Seed-Stage Due Diligence: What We Are Actually Examining
The phrase "due diligence" in venture capital is frequently used as a synonym for the activity that fills the period between a verbal commitment and a signed term sheet. In this usage it implies a checklist: legal review, cap table verification, reference checks, competitive landscape analysis. At the seed stage, this interpretation is incomplete to the point of being misleading. A company that has been operating for six to eighteen months does not have the institutional history, the revenue track record, or the customer relationships that make checklist-style due diligence meaningful. What seed-stage due diligence actually involves is something different — a structured attempt to form a conviction about a set of propositions that cannot, by definition, be verified with historical data because the relevant history does not yet exist.
This piece describes what we are actually examining at KnownWeil Capital when we conduct seed-stage diligence. It is not a comprehensive account of every process step; it is an account of the questions that we believe matter most and the methods we have found most useful for making progress on them. We write it in the hope that it is useful to founders who are preparing for seed diligence conversations, and to our peers in the venture community who are thinking about how to improve their own processes.
The Central Problem: Evaluating Potential Without Evidence
The fundamental challenge of seed-stage investing is that we are being asked to form a conviction about the quality of a company — its team, its product, its market, its business model — at a moment when the most important evidence about each of these dimensions either does not exist or is too early to be reliable. A company with three months of ARR growth does not have a revenue track record; it has a data point. A founding team that has been working together for a year has not yet been tested by the specific pressures that reveal character — the near-death experience, the competitive threat, the enterprise customer who demands customisations that would break the product architecture. A market that appears large in a bottom-up analysis has not yet revealed whether it will actually pay for a new entrant's solution or remain loyal to entrenched incumbents for reasons that are difficult to model in advance.
This means that seed-stage diligence is fundamentally a judgment exercise rather than an evidence evaluation exercise. We are not primarily asking "what does the data show?" because the data is insufficient to show much. We are asking "what would have to be true for this company to succeed, and do we believe those things are true?" This reframing changes both the questions we ask and the evidence we seek.
Diligence Dimension One: The Founding Team's Intellectual Architecture
The first and most important dimension of seed-stage diligence is understanding the quality of the founding team's thinking about their problem. By "intellectual architecture" we mean the coherence, depth, and originality of the mental model that the founders have constructed to explain why their market is the way it is, why existing solutions have failed, and why their approach will succeed where others have not.
We assess this through a combination of structured conversation and stress testing. The structured conversation begins with a request that feels simple but reveals a great deal: we ask the founders to explain their thesis as if we know nothing about the category. The purpose of this request is to observe the order in which founders present information — whether they begin with the problem or the product, whether they reach for market size data before establishing that the problem is real, whether they acknowledge the strongest version of the case against them before explaining why they disagree with it.
We follow this with a series of stress-testing questions designed to probe the assumptions that the thesis depends on. If a founder believes that enterprise buyers in their category will switch from an entrenched incumbent to a new entrant based on a 30% performance improvement, we ask for evidence that this switching threshold is real. If a founder believes that their go-to-market can be primarily product-led despite targeting mid-market buyers with long procurement cycles, we ask for specific evidence from comparable companies that this motion has worked. The purpose of stress-testing is not to be adversarial; it is to understand how founders respond to pressure on the foundations of their thesis. Founders who have genuinely built their thesis from first principles tend to engage with stress-testing productively — they acknowledge which assumptions are contested, explain what evidence they have or intend to gather, and are comfortable sitting with uncertainty on questions that cannot yet be resolved. Founders who have built their thesis primarily from a narrative template tend to become defensive when the specific empirical foundations of the narrative are challenged.
We also invest significant time in understanding the founders' prior intellectual history: the jobs they had before the company, the technical problems they worked on, the domain expertise they developed, the specific observations that led them to their current thesis. A founder whose thesis emerged from three years of domain immersion is different from a founder whose thesis emerged from a market mapping exercise. Neither is automatically superior, but the evidence that supports conviction is different in each case and we evaluate accordingly.
Diligence Dimension Two: Early Customer Evidence
The second most important dimension of seed diligence is the quality of early customer evidence. At the seed stage, "customer evidence" does not necessarily mean revenue. It means systematic evidence of a pattern in how potential buyers respond to the product, the pricing, and the value proposition — evidence that points toward a repeatable go-to-market motion even if that motion has not yet been fully executed.
