David Tisch on Seed Investing: Ownership Is Only Part of the Equation

1. Guest Background

David Tisch approaches this conversation as an early-stage investor with, by his account, seventeen years of experience. Interviewed by Harry Stebbings on 20VC, he explains a model that the host had repeatedly questioned: accepting relatively small ownership positions across a broad portfolio. Stebbings now acknowledges that ownership requirements led him to reject strong businesses and says he would accept a 1%–2% position in an exceptional opportunity. That admission gives the episode its central tension: which investment rules remain useful when access, potential outcomes, and competition change?

Tisch grounds his answers in BoxGroup's practices. He says the firm seeks to participate at the beginning, has no uniform ownership minimum, takes no board seats, and does not issue competitive term sheets, although it can lead investments. Initial ownership can run from 1% to 10%. These choices describe how his organization tries to work with founders; the interview does not establish them as universally superior terms.

He describes a team of ten investors, including six partners. His personal interests are also explicit: he especially enjoys consumer founders because their end users are so numerous and difficult to understand. Yet he argues for being equally engaged by founders working in science or physical-world models. His consumer perspective helps explain his curiosity without implying an exclusively consumer investment mandate.

2. What the Episode Covers

The conversation moves from interpreting the market to designing an investment organization for it. Stebbings characterizes 2021 as a bubble associated with temporary pandemic behavior rather than fundamental technological innovation. Tisch's response takes a different approach: every period has distinct macroeconomic and company-level conditions, and early investors must judge whether a particular company can transcend them. He describes AI as an early horizontal deployment, with model releases, application revenue growth, and acquisitions producing microcycles roughly every two or three months. That cadence is his interpretation of the market, rather than an independently measured regularity.

BoxGroup's construction gives this argument practical form. Tisch says its $275 million seed fund will invest in approximately 150 companies, at a pace slightly above fifty investments annually. Another $275 million fund makes roughly twenty to thirty substantial follow-on investments, almost all within the existing portfolio, across Series A and B. Cumulative investment in an individual company has reached about $40 million. Broad entry and selective subsequent commitment therefore belong to the same account of the strategy; initial ownership alone does not describe its full capital allocation.

The company examples complicate any expectation of uniform speed. Tisch says partner Claire knew Michael through Dorm Room Fund, enabling BoxGroup to participate in Cursor's earliest round. Asked about realized returns, he reports that the investment returned its fund multiple times. The supplied discussion does not establish the underlying transaction terms, so that account should remain attributed rather than expanded into a reconstructed deal history.

He places slower development beside that exceptional result. Clay, he says, followed a bumpy four-year path involving learning, iteration, and go-to-market development before explosive growth. Baseten needed the inference market to emerge. These stories explain why an early investor might distinguish an unfinished development process from an exhausted opportunity, although they do not provide a general test for telling the two apart.

Rogo illustrates changing investor assessments. Tisch recalls helping it approach approximately forty prospective Series A leads, with thirty-nine passing before Keith led the round. Sequoia and Kleiner subsequently invested. He interprets the episode as a company being misread at one stage, continuing to execute, and eventually attracting a different assessment. Rejection itself is not presented as a reliable investment signal; the relevant element is what the company builds afterward.

The interview also examines follow-on access, liquidity, and the consequences of growing a fund organization. Its closing discussion widens the frame further. Tisch hopes AI's contributions to health, science, and other industries will reach a global population, and argues that abundance should benefit everyone rather than a few. This is an expressed social ambition, extending the conversation beyond whether individual companies or funds become larger.

3. Core Views: Reasoning, Examples, and Limits

Tisch's defense of small ownership starts by enlarging the economic question. Stebbings describes $10 million–$20 million seed rounds for pedigreed founders as increasingly normal. Tisch argues that higher entry costs can coexist with attractive returns if potential outcomes expand and early shareholders retain more of their stakes. His observation is that rapidly growing companies can raise later rounds with 3%–8% dilution, compared with 12%–20% previously. Repeated across several financings, a difference in dilution changes the eventual ownership associated with the same initial investment.

The qualification matters as much as the mechanism. Tisch is discussing companies that grow exceptionally quickly, not claiming that every portfolio holding now receives favorable financing. He also concedes that filling an entire portfolio with $20 million seed rounds at $200 million pre-money valuations would fundamentally change the economics. The argument therefore does not erase price sensitivity. It asks investors to evaluate price together with retained ownership and plausible outcomes, rather than using an initial percentage as a sufficient measure of investment quality.

Access introduces another variable. To acquire 10%–20% of a genuine outlier at seed, an investor must identify the company, defeat capable competitors, and persuade its founders to accept the proposed arrangement. Tisch regards established multistage firms as increasingly effective at seed, with stronger accessibility and more efficient processes. He says competitive rounds may attract ten to fifteen term sheets. In that environment, a concentrated portfolio can reflect the deals a manager could obtain as well as the companies the manager most wanted.

His numerical illustration is deliberately provocative. He personally estimates that concentrated managers may win closer to 20% of their desired deals, imagining 120–150 offers to assemble thirty holdings. The interview supplies no independent dataset supporting that estimate. Its analytical value is the distinction it forces: confidence in the final portfolio cannot establish that it contains the manager's preferred opportunities. Assessing concentration requires examining what was excluded by competition and ownership requirements, not simply listening to explanations of the companies that remained.

