Venture Capital interview preparation

Questions

100

Traced to a firm

31

Firms

12

Updated

September 2026

Asked at

All firmsGeneral Atlantic9Insight Partners7Silver Lake6Vista Equity Partners4Bessemer Venture Partners3ACAccel2Advent International2Battery Ventures2Andreessen Horowitz1Coatue Management1Sequoia Capital1WPWarburg Pincus1

Topic

All topicsSourcing and deal flow5Market sizing and estimation8Founders and teams5Unit economics and cohorts11Term sheets12Cap table and dilution7Early-stage valuation7Portfolio construction6Board and governance4Down rounds and secondaries4Exits and liquidity4Fund economics5Sector theses and markets6India venture market6Fit and motivation10

Level

AnyCoreIntermediateHard

Type

AnyTechnicalFitCaseMarket viewBrainteaser

1. 001How do you source companies?Sourcing and deal flowIntermediatephone / first roundGeneral AtlanticTechnology, Media and Telecom · New York · 2016General AtlanticGeneralist · Beijing · 2014

   Say this

   Thesis first, then a systematic channel to work that thesis, then relationships that make the outreach land. I would not describe sourcing as networking, because networking has no denominator. It is a funnel you can count.

   Then walk it

   1. Start with a thesis: a market shift you believe in, written as a sentence. 'Vertical SaaS for Indian pharma distribution' is a thesis. 'Interesting AI companies' is not.
   2. Then map the space exhaustively. Every company in the category, from Tracxn, Crunchbase, app-store rankings, GitHub stars, job postings, conference speaker lists. Twenty to fifty names, not five.
   3. Rank them on signals you can see from outside: hiring velocity, web traffic trend, review volume, who the angels were. The last one matters most early: a great seed round with three operator angels from the same category is a real signal.
   4. Then outbound. A specific, short email that shows you have used the product and understand the wedge. Response rates on a thesis-led email run several times higher than a generic one, and founders talk to each other about which VCs send lazy notes.
   5. Relationships are the compounding layer on top, not the substitute for it. The best repeat channel is founders you already backed, and second-time founders from companies in your thesis.
   6. And keep the denominator. I would track companies mapped, first meetings taken, second meetings, term sheets. If the conversion from first meeting to second is under a fifth, my filter is wrong, not my outreach.

   Where candidates lose it

   Answering 'I'd use my network and go to events'. That tells the interviewer nothing and describes what everybody already does. They want a repeatable process with a thesis at the front and a number at the back. Name real tools and one live thesis you are working.

   Expect next

   • Give me a thesis you are working right now and the ten companies in it.
   • How would you source in a sector where you have no network at all?
   • What is your reply rate on cold outbound, and what makes it better?

   Reported by candidates at General Atlantic (Technology, Media and Telecom, New York, 2016); General Atlantic (Generalist, Beijing, 2014). Source: Wall Street Oasis.

2. 006How do you size a market for a company that is creating a category that does not exist yet?Market sizing and estimationHardtechnicalEarly-stage VC

   Say this

   You cannot size the category, so you size the behaviour it replaces and then size the behaviour it unlocks. Two numbers: the budget or time being spent on the old way today, and the population that could not participate before because the old way was too expensive.

   Then walk it

   1. Start with substitution. Find the spend that already exists in an adjacent, ugly form — the agency fee, the manual process, the spreadsheet plus two analysts. That is a floor you can defend with real data.
   2. Then expansion, which is where the real answer lives. New categories are usually big because they drop the price by an order of magnitude and bring in users who were priced out. Ride hailing was not sized correctly off the taxi market; it was several times bigger because at half the price people stopped taking the bus.
   3. So build it as price times units at the new price point, not at the old one. That single step is what separates a serious estimate from a top-down slide.
   4. Sanity-check with a revenue-per-user bound. If you claim a billion users at $50 a year in a country where average annual discretionary spend on that category is $8, the number is wrong and you should say so.
   5. Then reverse the question, which is the answer interviewers actually want: forget TAM, what does this company have to be true to return my fund? If I need a $3bn exit, that is roughly $300m of revenue at a 10x multiple, which is 3 million users at $100. Is 3 million users plausible in ten years? That question is answerable; 'what is the TAM' is not.
   6. And say the limitation out loud: for genuinely new categories the TAM number is theatre. It is a test of whether your reasoning holds, not a forecast anyone believes.

