What this covers
This is a conversation with David Tisch of BoxGroup, a New York seed fund, about early-stage investing outside Silicon Valley. His method is volume: 30 to 40 investments a year, drawn from 3,000 to 4,000 companies seen. He says he cannot reliably predict which companies will reach the scale that drives returns, so the approach depends on tolerating heavy losses. He also argues that a sin of omission costs a fund more than a failed bet, and he names Casper, Zynga, Via and Remind101 as passes that bother him most.
The middle of the conversation is about accelerators. He explains why Y Combinator and TechStars appealed to him for community more than unit economics, and why the 90-day format ending in a Demo Day was deliberately harsh. He predicts most lower-tier programs will wind down, though specialized ones with domain networks, such as in fintech, may hold up. He also traces how Google's New York office, and later others, fed the local startup scene with talent.
The most contestable claims concern the current consumer landscape. Facebook and Google are now strong enough to build the future themselves. Users give a new app roughly one chance, which he says makes lean-startup iteration obsolete. No immature acquisition channel remains. Reaching the American consumer costs on the order of $100 million. He also says public-market pressure has reduced risk-taking by older companies, and that brands without real customer connection will die. For diligence, he suggests identifying the one or two skills a business depends on and checking whether the team has them.
Tisch argues that as consumer technology matured, startup advantages shifted from building infrastructure and exploiting platform growth loopholes to great product design, brand connection, and costly marketing, making accelerators and cheap acquisition channels far harder to win.
- Infrastructure is now a service, so founders focus on features and consumer experience rather than deep tech
- Platform arbitrage and early app-store and Facebook/Google growth channels have matured and are closed
- Accelerators are an easy business to start but nearly impossible to run well beyond the top few programs
Consumer startups must reach mass scale and retain users, since a great product guarantees nothing.
- Over the past seven to ten years, technology founders shifted from building infrastructure to building consumer-facing experiences, because the infrastructure layer is now a service they can plug into, letting them focus only on the core features of their product and reach a wider set of customers across healthcare, fintech, consumer tech, and other industries.
“The infrastructure today is a service.”
- From roughly 2007 to 2012, design was a standout differentiator, so a beautifully designed product could win on the App Store or web without a large marketing budget, but after about 2012 a great product became a prerequisite that guarantees nothing about success.
“Starting in 2012, a great product was a prerequisite no matter what.”
- GroupMe was funded in 2009 when the team literally had to invent how to do group text messaging, whereas by the time of the interview group texting was a simple feature many services offered, illustrating how much the cost of building consumer functionality has collapsed.
“Today, it's a lay up, you can do that on many services.”
- Facebook and Twitter raised consumer expectations for software quality so much that workers who used outdated point-of-sale systems at work preferred to play with Facebook at night, forcing enterprise software to catch up to consumer design standards.
“So what happened was the enterprise software needed to catch up to what the consumer software was.”
- Once a user has abandoned an app, the only realistic way to win them back is for about five friends to recommend it to them, making first-session value even more critical.
“The only way you get them back today, if you lost them already, is I think it's like five friends”
- Niche consumer products that are magically fun for a small group do not scale to the 5 to 10 million or more users needed for venture returns, so the challenge is creating a consumer company that can cut through and reach mass scale.
“So the niche-happy products don't scale in terms of venture returns.”
Founders and investors should be judged on vision, leadership, and ability to attract people.
- Classtivity applied to TechStars as a class search engine, and Tisch suggested in response that they should just build the subscription product ClassPass, but the founder rejected this and built a search engine anyway, pivoting about a year later to ClassPass, which then took off.
“Why don't you just do ClassPass?”
- Accelerator applications should elicit a founder's vision and how they think about building a company rather than what they have already done, because venture capital funds the future rather than rewarding past accomplishments, so questions like what the next 10 hires would be are more revealing.
“In venture capital, you don't get money for what you've done.”
- Investors should identify the one or two skills essential to a business and check whether the founding team has them or a credible plan to add them, because a non-technical founding team can be fine for some businesses while a team of technical co-founders may still be wrong for a marketing business.