The exchange about abundant funding follows a similarly conditional structure. Stebbings worries that rapid additional financing distracts founders. Tisch counters that some teams can convert capital, valuation, and visibility into recruiting advantages, customers, and early partnerships. He attacks lean startup thinking as a general prescription for venture-backed companies, while acknowledging bootstrap exceptions and explicitly separating funding from discipline. The useful reasoning is that capital changes the available course of action. Whether a team can execute that course remains an open question; the interview does not establish that raising more money reliably produces better operating results.

Founder assessment must also account for the time between investment and market exposure. Tisch looks for unusual hiring, access to customers that should be difficult to reach, and the ability to navigate change. Although he observes more technical founders, he asks whether sales and go-to-market capabilities might become differentiators again in two years. If so, he argues, a seed investor should consider those founders now because their companies may reach larger markets one to four years later. This is a conditional timing argument, not a prediction that technical ability will stop mattering.

BoxGroup's internal process expresses his preference for affirmative conviction. Tisch says the firm meets fifty to one hundred companies weekly and discusses investments on Tuesdays and Thursdays without voting. A partner must genuinely want to make an investment; colleagues who have not spent time with the founders cannot casually dismiss it. Ethical and reputational concerns remain legitimate inputs. This may preserve room for nonconsensus judgments, but the interview offers no comparison demonstrating improved selection accuracy. His loose references to 70% or 90% early-stage failure are explicitly uncertain, and he distinguishes seed risk from Series A risk.

The autonomous-driving exchange supplies a final test of evidentiary discipline. Tisch asserts that mandated adoption could save hundreds of thousands of lives, without supplying a calculation. Stebbings challenges readiness in complex cities and difficult weather. Tisch's reference to an 80% reduction in deaths is hypothetical, offered to explain what he would consider sufficiently solved. Relative safety, performance in particular environments, and the case for compulsory adoption remain separate questions. An optimistic answer to one does not establish the others.

4. Learning and Application

For an investment team, the most actionable distinction is between finding opportunities, winning access, and selecting companies. Tisch roughly estimates that BoxGroup sees 50%–75% of what it wants to see and reports win rates of 70%–90% depending on conditions. Those figures should retain their labels as estimates and self-reports. The operational lesson is to record when a relationship began, when a decision was reached, and why an allocation was lost. Late discovery suggests a coverage problem; slow response suggests a process problem. Neither diagnosis automatically remedies poor selection, and Tisch does not claim a reproducible formula for improving investment taste.

Stage selection can likewise begin with the evidence an investor requires. BoxGroup stopped seed-stage brand investing because Tisch wants indications of resonance, actual purchasing, lifetime value, and acquisition economics. A team using similar criteria should ask whether those observations can realistically exist at its proposed entry point. Waiting for momentum or investing at Series A may better match the evidence requirement, with the tradeoff that entry conditions can change. This is a question of strategic fit, not a conclusion that brands cannot become substantial businesses.

Follow-on planning requires separating contractual rights, available allocation, and founder support. Tisch acknowledges that BoxGroup sometimes cannot obtain a larger allocation and sometimes simply misreads a company or reacts too slowly. An initial investment should therefore not be modeled as guaranteeing later access. Before a financing, investors and founders can clarify existing rights and incoming ownership demands. Tisch's illustration of a new investor seeking 22% while existing holders need 6% pro rata shows how contractual commitments can support negotiation. His warning about forcing participation against a founder's wishes concerns economic and reputational consequences; it does not settle the interpretation of any particular contract.

Liquidity decisions benefit from a similar separation of interests. BoxGroup generally avoids early sales because Tisch worries about relationships and believes in late compounding. Yet he thinks personal financial security can help founders take longer-term risks, and employee tenders can support retention. A practical discussion should identify the fund's return needs, the founder's personal pressures, and employees' needs separately before considering timing and scale. These proposed benefits are his judgments, not demonstrated outcomes, and a founder's sale alone does not establish that a fund should reduce its position.

For LPs evaluating a manager, the fourth-fund question turns ambition into diligence. Ask how greater assets would change stages, checks, concentration, and the capabilities needed to operate. Tisch acknowledges BoxGroup's increased scale while maintaining its earliest-stage identity, and recognizes that other organizations can expand differently. The task is to assess consistency between the intended destination and the organization being built, rather than treating size itself as a verdict.

Product and organizational decisions should also start with the relevant constraint. Tisch notes that not every company needs frontier models, while compute access, pricing, energy, and different adoption speeds complicate forecasts. A team can identify what its product actually requires and which improvements it can wait for. Stebbings's AI-first school example raises a related design question: what becomes easier when an organization is built afresh, and what must change when an existing one is adapted? Tisch asks whether both paths can be pursued. Because the discussion supplies no comparative results, it offers a question to investigate, rather than a general justification for abandoning existing institutions.

Source

• Original episode: David Tisch: Why No Manager Wants to Stay Small & Why Competing with Mega Platforms is a Losing Game