   Where candidates lose it

   Pulling a Gartner number off a slide. The interviewer wants to watch you build it. And sizing the incumbent market only — that is the error that made every early taxi-market analysis of ride hailing too small by a factor of five.

   Expect next

   • So what does this company need to look like for us to make 10x?
   • When is a small market actually fine?
   • How would you size the market for an AI coding agent?
3. 007Bottom-up or top-down market sizing — which do you trust, and why?Market sizing and estimationCorephone / first roundGrowth equity

   Say this

   Bottom-up, always, and I use top-down only as a sanity check. Bottom-up is units times price, built from things you can count. Top-down is a big industry number times a percentage you made up, and that percentage is doing all the work.

   Then walk it

   1. Bottom-up: number of potential customers, times the share you can realistically win, times what each one pays. Every input is arguable on its own terms, which is the point — the interviewer can push on one number rather than the whole thing.
   2. Top-down: 'the global logistics market is $10 trillion and we only need 1 percent.' That sentence has appeared in every failed pitch deck ever written. The 1 percent is unjustified and usually off by a factor of ten.
   3. Worked example. Indian restaurant POS software: roughly 500,000 addressable organised restaurants, maybe 40 percent can pay for software, at ₹2,000 a month that is ₹4.8bn a year, call it $58m of Indian SaaS revenue. Now you can argue about penetration and price with real edges.
   4. And notice what bottom-up just told you: a $58m market cannot support a venture-scale outcome on software alone, which is exactly why every Indian restaurant-tech company ends up in payments or lending. Top-down would never have surfaced that.
   5. Use top-down to check the order of magnitude. If bottom-up gives you $58m and the top-down says $6bn, one of them is wrong and finding out which is the real work.
   6. The honest limitation: bottom-up systematically underestimates genuinely new categories, because it prices at today's price point. So for a category-creating company I build bottom-up at the new price, not the old one.

   Where candidates lose it

   Saying 'both, they're complementary' and stopping. That is true and empty. Commit to bottom-up, then show one worked build with real numbers. The follow-up is always 'size it for me now', so have a live example ready.

   Expect next

   • Size the Indian SaaS market for restaurants, out loud, right now.
   • When does bottom-up mislead you?
   • What is the difference between TAM, SAM and SOM?
4. 009What is the difference between TAM, SAM and SOM, and which one actually matters?Market sizing and estimationCorephone / first roundGrowth equity

   Say this

   TAM is everyone who could conceivably buy the category, SAM is the slice this product and business model can actually serve, and SOM is the share you can realistically win in your planning horizon. SAM is the one that matters for the investment decision.

   Then walk it

   1. TAM: total addressable market, the whole category with no constraints. Useful only for establishing that the ceiling is not the binding problem.
   2. SAM: serviceable addressable market. Constrained by geography, segment, regulation, language, price point and what your product does today. This is where the honest number lives.
   3. SOM: serviceable obtainable market, your realistic share given competition and your distribution. For a seed company this is the five-year revenue ceiling, and it should be big enough to return the fund.
   4. Worked example. Global payroll software might be a $30bn TAM. Payroll for Indian companies with 50 to 500 employees is maybe a $250m SAM. Winning 15 percent of that is a $38m revenue business — a real company, and possibly too small for a $500m fund. That comparison is the entire decision.
   5. So the question I actually answer is: does the SOM support an outcome that returns the fund at the ownership I can get? Everything else is framing.
   6. The limitation worth naming: these boundaries are soft and companies move between them. Every great company's SAM expanded — Amazon's was books. So I hold the SAM number loosely and ask whether the expansion path is credible rather than assumed.

   Where candidates lose it

   Getting the definitions right and then failing to say which one drives the decision. Definitions are a two-mark question; the judgement is in connecting SOM to fund returns at your likely ownership. And do not confuse SAM with 'the market we're targeting first' — that is the beachhead, which is smaller again.

