Cornerstone Research has put two complete interview cases on its careers site, with the questions, the hints, and the analysis it wants to hear. Most candidates skim them and go back to practicing market sizing.
Read them side by side and the thesis of this guide writes itself. Both cases are the same case. Someone has attributed an entire observed change to one alleged cause, and your job is to take everything else away until only the defensible part is left. One case does it to a $40 stock drop and ends at $3. The other does it to a manufacturer's lost revenue and ends near zero.
That skill, subtraction under scrutiny, is what a litigation-economics firm sells. Below: the published shape of the process, the four things the firm says it assesses, an event study you can run by hand, and why the behavioral half weighs more than you expect.
What the firm publishes, and what it does not
The firm's own careers pages are unusually specific for this industry, so start there rather than with folklore.
| Item | What Cornerstone Research states |
|---|---|
| Rounds | Two: two virtual 30-minute first-round interviews, then a multi-hour final round in an office |
| Cases | A different case in each round; behavioral and case questions in both |
| Assessed | Thought process, comprehension, communication skills, and the ability to receive and apply real-time feedback |
| AI | Welcome for preparation; not permitted during interviews or assessments |
| Eligibility | Analyst: college seniors or master's students. Summer Analyst: college juniors |
| Materials | Cover letter, resume, transcript |
| Firm | Founded 1989; more than 1,000 people across nine offices |
Sources: the firm's interview process page, its Analyst Experience brochure for 2026, and its core values page. The Analyst and Summer Analyst processes are stated to be identical.
Not published: any online test, any pay figure, any list of behavioral questions, or comparable detail on the PhD and experienced-hire tracks. Salary numbers live on Glassdoor and in group chats. Treat them as folklore and confirm with recruiting.
Both published cases are the same case
The two cases on the firm's case examples page look different. One is a securities matter: Zilo, a high-tech IPO whose stock falls after it loses two large customers, and whose shareholders claim $40 per share. The other is an intellectual-property matter: Duff, found to have used ACME's patented manufacturing techniques, with ACME claiming hundreds of millions in lost profits.
Strip the industry away and the same four beats appear in the same order.
| Beat | The question behind it | Zilo (securities) | ACME v. Duff (IP) |
|---|---|---|---|
| 1. Name the assumption | What is wrong with the claimant's number? | $40 treats every move in the class period as the misconduct | Assumes Duff's customers would otherwise have bought from ACME |
| 2. List what else | What other explanations exist? | Economy-wide, industry-wide, other Zilo news | A downturn, new competitors, a new technology |
| 3. Propose the test | What data would separate them? | Regress Zilo on the S&P 500 and an industry index in a clean pre-period; flag significant residuals in the class period | Compare the two customer bases; test the technology explanation econometrically |
| 4. Finish the arithmetic | What is left after honest subtraction? | About $3 of the $40 | Little revenue lost; after variable and fixed costs, little profit lost |
Two things to notice. The firm's own hint for beat 1 says the problem is "similar to how an economist would think about measuring the causal impact of one variable on another." That is the whole interview in one sentence. And the hint for beat 2 hands you the three buckets: economy, industry, company. Use them by name; an interviewer who wrote the case will hear their own structure coming back and relax.
You will get a case you have not seen. It will have these four beats anyway. Move a fresh fact pattern through them out loud in 30 minutes and the content half of the interview is handled.
The four things on the scoring sheet
Cornerstone Research says its process lets it assess "your thought process, comprehension, communication skills, and ability to receive and apply real-time feedback." Candidates prepare for the first three. From the interviewer's chair, the fourth separates offers from rejections at the margin.
A Cornerstone interviewer is not scoring a solo performance. The published tips end with "Listen actively—your interviewers are there to guide you," and the case hints are written as nudges: "take a step back," "think about a statistical analysis that could be helpful here." In the room, the nudge is the scored moment, and three things can happen.
- You take it and say what changes: "Better frame. If the rival's sales did not rise, the customers did not go there, so the stolen list is not the mechanism." Full marks for the beat and for feedback.
- You take it silently and drift toward the answer. Partial marks. I cannot tell whether you understood or complied.
- You defend the original answer. You have just shown the one trait a litigation shop cannot employ. Analysts' work goes into expert reports that opposing economists take apart line by line, and an analyst who argues with edits is a liability.
Ned's rule. At Cornerstone, the first number anyone hands you is a ceiling, not an answer. Say what it assumes, name what it leaves out, and propose the data that would shrink it. The candidate who subtracts gets hired over the candidate who multiplies.
