Analysis Group's careers FAQ describes its interview process in about 200 words, and the word "case" is not one of them. The first round is "behavior based" and virtual. The final round is three to five conversations with consultants for analyst candidates and four to eight for associates, about "your research, prior work and internship experience, teamwork and leadership, and career interests." The one concrete thing the firm asks analyst candidates to prepare is to "talk through your own experience working on a research or quantitative project."
The thesis is short: at Analysis Group, the case is your own research project. You bring it, and a room of economists interviews it. A candidate who spends a month on profitability frameworks is rehearsing a test the firm does not describe giving, and skipping the one it describes in detail.
By the end you will know what the firm states, what it used to state, how economists score a research walk-through, and what to do with the next four weeks.
What the firm publishes, and what it does not
Two pages carry the whole published process: the careers FAQ and the analyst page. Analyst means a bachelor's degree, or a master's with under four years of relevant experience; associate means an MBA, a PhD, or a master's with four or more. Final rounds happen "in both formal and conversational settings," typically at the office you applied to, and the project the FAQ wants analysts to bring can be a data analysis, an internship model, or "a research paper that you wrote for a college course."
The analyst page puts it more directly than the FAQ does: interviews "cover a mix of behavioral, teamwork, and analytical topics." Hold on to "analytical." It is the closest the firm comes to describing anything case-like.
Now the gaps. The firm publishes no question bank, rubric, interviewer titles, or pay; "the standard Analysis Group case" is folklore. The nearest candidate-level account is the summer program profile the firm hosts, where one intern describes a final round of "an AG/consulting interview, a behavioral interview, an economics quantitative interview, and a partner interview." Economics questions, yes; a framework case, no.
Five years of one sentence
The thesis leans on a sentence that has left the page; here is its history from archived copies.
| Snapshot | What the FAQ said about cases | First round |
|---|---|---|
| December 2021 | "Analysis Group does not conduct case interviews." | "Over the phone or on campus" |
| March 2025 | "Analysis Group does not conduct formal case interviews." | Virtual |
| February 2026 | Same, with "formal" | Virtual |
| Live page, 2026-09-24 | The sentence is gone; every other line is unchanged | Virtual |
I read the two edits like this. "Formal" arriving in 2025 tells me the firm wanted room to ask analytical questions without promising a case. The sentence leaving in 2026 could mean the process is shifting, or that recruiting grew tired of candidates hearing "no case," skipping quantitative prep, and going quiet when handed a table. The firm has not said which, so prepare for the stricter reading: no framework case, but a number, a chart, or your own regression, and the question of how you know.
Why economists interview like this
The firm explains the job on its What Is Economic Consulting? page: management consultants "deliver strategic recommendations," while economic consultants "produce detailed analyses, expert reports, and courtroom testimony." An analyst's regression can end up as an exhibit, and the other side hires its own economist to find the assumption that breaks it.
In US federal court that fight has rules. Federal Rule of Evidence 702 requires expert testimony to be "based on sufficient facts or data," to be "the product of reliable principles and methods," and to reflect "a reliable application of the principles and methods to the facts of the case." Replace "expert" with "analyst" and you have the interview. They are not asking what you found. They are asking whether it survives someone paid to disagree with it.
I have never sat on an Analysis Group panel and will not pretend otherwise. I have sat on a great many consulting panels, and the difference between a strategy interviewer and an economist is the follow-up question. Mine is "so what would you do?" Theirs is "compared to what?" A candidate trained on the first answers the second with a recommendation, and the economist notes that you did not hear the question.
