.

Four things follow from that table.

The GPA line is a filter. The posting states it as a condition of consideration, ahead of anything a cover letter could say. The deadline belongs to the posting rather than to the firm: the full-time and internship postings we read carried different dates, and the FAQ tells you to check the relevant posting. The transcripts requirement means the application is slower to assemble than a one-click apply, so start earlier than you think. And the cover letter carries a functional job, naming your office preferences, which the online form also asks for.

The internship is a separate posting with a separate date, and the FAQ puts a length on it: the program "runs for 10-12 weeks, typically starting in early June and running through the middle of August." The same analyst and associate split that governs the second round governs the internships, so read the posting that matches your degree rather than the one a search engine served.

One more line matters for anyone a few years out of school. The same posting says candidates with three or more years of directly relevant experience may be considered for a senior analyst title instead, reviewed from the same materials.

Six ways candidates misread the Analysis Group interview

1. Rehearsing a profitability tree. The firm describes questions tied to legal disputes, regulation and public policy. A tree with no counterfactual branch answers a question nobody asked.
2. Turning up without a project ready. The FAQ names this exercise in writing. It is the one part of the day you can prepare word for word, and the easiest to leave to improvisation.
3. Landing neither half of the answer. Describing the method and never stating the finding fails on Answer, which carries the most weight on our rubric. Stating the finding and naming no weak joint fails the other way, because an opponent will look for that joint whether you do or not.
4. Treating the conversational meetings as breaks. The firm's own phrase is "formal and conversational settings", and both are part of the round.
5. Preparing econometrics you do not have. The analyst page says you do not need coding experience and that training is provided. Claiming a method you cannot defend is worse than naming the one you know.
6. Assuming the associate process applies to you. The job talk belongs to doctoral candidates. An undergraduate rehearsing a research presentation is preparing for someone else's day.

The first three are about the analysis. The last three are about reading the right page for your own application, which is cheaper to fix and nearly as common. Our practice hub shows what else is here for a specific firm.

What Analysis Group does not publish

How long an interview runs. No page gives a duration, for either round. Any minute count you find elsewhere came from somewhere other than the firm.

What "analytical topics" contains. The analyst page names the category and stops. We are not filling it in. It might be a discussion of your own project, a reasoning question, or arithmetic on paper, and the firm does not say which.

Whether there is an online assessment. Nothing we read describes a test, a game or an automated screen. That is not proof none exists. It means nobody should tell you to prepare for one.

A passing bar, an acceptance rate, or a number of openings. None of these appears on any firm page or posting. The GPA floor is the only published threshold, and it gates the application rather than the interview.

How long the decision takes. The FAQ says you will hear back either way once your application is reviewed, and gives no interval for the stage after the final round.

Anything about the firm's own scoring. It publishes a salary estimate and a GPA floor, and no criteria for judging an answer. The weights on this page are ours, from our grader, and we say so wherever they appear.

If your invitation email contradicts anything above, the invitation is current and this page is not. Write to us with what you saw and the date, and we will read it back against the firm's pages before changing a word.

What changed on the Analysis Group page

We correct in public and we date the correction. The body above carries what the firm publishes today; this log carries what this page used to carry.

29 September 2026. The page previously said Analysis Group's screening steps could not be described, and declined to describe them. That was the page's central claim and it was false. The two rounds sit in the careers FAQ and in the analyst role page, and the GPA floor, the salary estimate and the deadline sit on a live 2027 posting on the firm's own applicant portal. All three documents are in the source register at the head of this page, with the date each was read.

29 September 2026. The claim that preparation guides skip the job posting is now written as our own reading of the guides we read, because we have counted nothing. The same treatment went to the sentence about the GPA line, which the posting itself states as a condition of consideration.

29 September 2026. The four math weights are named as ours at the point where the chart appears. Where we point at the firm's published case as a reason one of those weights is high, the page now says that the connection is our inference. Analysis Group publishes no marking scheme of any kind.