Scoring a weak and a strong answer
Here is how I would score two answers to an original prompt against the firm's four criteria, on a 1 to 5 scale of my own. My heuristic, not the firm's sheet.
Prompt: a regional grocer says its sales fell 30% after a rival allegedly poached its shoppers with a stolen customer list, and claims the whole 30% as damages. What is wrong with that?
| Criterion | Weak answer | Strong answer |
|---|---|---|
| Thought process | "Damages are 30% of revenue times margin." Straight to arithmetic. 2 | Names the assumption, then the buckets: regional grocery spending, a discounter opening nearby, the grocer's own price rise or closures. 4 |
| Comprehension | Misses that damages are profit, not revenue. 2 | Revenue is the wrong unit; variable costs come off, some fixed costs do not. 5 |
| Communication | Fluent but unstructured. 3 | Numbered, short, uses the interviewer's words back. 4 |
| Feedback | Nudged with "what if the rival's sales did not grow at all?" and repeats the 30%. 1 | Updates aloud: then the list was not the mechanism; match the list to the rival's new customers and their spend. 5 |
| Total of 20 | 8 | 18 |
Neither answer needed econometrics. A firm interviewer says on the careers page that the cases are "designed to be accessible to a wide variety of experiences and academic backgrounds," and the published analysis bears that out. The strong answer needed one reflex: ask "what else?" before touching a calculator.
A worked event study you can reproduce
The firm's first published case is a securities matter, and securities litigation is the practice its public research is best known for, so know what a market-model event study is. It is the tool behind Zilo's $40-to-$3 result. Here is one with original numbers.
Marlowe Diagnostics, a fictional device maker, discloses in an 8-K that it received an FDA warning letter seven weeks earlier. The stock closes at $41.00, down from $50.00. Plaintiffs' first cut: $9.00 per share times 120 million shares, a $1.08 billion headline.
Now subtract.
| Step | Figure | What it tells you |
|---|---|---|
| Raw change | $50.00 to $41.00, or -18.0% | The plaintiffs' number |
| Same-day market and peers | S&P 500 -2.5%; device-peer index -3.0% | A bad day for everyone |
| Betas from a 250-day pre-period regression | 0.8 on the market, 0.6 on the peer index | How Marlowe normally moves |
| Predicted return | 0.8 x (-2.5%) + 0.6 x (-3.0%) = -3.8% | Expected close absent firm news: $48.10 |
| Abnormal return | -18.0% minus (-3.8%) = -14.2%, or $7.10 per share | The firm-specific component |
| Residual standard deviation 2.2% a day | t is about -14.2 / 2.2 = -6.5 | Far beyond noise; a two-sided 5% test needs roughly 2 |
| Same-day confounders | The 8-K also announced the CFO's exit and a guidance cut | The $7.10 is a ceiling, not the answer |
Two lines of arithmetic have removed $1.90 per share, about $228 million, from the claim. The remaining $7.10 still has to be apportioned between the warning letter and the unrelated guidance cut, which is where experts spend months on intraday prices, analyst reports, and comparable disclosures.
A Cornerstone-shaped answer stops here and says so: "The abnormal return is large and significant, but it is an upper bound because of the same-day news. Class-wide damages would also need a per-share inflation path over the class period and a trading model for damaged shares. I know those exist; I would not build them in 30 minutes." That last sentence scores. Pretending you would build them does not.
Why the firm cares this much about subtraction
The firm's own research shows the gap its analysts work in. Its 2025 filings report, produced with the Stanford Law School Securities Class Action Clearinghouse, counted 207 new securities class actions in 2025 and a record $694 billion of Disclosure Dollar Loss, the fall in defendants' market value around the end of the class period. The same report says DDL "should not be considered an indicator of liability or measure of potential damages." Its 2025 settlements report counted 74 settlements totaling $3.0 billion, a median of $17.3 million, a median 3.5 years from filing to settlement hearing, and a median settlement equal to 6.5% of plaintiff-style damages, the report's proxy for claimed investor losses.
The distance between hundreds of billions of headline loss and a $17.3 million median settlement is not a scandal. It is the Zilo lesson at national scale: the raw drop is where the argument starts.
Courts have made this an analyst's problem specifically. In Halliburton Co. v. Erica P. John Fund (2014), the Supreme Court described event studies as "regression analyses that seek to show that the market price of the defendant's stock tends to respond to pertinent publicly reported events," and held that defendants may use price-impact evidence at class certification. A regression an analyst builds under an expert's name can bear on whether a class exists at all. That is why a 30-minute case asks whether you would control for the market before you multiply.