“what are the core competencies needed.”
- Tisch assessed founding teams by asking about their biggest co-founder fights and who usually wins them, and by having applicants submit a short team video to judge whether chemistry is visible on screen.
“what are the biggest fights that you as co-founders have gotten into?”
- Founders must be able to present themselves on stage and attract people, because they will have to convince investors, customers and especially engineers to bet on an unfinished product, and this matters more because great engineers are scarce.
“as an entrepreneur, you are going to have to be a showman”
- BoxGroup is a four-person seed-stage fund that writes $200,000 to $500,000 checks across 30 to 40 companies a year, aiming to be the first check in and usually investing around company formation.
“we write about a $200,000 to $500,000 check and we do that across 30 to 40 companies a year.”
- BoxGroup does not invest by sector theses but by holding views about the world 10, 15, or 50 years out, and funds companies whose vision plays into that future, since a company should be judged by where it is going rather than by what it has done.
“We don't come in with theses on spaces, we have views of the future”
- To make 30 to 40 investments a year, BoxGroup must see 3,000 to 4,000 companies, and sourcing is the most important part of the job, because missing a deal is a worse outcome than passing on one after seeing it.
“sourcing those 3000 to 4000 deals is the most important part of our job”
- The strongest correlation in BoxGroup's investments is that the longer Tisch has known a founder, the earlier BoxGroup is willing to invest, while strangers with only an idea and no execution history are essentially unfundable today because free tools let anyone get started.
“the longer I've known somebody, the earlier we'll be willing to invest.”
- The most important founder trait is leadership, judged by whether a founder could attract and motivate 100 people at a future all-hands, because investors too often focus on the present rather than whether the founder can scale.
“More now than ever, leadership.”
- Being aggressively fired from a corporate job in 2009 led to six months of depression during which Tisch studied the accelerator model, and he considers that firing the kindest thing that happened to him because it pushed him toward building the New York ecosystem.
“the guy who fired me did the kindest thing anybody could have done for me.”
Consumer tech growth is harder as platform exploits close and customer acquisition costs rise.
- Airbnb's growth by hijacking Craigslist's network was an avenue that is now closed, because the newest scaled companies are either smart enough or protective enough not to open their platforms to such exploitation, and Instagram remains the best platform for building an audience.
“That avenue is gone.”
- Tech companies are structurally underfunded: reaching the American consumer costs on the order of 100 million dollars, and the old benchmark of about 60 million dollars to get 30 percent of the country to hear a name three times has grown, so trying to raise about 4 million dollars to go viral is unrealistic.
“every single tech company is underfunded, truly.”
- There has never been a period on the internet where companies could not hack someone else's scale to their benefit, so the Airbnb-on-Craigslist example is a recurring pattern, although the newest scaled platforms are protective and less likely to open up.
“I don't think you can find a period on the internet where this was true, where there wasn't an ability to sort of hack someone else's scale to your benefit.”
- A feature from Apple's App Store was once enough to make a company win, but that no longer creates a company today because the App Store has matured and cannot be arbitraged the way it was in its early years.
“If Apple features you, I think if you go back like 10 years or 7 years at the beginning of the App Store, if you've got that Apple like hero feature, you won.”
Exits are harder due to incumbents, public markets, and weak acquirer demand.
- Large incumbents like Facebook and Google are now strong and well-positioned to build the future themselves, unlike MySpace or Friendster, so startups can no longer win by building a small feature that disrupts a big company.
“they are positioned to do the future, not to be disrupted.”
- Tesla has far fewer sales than established car companies yet a larger market cap, illustrating the valuation gap that makes non-tech incumbents hesitant to buy tech companies at the prices founders expect.
“Tesla has nowhere near the sales, but is a bigger market cap than the huge car companies.”
- Public markets judge companies on quarter-by-quarter profitability rather than vision, and public-company CEOs are no longer incentivized for 20-year horizons as founders of earlier eras were, so older companies are less willing to acquire risky modern companies and overall risk-taking has declined.