We look at several types of early customer evidence. The first is customer conversations — not surveys or aggregate NPS scores, but the detailed qualitative accounts of specific conversations with specific buyers. We ask founders to walk us through their last ten customer conversations in detail: who was the buyer, what problem were they describing, how did they describe their current solution, what specifically excited them about the product, what were their objections, what happened at the end of the conversation. The richness of this account tells us a great deal about the quality of the founders' customer discovery process. Founders who can give detailed, specific, differentiated accounts of individual buyer conversations are doing real customer discovery. Founders who summarise these conversations in aggregate terms — "customers generally respond well to X" — are often abstracting over a much more ambiguous underlying reality.
The second type of evidence is usage behaviour. For companies with a product in market, we look at usage data with a specific question: is anyone using this product in a way that reveals genuine value, not merely curiosity? The distinction matters. A product that has many trial signups but low conversion to regular use is showing something different from a product that has fewer signups but a high proportion of users who return daily. We look for evidence of what Ben Horowitz called "the product crack" — the specific user behaviour that reveals that at least some subset of early users have found the product genuinely valuable in their work.
For companies at the pre-revenue seed stage — a category we invest in selectively — we substitute reference customer conversations for revenue evidence. We ask to speak with the five to ten people who are most familiar with the product and the problem it addresses, and we conduct those conversations ourselves. The signal we are looking for is not enthusiasm; early users of interesting products are often enthusiastic. The signal we are looking for is specificity: can the reference customer describe, in concrete terms, the workflow change that the product would enable, and can they articulate a plausible budget owner and procurement pathway for the product in their organisation?
"Seed diligence is not about verifying what is true. It is about forming a conviction about what might become true, and understanding the specific conditions under which it will."
Diligence Dimension Three: Technical and Product Architecture
At KnownWeil Capital, we invest primarily in B2B technology companies — enterprise software, developer tools, applied AI. This means that technical diligence is a meaningful component of our process. We are not looking to certify the quality of the code; that is not a productive use of pre-investment time. We are looking to understand whether the technical architecture of the product is consistent with the long-term product vision, and whether it is designed in a way that creates genuine differentiation or merely replicates existing approaches with different branding.
The most important technical question we ask at the seed stage is: what is the hardest technical problem this company needs to solve, and is the founding team capable of solving it? For companies like Trigger.dev — which raised a $1.7M seed to build durable background job infrastructure for JavaScript developers — the hardest technical problem was the correctness and reliability of job execution in distributed environments, a problem that is non-trivial and where the founding team's prior experience with distributed systems was directly relevant. For companies in the enterprise AI space, the hardest technical problem is often not model capability but the integration architecture that connects AI capabilities to enterprise data sources in a way that is secure, reliable, and maintainable at scale.
We also pay close attention to the architectural decisions that have already been made and what they reveal about the founders' technical philosophy. Companies that have built on clean abstractions, with thoughtful separation of concerns and explicit interfaces between components, tend to be companies where the founders have the disciplined technical thinking that scales to larger engineering organisations. Companies that have accumulated significant technical debt in the early prototype — justified by speed — tend to be companies that will face a painful refactoring period precisely when they can least afford to slow down engineering velocity.
This is not a rigid criterion. Some technical debt is appropriate in early products, and the ability to ship quickly is itself evidence of valuable technical judgment. What we are looking for is founders who are conscious of the trade-offs they are making — who can articulate which corners they cut intentionally, why, and what the plan is for addressing them — rather than founders who have accumulated technical debt as an unintended consequence of undisciplined engineering practices.
Diligence Dimension Four: Competitive and Market Structure
The competitive analysis component of seed diligence is the area where the gap between investor practice and founder experience is widest. Founders often experience competitive analysis as an adversarial exercise — an attempt to find a competitor that invalidates their thesis. Investors who conduct competitive analysis poorly do indeed use it this way, identifying incumbent players or well-funded competitors as a reason to decline investments that would have been worth making. The most valuable competitive analysis at the seed stage is not about identifying risks; it is about understanding market structure in a way that reveals whether the founding thesis is well-targeted.