The 3.5-year median also explains the fourth scoring criterion. You will join cases that outlast your first job title, and your work will be edited by managers and experts for years. Taking feedback is not an interview nicety; it is the job.
The behavioral half is not the warm-up
A manager in the firm's Washington office says on the careers page that candidates should "consider the behavioral components as much or even more than the case element," and names the dimensions: "collaboration, conflict management, and mentorship." When a firm tells you what it is listening for, believe it.
| Dimension the firm names | What I listen for | The tell of a thin answer |
|---|---|---|
| Collaboration | A moment you changed your plan because of someone else's work or constraint | "We" throughout, no decision you personally made |
| Conflict management | A disagreement about substance, settled with evidence rather than seniority | A roommate story, or a conflict that ended because the other person left |
| Mentorship | A person you made measurably better, and how you knew | "I was always available for questions" |
Add a fourth story the firm does not name but every litigation interviewer wants: the time you found an error in your own work and told the team before anyone else did. Opposing experts read footnotes for a living. Keep each story near 90 seconds, give numbers where they exist, end with what you would do differently, and expect follow-ups.
Practice this today
One assignment, 40 minutes. Pick any public company that filed an 8-K this month and fell more than 10% the next trading day. Write half a page: the raw move; what the S&P 500 and the nearest sector ETF did that day; a rough abnormal return using a beta you look up; every other piece of news in the 8-K; and the one piece of data that would falsify a "the drop was the disclosure" story. Read it aloud in under three minutes. That is closer to a Cornerstone case than any market-sizing drill.
Then rehearse the half most candidates skip. CoachNed's behavioral interview runs your collaboration, conflict, and mentorship stories with four follow-ups each.
Be honest about fit. CoachNed's cases are business cases, not litigation cases, so use them for the communication layer, structuring out loud and taking a nudge without defensiveness, and use Cornerstone's own published cases for the content. The math drill covers arithmetic under time pressure, and the five-minute first rep gives you three typed turns with instant scores. Everything is open for seven days, no card; then $120 for a recruiting season or $49 a month.
CoachNed is independent and not affiliated with Cornerstone Research or any other firm named in this guide.
Frequently asked questions
Does Cornerstone Research use case interviews for analysts?
Yes. The firm states that each of its two rounds includes a case and that the case differs between rounds. Its published examples are a securities damages problem and an IP lost-profits problem, both about isolating the alleged conduct's effect from everything else.
How long is the Cornerstone Research interview process?
As published: two virtual 30-minute first-round interviews, then a multi-hour final round in one of the firm's offices. The gap between rounds is not published; ask your recruiter.
Do I need econometrics to pass the Cornerstone Research case?
No. The firm's interviewers say the cases are accessible across academic backgrounds. You need the instinct to ask what else could explain a change, the economy-industry-company buckets, and enough comfort with a regression to say what a pre-period model predicts. The worked example above is the level.
Can I use AI during a Cornerstone Research interview?
No. The firm states that AI is welcome for preparation but not permitted during interviews or assessments, and asks that applications reflect your own thinking and voice.
What does Cornerstone Research pay analysts?
The firm does not publish analyst compensation. Online figures are candidate-reported and vary by office and year. Confirm base, bonus, and any signing payment with recruiting.
Sources
- What We Look for in Analyst Candidates — rounds, assessed skills, AI policy, interviewer quotes on behaviorals. Checked 2026-09-24.
- Analyst Interview: Case Examples — the Zilo and ACME v. Duff cases with the firm's hints and analysis. Checked 2026-09-24.
- The Analyst Experience (US, 2026) — eligibility and application materials. Checked 2026-09-24.
- Cornerstone Research core values — founded 1989; more than 1,000 people across nine offices. Checked 2026-09-24.
- Securities Class Action Filings: 2025 Year in Review — 207 filings, $694 billion DDL, and the DDL caveat. Checked 2026-09-24.
- Record High Median Securities Class Action Settlement Amount in 2025 — 74 settlements, $3.0 billion, $17.3 million median, 3.5 years to settlement hearing. Checked 2026-09-24.
- Securities Class Action Settlements: 2025 Review and Analysis (PDF) — median settlement at 6.5% of plaintiff-style damages. Checked 2026-09-24.
- Halliburton Co. v. Erica P. John Fund, Inc., 573 U.S. 258 (2014) — the Court's description of event studies and its holding on price-impact evidence. Checked 2026-09-24.