“the public markets are so volatile on a quarter by quarter basis”
- The M&A market is weaker for the current wave of startups because Facebook, Google, and Twitter bought dozens of small companies on their way up, while Airbnb and Uber are not doing so, which makes exits harder for the current cohort.
“Facebook and Google are not buying 50 different cute startups a year like they were”
- Founders should not present an exit slide in a pitch, but investors need some comfort that a plausible set of acquirers exists, since a company whose only exit is becoming a public company is daunting to fund.
“I don't want the founder to put a slide in that says exits”
Venture returns require many bets, tolerance for losses, and accepting regret over passed-on winners.
- Passing on a great company you saw is the omission that costs a fund its returns, and Tisch loses more sleep over companies he passed on, such as Casper, Zynga, Via, and Remind101, than over investments that failed.
“that's the omission that cost you the fund.”
- Making 30 to 40 investments a year requires a willingness to be wrong and lose badly, and venture investors need an unusually high risk tolerance and quick emotional recovery from losses, because big hits outweigh every miss in the portfolio.
“I think I'm in a business of hits, you have to get the big hits”
- Tisch passed on Remind101 during a TechStars batch after getting it through an early round, and it later became the third-ranked free app in the App Store, which he cites as an example of a regretted omission.
“I looked at the App Store today and they're the third ranked free app in the App Store.”
- Tisch is skeptical that he can reliably identify which companies can reach the 'X-bagger' scale that drives venture returns in advance, so his approach relies on a willingness to be wrong and lose heavily while making many bets.
“I don't think I'm smart enough to truly identify that this can or can't be a certain scale.”
- Investor losses are far less personal than founders' losses, so investors should not trivialize failure, and the venture business requires an unusually high risk tolerance coupled with the ability to move on immediately from losses.
“the losses are less personal for us than they are for the founders that we're investing in, they put their life into this.”
New York's industry mix and finance talent make it a fertile place to test consumer technology.
- Large technology companies opening New York offices, starting with Google and later Facebook, Microsoft, eBay, Snapchat, Pinterest, Uber and Square, fed the local startup ecosystem because their well-funded offices recruited talent that then bled down into startups, with ex-Google employees early on being a major source of founders and early hires.
“a lot of it was ex-Google - just people leaving Google would feed the ecosystem.”
- After 2008, New York talent that had depended on finance stopped treating finance as a guaranteed career path and began taking career risk into their own hands, which was the real shift that seeded New York's startup ecosystem and attracted others to do similar things.
“People did not come to New York and view finance as a guarantee.”
- Historically, venture-funded tech developed logically: first infrastructure layers were built, then services connecting that infrastructure to consumer-facing products, and because this required heavy R&D, it was built in Silicon Valley.
“You at first were building an infrastructure layer, and then you were building services to connect the infrastructure layer with other sort of more consumer facing services”
- New York's diversity of industries, from fashion and retail to sports, finance, and healthcare, plus its varied demographics, makes it a fertile ground for testing consumer technology, and this was not feasible 10 to 15 years ago when consumer touchpoints required building too much technology.
“So the complexity of the consumers in New York is a fascinating and fertile ground to test things on.”
Consumer products must deliver core value immediately because users no longer try new apps eagerly.
- Users have stopped eagerly trying new apps, so a new app gets roughly one chance to deliver its core value on first use, and if it fails, the user is effectively lost, meaning iterative lean startup and MVP approaches from 2007 to 2012 are largely irrelevant now.
“A user has no tolerance for mistakes, a user has no tolerance for bad product.”
- A consumer product must deliver value tied to core human drivers such as money, sex, happiness, or relieving boredom, because products that only generate a launch spike, go viral, and die within days are novelties that fail to capture enough of that deep appreciation.
“It's money, it's sex, it's happiness, it's, you know, filling boredom.”