   Expect next

   • What SOM do you need for this to return a $200m fund?
   • Give me a company whose SAM expanded dramatically.
   • How would you size this bottom-up?
5. 014What do you look for in a founding team?Founders and teamsCorephone / first roundEarly-stage VCSeed funds

   Say this

   Four things, in this order: earned insight into the specific problem, the ability to recruit people better than themselves, unusual speed of learning, and enough resilience to survive three years of it not working. At seed, the team is most of what you are underwriting.

   Then walk it

   1. Earned insight. Not domain experience as a line on a CV, but a specific, slightly contrarian belief about the market that came from doing the work. The test is whether they tell me something about the industry I did not know and could not have read.
   2. Recruiting ability. The first ten hires determine the company, and the only evidence that matters is who has already said yes to them. If a genuinely impressive engineer left a good job to join, that is a stronger signal than any reference.
   3. Rate of learning. I compare what they said three months ago to what they say now. Founders who update fast on evidence compound; founders who defend the original plan do not.
   4. Resilience, which is the one nobody can fake for long. Most companies spend a long stretch looking dead. I look for prior evidence of finishing something hard with no external pressure to do so.
   5. On co-founder dynamics: clear decision rights, complementary skills rather than duplicated ones, and a track record of disagreeing productively. I would rather see two people argue in front of me than perform agreement.
   6. The honest limitation: founder assessment is where investors are most overconfident. My pattern-matching is largely a bias toward people who remind me of people who already worked, which is how whole categories of founders get missed. So I weight evidence from the business over my read of the person wherever I can.

   Where candidates lose it

   Giving the generic list — passionate, smart, hard-working. Everyone the fund meets is those things. The differentiators are recruiting evidence, rate of learning, and earned insight, all of which are observable. And you must name the bias problem, because the honest answer to 'how do you judge founders' includes 'imperfectly'.

   Expect next

   • How do you tell conviction from delusion?
   • Would you back a solo founder?
   • What is the strongest founder signal you have ever seen?
6. 015How do you tell conviction from delusion in a founder?Founders and teamsHardsuperdayEarly-stage VC

   Say this

   By how they handle disconfirming evidence, not by how strongly they believe. Both look identical from the front. The difference is that the convicted founder can state exactly what would change their mind and can recite the counterargument better than you can.

   Then walk it

   1. Test one: ask for the strongest case against the company. A convicted founder gives you a sharper bear case than your own and then tells you why they are taking the risk anyway. A deluded one tells you there isn't one.
   2. Test two: ask what data would make them stop. 'We'd know by Q3 whether the enterprise motion works, and if payback is still over 30 months we pivot to self-serve' is conviction. 'It will work' is not.
   3. Test three: look at what they have already changed. Every founder who has been at it eighteen months has been wrong about something. Ask what, and what they did. Someone who has never revised anything either has not shipped or is not listening.
   4. Test four: separate the belief about the destination from the belief about the route. Stubborn on the mission, flexible on the path, is the combination that works. Stubborn on both is the failure mode.
   5. Watch how they talk about customers who said no. Delusion sounds like 'they didn't understand it'. Conviction sounds like 'they didn't have the budget line, so we changed who we sell to'.
   6. And the limitation I would admit: this call is genuinely hard and the same trait produces both outcomes. Several of the best companies of the last twenty years looked delusional at seed and their investors have said so. So I would rather be wrong by backing a few founders who turned out deluded than build a filter so tight it screens out the outliers.

   Where candidates lose it

   Framing it as a personality read — 'you can just tell'. Interviewers hear that as pattern-matching with no method. Give behavioural tests that produce observable answers, and admit that the best outcomes often looked like the failure mode early.

   Expect next

   • Give me a company that looked delusional and worked.
   • What would make you pass on a founder you liked?
   • How do you avoid being sold to in a founder meeting?
7. 017How do you reference-check a founder?Founders and teamsIntermediatetechnicalGrowth equityEarly-stage VC

   Say this

   Off-list references are the only ones that matter, and the useful calls are with people who worked for the founder rather than above them. On-list references tell you the founder can pick three friends.