How a research walk-through gets scored
The firm publishes no rubric, so this one is mine: what economists listen for when a candidate describes their own work, scored 0 to 2 on five lines.
| Line | 0 points | 1 point | 2 points |
|---|---|---|---|
| The question | "I studied X" | A topic with a direction | A comparison built in: "did A change B, relative to C" |
| The data | "A dataset" | Source and size | Source, unit of observation, period, and what is missing |
| Identification | "I ran a regression" | Names the method | Why the design isolates the effect, and what must be true for it to work |
| The result | "It was significant" | A sign | A magnitude with units and a rough sense of uncertainty |
| What would change your mind | Nothing | One threat | The threat, how you checked it, and what you could not check |
A typical weak answer scores 1, 1, 0, 1, 0: three out of ten. A strong one scores 2, 2, 2, 2, 1: nine. The gap is never the econometrics. It is whether the candidate can say the third and fifth lines aloud, and most have never been asked to.
Ned's rule. The interviewer is not grading your result. They are grading the sentence after it: "this would be wrong if..." If you cannot finish that sentence about your own project, you have finished a regression, not a piece of research.
Worked example: a senior thesis in five sentences
Every number below is invented so you can reproduce the arithmetic. Suppose your thesis studied Harbor City, which made six of its eighteen bus routes fare-free for twelve months as a pilot. Here is the five-sentence answer to "tell me about your research."
- Question: did removing fares raise ridership in Harbor City, relative to what would have happened anyway?
- Data: monthly boardings for all eighteen routes over twenty-four months, from the transit agency's open-data portal; six routes went fare-free in month thirteen.
- Design: difference-in-differences, comparing the change on the six free routes with the change on the twelve paid routes over the same months.
- Result: free routes rose from 40,000 to 50,000 boardings a month, up 25%; paid routes rose from 55,000 to 60,500, up 10%. Had the free routes grown like the paid ones they would sit at 44,000, so the pilot added roughly 6,000 boardings a route a month, about 14%.
- Caveats: the city chose which routes went free, riders can switch from a paid route to a free one, and six treated routes is a small number of clusters. I would call it 10 to 15% with wide error bars, and probably an upper bound.
That takes ninety seconds and scores nine on the table above. Then the interview starts.
| Follow-up to expect | Answer that leaks points | Answer that scores |
|---|---|---|
| "Why are the paid routes the right comparison?" | "They are in the same city." | "In the twelve months before the pilot both groups grew 3 to 4%. Had they diverged, I would have no design." |
| "What does rider switching do to your number?" | "It would be a problem." | "It pushes both groups toward a bigger estimate: free routes up, paid routes down. So 14% is an upper bound." |
| "How sure are you about 14%?" | "The p-value was under 0.05." | "With six treated clusters, clustered standard errors are unreliable. I would run a permutation test and say 10 to 15%." |
None of this needs a PhD. It needs one honest hour spent on your own project before the interview.
The analytical question you might still get
"Analytical topics" is the firm's phrase, and it says no more. What follows is my illustration of the kind of question applied economists ask, not a leaked one.
Market definition is everyday antitrust work. The Department of Justice's 2023 Merger Guidelines, section 4.3 describe the hypothetical monopolist test: would a single firm controlling a candidate group of products profitably impose "a small but significant and non-transitory increase in price," for which the agencies "will often use" 5%? Take a margin M (price minus incremental cost, as a share of price) and a price rise S. The monopolist breaks even when the lost share of volume L satisfies (M + S)(1 − L) = M, so the critical loss is L = S / (M + S).
| Input | Value | What it means |
|---|---|---|
| Margin M | 40% | 40 cents of margin per dollar of price |
| Price rise S | 5% | The agencies' usual SSNIP |
| Critical loss L = 0.05 / 0.45 | 11.1% | Lose more than this and the price rise is unprofitable |
| Predicted loss if own-price elasticity is −3 | 15.0% | Above 11.1%: the candidate market is too narrow, so widen it |
| Predicted loss if own-price elasticity is −1.5 | 7.5% | Below 11.1%: the candidate market passes |
Nobody is scoring recall of the formula; they will hand it to you. They are scoring whether you can derive it in a line, notice that the margin drives the answer, and ask where the elasticity came from. Assumption, magnitude, what would change your mind, again.