- For the first time in internet history there is no immature customer acquisition channel, because newer platforms like Snapchat do not let other companies grow off them, and Facebook, Google, and the App Store have matured so that paid acquisition costs on Snapchat and Instagram now match Facebook and Google.
“this is the first time in the internet's history that there is not an immature customer acquisition channel.”
- The cheapest way to acquire users is word of mouth, created through virality, stunts, or a magical product, and after that a company must be strong at performance marketing, finding small arbitrage pockets, and social presence.
“The cheapest way to acquire users is word of mouth.”
Brands must differentiate through real meaning, storytelling, and a clear point of view.
- Enterprise buyers receive hundreds of tech pitches a day and archive most of them, so enterprise startups must use storytelling and narrative to rise above the noise or find a sales channel that breaks through.
“It's hundreds, right? It's not even like 10, they are getting inundated.”
- Tech companies that ignored marketing are now at a disadvantage because nearly every television and podcast ad is for a tech company, so they must tell a brand story quickly and efficiently to get people to download apps they have no interest in.
“you have to be great at marketing”
- A useful litmus test for a brand, service, or product is whether anyone would care if it disappeared tomorrow, and by that test long-tail fast food brands without a point of view are at risk while brands like Shake Shack and In-N-Out stand for something.
“is if this thing - a brand, service product, whatever literally just disappeared tomorrow would I care?”
- Brands without meaning or a real connection to customers will die, because consumers spend their money on brands they feel connected to, and Tisch cites a figure that 90 cents of each millennial dollar goes to brands they believe they have a connection with.
“companies without a heart are dead.”
- Long-tail fast-food chains that do not stand for anything are in trouble, while McDonald's remains durable despite declining same-store sales and brands like Shake Shack and In-N-Out mean something to customers.
“If I am Wendy's, or like one of these long tail fast food joints that doesn't stand for something”
- Airlines are protected by monopolies and barriers to entry, but customer resentment toward them will eventually be cut through by a company that reinvents itself with better customer service, and established fast-food brands like Arby's are at risk because they do not stand for anything.
“there will be something that cuts through that noise - it might be a company reinventing who they are and delivering better customer service.”
Software and digital services are reshaping incumbents, banking, transport, and entertainment.
- Every company, including old non-tech companies, must use software to drive its business or it is in trouble, so incumbents will increasingly need to buy modern tech companies, as John Deere did when it bought an AI-driven farming company.
“every company needs to become a software company. If you're not using software to drive your business, you're in trouble.”
- Uber trained society that pressing a button on a phone can trigger a real-world offline action, which opens the door for many other services built on the same pattern of digital requests producing physical outcomes.
“when Uber trains society that you can press a button on your iPhone and something will happen offline, that opens up every other door”
- Technology has not made people's appetite for entertainment any different than before, as the same pattern existed with people staring at newspapers, so the trend of consuming content on phones and podcasts is a natural evolution rather than a health problem.
“I don't think people's sort of voracious appetite for content or entertainment is in any way different.”
- Regulation has forced big banks into homegrown, bank-by-bank technology stacks that limit innovation, which is why money cannot be wired outside business hours, and the speaker expects these constraints to disappear in the near future.
“regulation has forced them to sort of limit their technology innovation inside of each bank”
- With autonomous cars, people will spend their time in vehicles watching videos and playing games rather than talking or sleeping, and this technology will make entertainment more accessible and affordable to a larger group.
“you're going to watch videos, and you're going to play games.”
Accelerators add value mainly through community, mentorship, and deadline-driven pressure.
- Tisch joined TechStars in 2009 after David Cohen asked him why he was not in New York City, and after six months of studying online YC and TechStars he concluded that YC would not franchise into New York and that TechStars' mentor-driven model was worth replicating himself.
“Why aren't you in New York City?”
- Early in the accelerator era it was rare for established entrepreneurs to mentor formally, so each startup could be exposed to around 100 mentors and speakers, but today entrepreneurs are asked to mentor so constantly that attention is much harder to win.