   Then walk it

   1. Get the list, call it quickly, and treat it as a formality. Then build your own list: former direct reports, a co-founder they parted from, customers who churned, and an investor from a previous company.
   2. Direct reports are the highest-signal call. Ask whether they would join this founder again, and listen to the pause before the answer. Ask who else on the team should I talk to, which quietly widens the list.
   3. Ask behavioural, not evaluative, questions. Not 'is she a good leader' but 'tell me about a time she changed her mind' and 'what happened the last time the company missed a quarter'. Stories are checkable; adjectives are not.
   4. Always ask the negative directly: 'what is the thing that will frustrate their next investor?' Referees will tell you, but only if you ask in a way that gives them permission.
   5. Then triangulate with customer calls, which for growth-stage deals are worth more than the founder references. Ask what would make them switch away and what the renewal conversation actually looked like.
   6. The limitation: references are systematically positive because the network is small and nobody wants to torch a relationship. So I read them for the shape of the concerns rather than a verdict, and I weight one specific negative story over five glowing generalities.

   Where candidates lose it

   Only calling the list you were given, and asking questions that can be answered with 'yes, she's great'. Also forgetting that founders find out you called. Off-list references need handling with judgement, especially with a live process and a signed term sheet in the market.

   Expect next

   • What would you do if one off-list reference was strongly negative?
   • How do you reference-check without damaging the relationship?
   • What do you ask a customer that you cannot ask the founder?
8. 019Walk me through CAC, lifetime value and payback, and tell me how each one gets manipulated.Unit economics and cohortsCoretechnicalGrowth equity

   Say this

   CAC is all the money you spent to acquire a paying customer, divided by the customers you actually acquired. LTV is the gross profit that customer produces over their life, discounted. Payback is how many months of contribution margin it takes to earn the CAC back. Payback is the one I trust.

   Then walk it

   1. CAC properly done is fully loaded: paid media, sales and marketing salaries and commissions, tools, and any onboarding cost, divided by new paying customers in that period. Not just ad spend.
   2. LTV is gross-margin based, not revenue based: ARPU times gross margin, divided by monthly churn, discounted if the life is long. Using revenue instead of gross profit inflates it by whatever your cost of service is.
   3. Payback in months is CAC divided by monthly gross profit per customer. Best-in-class B2B SaaS is under 12 months, acceptable is 12 to 18, and above 24 months you are running a financing business rather than a software business.
   4. The three manipulations to look for. One, blended CAC that folds organic and word-of-mouth customers into the denominator while only counting paid spend in the numerator — always ask for paid CAC on paid customers. Two, LTV built on an early cohort's churn, which is always the best cohort. Three, a churn assumption of 1 percent monthly applied to a company that is eighteen months old and has never observed a five-year life.
   5. The practical rule I would state: LTV/CAC above 3 is the convention, but it is nearly meaningless without payback, because a 5x LTV/CAC with a 36-month payback will kill the company on cash before the ratio ever pays out.
   6. One real number to anchor it: at $12,000 CAC, $1,000 monthly revenue and 80 percent gross margin, payback is 15 months. If monthly churn is 2 percent, implied life is 50 months and LTV is $40,000, so LTV/CAC is 3.3x. Both numbers are fine; the fragile input is that 2 percent.

   Where candidates lose it

   Quoting LTV/CAC above 3 as if it settles the question, and using revenue instead of gross profit in LTV. Also: not asking over what period CAC was measured. Founders present a good quarter. Ask for twelve months and for paid-only CAC, and the ratio usually halves.

   Expect next

   • Why do you prefer payback to LTV/CAC?
   • What is a good payback period for consumer versus enterprise?
   • How would you calculate CAC for a marketplace?
9. 020What is the burn multiple and why do investors like it?Unit economics and cohortsIntermediatetechnicalGrowth equityLate-stage VC

   Say this

   Net burn divided by net new annual recurring revenue over the same period. It answers one question: how many dollars did you set on fire to buy a dollar of new recurring revenue. Under 1.5 is excellent, 1.5 to 2 is fine, above 3 means the growth is bought rather than earned.