The behavioral round is not a warm-up
The first round is behavior based and decides who reaches the office. The recruiting profile the firm hosts wants someone "able to hold independent views and defend them, and a team player," the two behaviors most likely to collide in a story.
| Question you will hear | What is being checked | The tell of a weak answer |
|---|---|---|
| "Why economic consulting rather than management consulting?" | You know the product is analysis and testimony, not recommendations | "I like solving problems for clients" |
| "Tell me about a result that surprised you." | Whether you changed your mind on evidence | The story ends with you having been right all along |
| "Tell me about disagreeing with a teammate or supervisor." | An independent view held, then resolved, without a casualty | The disagreement was avoided rather than settled |
For PhD candidates the job talk is the case and the room is your discussant; budget as much time for questions as for slides.
Four weeks, in order
Week one is writing. Put your project into the five-sentence form above, then write the five hardest questions an opposing expert would ask and answer each in three sentences. Most candidates find the fifth scorecard line empty at this point, which is why it comes first.
Week two is speaking. Say the five sentences aloud until they take ninety seconds, then rehearse follow-ups. CoachNed's behavioral interview runs one story and presses with four follow-ups, the closest a machine gets to a "formal and conversational" economist meeting. Count how many answers ended in a recommendation rather than a comparison.
Week three is analytical reps, the part of case prep that does transfer: the chart drills for stating what a table shows and what it cannot, the math drills for clean arithmetic aloud, and the critical loss derived once by hand. Skip the live MBB voice case for this firm. It trains the "so what would you do" reflex you are trying to unlearn.
Week four is the behavioral table above and the firm's own material: its Pear TV mock case on the economic consulting page, and the FAQ, reread the week before you interview, because this guide has just shown you that it changes.
Everything on CoachNed is open for seven days, no card; then $120 for a recruiting season or $49 a month. CoachNed is independent and not affiliated with Analysis Group or any other firm named in this guide.
Frequently asked questions
Does Analysis Group do case interviews?
The firm's current FAQ does not mention one; it describes a behavior-based first round and a final round built around your research and prior work. Archived versions said the firm "does not conduct formal case interviews," so expect quantitative questions about your own project rather than a framework case.
How many interview rounds does Analysis Group have?
Two: a virtual behavioral first round, then a final round, typically in person, of roughly three to five consultant meetings for analysts and four to eight for associates.
Should I apply to Analysis Group as an analyst or an associate?
Per the firm's FAQ, a bachelor's or a master's with no relevant work experience means analyst; an MBA, a PhD, or a master's plus four years of experience means associate. Analysts who later want promotion must earn an advanced degree.
When does Analysis Group recruit?
As of September 2026 the firm says its full-time consulting cycle runs September to March and internship recruiting October to March, with applications accepted at any time. Summer analyst internships are for rising seniors or master's students graduating the following calendar year.
Sources
- Analysis Group careers FAQ — interview format, degree mapping, recruiting calendar. Checked 2026-09-24.
- Analysis Group analyst page — "analytical topics," staffing model. Checked 2026-09-24.
- What Is Economic Consulting? — the firm on management versus economic consulting. Checked 2026-09-24.
- Summer program profile (PDF) — intern's four-interview final round. Checked 2026-09-24.
- Recruiting profile (PDF) — "independent views" quote. Checked 2026-09-24.
- Archived FAQ, December 2021 — "does not conduct case interviews." Checked 2026-09-24.
- Archived FAQ, March 2025 — "does not conduct formal case interviews." Checked 2026-09-24.
- Archived FAQ, February 2026 — last archived copy with the sentence. Checked 2026-09-24.
- DOJ 2023 Merger Guidelines, section 4.3 — hypothetical monopolist test, 5% SSNIP. Checked 2026-09-24.
- Federal Rule of Evidence 702, Cornell LII — expert testimony standard. Checked 2026-09-24.