“every company would get exposed to upwards of 100 people”
- The accelerator model that Y Combinator and TechStars pioneered was attractive to Tisch mainly for its community-building effect, the shared learning and shared experiences that make all participants better, rather than for its unit economics, which he considered interesting mainly at scale.
“It's not fascinating because of the economics”
- The 90-day accelerator boot camp was intentionally harsh, with a fixed deadline ending in a Demo Day in front of about 800 people, because the pressure and visible progress of peers push every company to make a big leap forward in that window.
“They had 90 days and they had 90 days to jump.”
Too many accelerators exist, and only the top programs reliably find great companies.
- There are too many accelerators because they are easy to start and fund, but beyond the top programs like YC and TechStars, lower-tier accelerators have almost no chance of finding great companies, so most will wind down after a few batches when nothing works.
“there's zero chance that they're going to get a great company.”
- The accelerator application process is extremely selective, with about 2,000 companies applying to TechStars for 10 to 14 spots per batch, making admission statistically harder than getting into Harvard.
“We used to say it's harder to get into than Harvard.”
- Tisch personally reviewed all roughly 2,000 applications in the first filtering round because the biggest mistakes happen at the first cut, and a missed gem is far harder to recover than one spotted early.
“the biggest mistakes would be made in that first round of cuts.”
- Specialized accelerators in fields like biotech or fintech can be effective because they offer domain-specific networks, such as access to banks, that generalist programs cannot match, though only the best of these succeed.
“there are accelerators that have specialized networks and specialized expertise”
Startups must show a credible monetization path and category-specific structure to be fundable.
- The idea that companies can figure out their business model later is nonsense: founders need not monetize on day one, but they must articulate how and when they will make money, and if they plan to burn money first, they must explain how they will fund reaching scale, potentially needing $500 million to $1 billion.
“I think that's nonsense.”
- Personal budgeting software is easy to build and so 25 to 30 products compete without any winning, whereas tax software is hard to build but its complexity creates defensibility, and the challenge in either case is customer acquisition and a product that keeps users coming back.
“that's why none of those companies have scaled to the level of transformative is because they're easy to do.”
- Fintech must be separated into infrastructure and end-consumer layers, because infrastructure startups need deep relationships with big banks while consumer fintech can plug into existing bank APIs, and regulation has forced banks into homegrown, siloed tech stacks that are now being unwound.
“you have to differentiate between infrastructure layer and the sort of the stuff that touches the end consumer.”
Good investors add value through bespoke, founder-specific help rather than generic advice.
- Y Combinator has scaled since 2009 and now runs a more mentor-driven model through a large surrounding group of non-active alumni and partners, though it is not expressed that way, and its class sizes have grown, while TechStars was a smaller, bespoke operation during Tisch's tenure.
“they have more of a mentor-driven model, it's just not expressed that way.”
- Introductions are a trivial value-add for an investor, taking about eight seconds, so a good investor must instead learn what each founder wants and fit into the specific piece of the founder's puzzle, such as leading a round or being a candid confidant.
“An introduction takes about eight seconds of my day. So, if that's my value add I suck.”
- Generalized startup ideas, advice, and tactics are irrelevant unless they are specific to the company, so a VC or accelerator is in a bespoke services business that must treat each company individually.
“They are irrelevant if they're not specific to your company. So as a VC, as an accelerator, you're in a bespoke services business”
Concerns about screen addiction and market saturation are overstated given enormous opportunity.
- Tisch does not believe history reliably predicts where venture investing is in its cycle, because transformative technology disrupts the course of society, and each scaled company like Uber or Facebook opens new doors for others, so the long tail of opportunities remains enormous even as top companies are entrenched.
“I also don't think history sort of tells the future”
- Concerns about screen addiction and the mining of people's time are overstated because consumers have always wanted entertainment and information, and the change is in the scale, speed, and pace of access rather than in the underlying appetite, with autonomous cars likely to make passengers consume video and games.
“I don't think people's sort of voracious appetite for content or entertainment is in any way different.”