   Then walk it

   1. The calculation: if you burned $12m in a year and added $6m of net new ARR, the burn multiple is 2.0. Net new ARR is net of churn and downgrades, which is the whole point — it punishes growth that is leaking out the back.
   2. Why it beats the alternatives: growth rate alone rewards companies that buy revenue, and efficiency ratios based on a single quarter can be gamed by pausing spend. The burn multiple is one number that captures both sides at once.
   3. Rough bands, and these hardened after 2022: under 1 is exceptional, 1 to 1.5 is very good, 1.5 to 2 is acceptable at scale, 2 to 3 needs a specific explanation, above 3 is usually a broken go-to-market rather than an investment phase.
   4. It is stage-sensitive and you should say so. A company going from $1m to $3m of ARR will have an ugly multiple because the fixed cost base dominates. From $20m to $40m it is a genuine judgement on efficiency.
   5. What it hides: a company can produce a lovely burn multiple by starving R&D and harvesting an existing base. So I read it next to net revenue retention and the R&D share of spend — good multiple plus deteriorating NRR is a company eating its seed corn.
   6. In practice this is now the first number a growth investor asks for, because it is the cleanest available proxy for whether more capital produces more company.

   Where candidates lose it

   Using gross burn or using total ARR instead of net new ARR. Both make the number look better and both are wrong. And quoting benchmark bands without adjusting for stage — a seed company's burn multiple is nearly uninformative, and saying so is part of a correct answer.

   Expect next

   • What is the magic number and how does it differ from this?
   • How would you fix a burn multiple of 4?
   • What is a good burn multiple for a Series A company?
10. 021What is the SaaS magic number, and what does a reading of 0.5 tell you?Unit economics and cohortsIntermediatetechnicalGrowth equitySaaS-focused funds

    Say this

    Magic number is the annualised increase in quarterly recurring revenue divided by the prior quarter's sales and marketing spend. A reading of 0.5 says every dollar of sales and marketing bought fifty cents of annual recurring revenue, which implies a payback of about two years. That is a hold, not a spend signal.

    Then walk it

    1. Formula: (current quarter revenue minus prior quarter revenue) times four, divided by prior quarter sales and marketing expense. The times four annualises it, and the one-quarter lag reflects that spend converts with a delay.
    2. Reading it: above 1.0 means gross payback inside a year, so step on the accelerator. Between 0.75 and 1.0 is healthy. Below 0.75 means fix the funnel before adding budget. Below 0.5 usually means the segment or the channel is wrong.
    3. So 0.5 implies roughly 24 months to recover the acquisition cost on a gross-revenue basis, and longer on a gross-profit basis. At that level, raising more money to hire more reps makes the company worse, not bigger.
    4. What I would do about it rather than just diagnose it: split the number by segment and channel. Usually one motion is at 1.2 and another is at 0.2, and the blended 0.5 is hiding the fact that they should stop selling to the small accounts.
    5. The weaknesses, which you should volunteer: it uses revenue rather than gross profit, so it flatters low-margin businesses. It is noisy quarter to quarter for small companies. And it treats sales and marketing as a single lump when brand spend and quota-carrying rep cost have completely different lags.
    6. Which is why in practice I would look at it alongside CAC payback on a gross-profit basis and the burn multiple. Magic number is the quickest read; it is not the deepest one.

    Where candidates lose it

    Getting the formula slightly wrong — forgetting to annualise, or using the current quarter's sales and marketing spend instead of the prior quarter's. And treating the benchmark as a verdict rather than splitting it by segment, which is where the actual insight is.

    Expect next

    • How does that differ from the burn multiple?
    • What is a good magic number at Series B versus Series D?
    • If it is 0.4, what do you tell the CEO to do on Monday?

Firm tags come from public, anonymous candidate reports on Wall Street Oasis: strong signal, not sworn testimony. Firms are named as the places a question was reported, not as partners of Fin Maverick. Answers are written for this page to show how to think out loud; they are not scripts to recite.