We approach competitive analysis by first mapping what we call the "solution landscape" — the full set of approaches that potential buyers are currently using to address the problem the company is solving, including non-software approaches. For a company building enterprise contract management software, the solution landscape includes not just contract management software vendors like DocuSign, Ironclad, and Juro, but also the Excel spreadsheets and shared Drive folders that a significant proportion of enterprise legal teams actually use for contract tracking. Understanding this full landscape is important because it reveals where the real switching cost lies and what the real alternative cost is that the company's pricing must be measured against.
We then try to understand what specifically makes the founding team's approach different from every other approach in the landscape, and why that difference is both real and durable. Many companies describe their differentiation in terms of features — they have Feature X that competitors lack. We are more interested in differentiation that is structural: an architectural decision that allows them to solve a problem that competitors cannot easily replicate, a data advantage that compounds over time, a go-to-market motion that reaches buyers that competitors are structurally unable to reach efficiently.
Mintlify, which raised $18.5M and built a documentation platform that developers actually want to use, is a useful example. Mintlify's differentiation from incumbents like ReadTheDocs and GitBook is partly feature-based — it has better design, better MDX support, better analytics — but it is also structural. Mintlify was built by developers for developers, with an understanding of what technical writers and engineering teams actually care about in documentation tools that traditional documentation vendors — built primarily around marketing and product teams — lack. This structural understanding of the buyer created a differentiation that was harder for incumbents to replicate quickly than any individual feature.
Diligence Dimension Five: Reference Checks and Team Quality
Reference checks are the most consistently underinvested component of seed-stage diligence in our experience of the venture industry. Many investors conduct reference checks in a way that is designed primarily to confirm the positive impression formed in direct meetings — they ask questions that invite positive framing, they speak to references selected by the founders themselves without supplementing them with independently identified references, and they weight glowing recommendations heavily without calibrating for the relationship between the reference and the founder.
Our approach to reference checks is structured around the specific behavioural patterns we believe are most predictive of founder success. We ask a standard set of questions across all references, and we ask them in a way that elicits specific, behavioural evidence rather than general assessments of character. We do not ask "Is X a good leader?" We ask "Can you give me a specific example of a situation where X had to make a decision under significant uncertainty with incomplete information, and describe how they approached it and what the outcome was?" The specificity of this question makes it much harder to give a pro forma positive answer, and the responses reveal a great deal about the founder's actual decision-making style.
We also conduct our most important reference conversations with people who have observed the founder under pressure: managers who supervised them during difficult periods, co-founders or early employees from previous companies who experienced the pressures of the early startup phase alongside them, and customers from previous ventures who observed how the founder responded to product problems or service failures. These conversations, conducted with sufficient depth and without the founder present, are among the most valuable inputs to our final conviction-building process.
The Investment Memo and Final Conviction
At the conclusion of our diligence process, we write an investment memo. The memo is not a summary of our findings; it is a structured argument for why we should or should not make the investment. It articulates the specific thesis that would need to be true for the investment to succeed, the evidence that supports each component of that thesis, the evidence that challenges it, the specific risk factors that we believe are most likely to prevent the thesis from playing out, and the specific milestones that would increase or decrease our conviction over the next twelve to eighteen months.
The memo is also a commitment device. By writing explicitly about our risks and our milestones, we create a record against which our future judgment can be measured. Investments that play out differently from the memo's predictions are learning opportunities — not just about the company, but about our own pattern recognition and our own analytical blind spots. Over time, the corpus of our investment memos is one of the most valuable sources of institutional learning we have, because it documents not just the outcomes of our decisions but the reasoning process that produced them.
Depot, which raised $6M to build Docker build infrastructure that is significantly faster than the Docker daemon, and Inngest, which built durable execution infrastructure for modern JavaScript applications, are both companies where the diligence process confirmed specific architectural insights about developer infrastructure markets that we had been building toward through earlier investments. The ability to say, at the point of conviction, "this company fits a pattern we understand deeply" — and to have the documented evidence to justify that claim — is the goal of a diligence process that is designed not just to prevent bad investments but to build institutional knowledge over time.
The ideal outcome of seed diligence is not certainty — certainty is not available at the seed stage. The ideal outcome is a calibrated conviction: a specific, documented belief about what must be true for the investment to succeed, grounded in the best evidence available at the time of the decision, and held with appropriate humility about the limits of that evidence. This is what we are actually examining when we conduct seed-stage due diligence. The checklist is the least important part.