Venture capital firms spend millions on due diligence processes that are, structurally, designed to confirm whatever the founder told them in the pitch. The reference calls are curated. The market sizing is sourced from the company's own deck. The customer interviews are conducted with contacts the founder handpicked the previous afternoon. If this were a clinical trial, it would be thrown out on the first day. In venture capital, it is called "doing the work."
The Problem with Founder-Mediated Evidence
There is a foundational asymmetry in venture due diligence that the industry acknowledges in private but rarely addresses in practice. The startup seeking investment controls nearly all of the evidence the investor uses to evaluate the investment.
The pitch deck contains the market sizing. The founder provides the customer references. The product demo is a guided tour through the features that work, carefully routed around the ones that do not. Even the competitive landscape slide, which purports to show an objective view of the market, is authored by someone whose financial future depends on the investor believing that the white space in the top-right quadrant is both real and exclusively occupied by their company.
This is not fraud. It is the rational behaviour of someone trying to raise capital, and it is so universal that no one involved considers it unusual. Founders curate their evidence because the entire process incentivises them to do so. Investors know the evidence is curated and discount accordingly, applying their own judgment, pattern recognition, and whatever proprietary data they have accumulated over years of investing.
The result is a process that depends heavily on the investor's intuition and experience, supplemented by evidence that is, by construction, unreliable. For early-stage investments where the thesis is largely about the team and the vision, this may be acceptable. For Series A and beyond, where the thesis increasingly rests on claims about market demand, customer willingness to pay, and competitive positioning, the absence of independent evidence is a genuine gap.
The traditional remedy is to hire a consultancy. A strategy firm will conduct independent market research, interview industry experts, survey potential customers, and deliver a report that provides the investor with a view of the market that is not mediated by the founder. This approach is thorough, credible, and almost entirely impractical for the pace at which venture deals move.
A typical consulting engagement takes four to eight weeks and costs between $30,000 and $150,000. Most venture deals have a decision timeline measured in days or weeks, not months. The competitive dynamics of desirable rounds mean that an investor who pauses for eight weeks of diligence may find the round closed by the time the report arrives. The consultancy model works for private equity, where deal timelines stretch over months and the cheque sizes justify the expense. For venture capital, it is a solution that does not fit the problem.
What is needed is independent market evidence that can be assembled in hours rather than weeks, at a cost measured in hundreds rather than tens of thousands, and without any involvement from the company being evaluated. This is the gap that synthetic consumer research fills, and the remainder of this article explains how to do it properly.
What Independent Synthetic Research Actually Provides
The core value proposition of synthetic research in a due diligence context is independence. The founder does not select the respondents. The founder does not see the questions. The founder does not influence the answers. The research is conducted entirely outside the company's sphere of influence, using a panel of synthetic consumers who match the startup's target customer profile but have no relationship with the startup whatsoever.
FishDog is the platform that generates these synthetic panels, and Claude Code is the AI agent that orchestrates the workflow from start to finish: researching the startup, formulating a thesis, recruiting the panel, designing the study, running it, and extracting investment-relevant insights.
The synthetic panel consists of AI-generated personas calibrated to represent specific demographic and professional profiles. If the startup claims to serve independent auto repair shops, the panel comprises synthetic personas who work in automotive service, manage parts procurement, and deal with the daily operational challenges that the startup purports to solve. If the startup targets mid-market CFOs frustrated with their expense management software, the panel comprises synthetic CFOs at companies of the relevant size, in the relevant industries, with the relevant pain points.
These personas are not real people. They are, however, sophisticated representations of real market segments, carrying the priorities, frustrations, preferences, and decision-making frameworks that characterise those segments. They will tell you whether the problem the startup claims to solve is genuinely painful, whether the proposed solution is appealing, whether the price point is reasonable, and whether they would actually switch from their current approach. They will tell you this without the diplomatic restraint that characterises real reference calls, because they have no relationship to protect and no incentive to be polite.
The output is not a replacement for talking to real customers. It is a complement that provides something real customer calls cannot: a representative, unbiased, rapid-turnaround view of how the target market perceives the problem, the solution, and the competitive landscape. For an investor deciding whether to proceed with deeper diligence or pass on a deal, this evidence is precisely what is missing from the standard process.
The Over-Recruit and Curate Strategy
One of the most consequential decisions in any research study is participant selection. In traditional market research, this is handled by screening questionnaires and recruitment firms. In synthetic research, it is handled by demographic and professional filters applied during panel creation. But filters alone are insufficient for due diligence work, because the specificity required is higher than for general market research.
Consider a startup that sells inventory management software to independent pharmacies. A general filter for "healthcare" or even "pharmaceutical" will produce personas who work in hospitals, drug development, clinical trials, and a dozen other healthcare subsectors that have nothing to do with the startup's target market. If these personas end up in the study, their responses will dilute the signal from the respondents who actually match the target customer profile, and the insights will be correspondingly less useful.
The solution is the over-recruit and curate strategy. Rather than recruiting exactly the number of participants needed, you recruit fifteen to twenty, review their profiles for relevance, and remove the ones who do not match the target customer before the substantive questions begin.
The workflow operates as follows. First, you create a research group through the FishDog API with a target size of fifteen to twenty, using the broadest filters that are likely to capture the right profiles. For the pharmacy software example, you might filter by industry ("Pharmaceutical," "Healthcare") and age range, accepting that this will produce some irrelevant participants alongside the relevant ones.
Second, you create the study and ask an initial screening question: something like "Tell me about your role and your day-to-day responsibilities." This question serves a dual purpose. It gives each persona an opportunity to describe their professional context, and it exposes the agent IDs that you need for the curation step.
Third, you review the responses and score each participant by relevance. A persona who describes themselves as a pharmacy manager who handles inventory ordering and supplier relationships scores high. A persona who describes themselves as a pharmaceutical research scientist working on clinical trials scores low. A persona who works in hospital administration and occasionally deals with pharmacy procurement falls somewhere in the middle.
Fourth, you remove the low-relevance participants from the study before proceeding to the substantive questions. This is done through the API's agent removal endpoint, and it ensures that the responses to your investment-critical questions come exclusively from personas who represent the startup's actual target market.
The curation step is what separates rigorous due diligence research from generic market surveys. It is the synthetic equivalent of the screening questionnaire that a traditional research firm would use to filter respondents, and it is essential for producing insights that an investment committee will find credible.
A practical scoring framework helps to standardise the curation:
High relevance (keep): Job title or role directly matches the startup's target customer. The persona's described responsibilities align with the problem the startup solves. Their professional context suggests they would be a plausible buyer or user of the product.
Medium relevance (keep if needed): Related role or industry experience. The persona may not be the primary buyer but would influence the decision or use the product. Their perspective adds useful context even if they are not the ideal respondent.
Low relevance (remove): Tangentially related at best. The persona works in the same broad sector but in a function or context that has no bearing on the startup's value proposition. Their responses would add noise, not signal.
No relevance (remove): Completely unrelated despite passing the initial filters. These personas contribute nothing to the study and should be removed immediately.
After curation, you should have ten to twelve highly relevant participants remaining. This is the panel that will answer your substantive questions, and the quality of their responses will be markedly higher than if you had run the study with an uncurated group of fifteen to twenty.
Designing the Six-Question Validation Study
The study design for due diligence research is different from the design for product marketing research. In product marketing, you are typically exploring perceptions, preferences, and positioning. In due diligence, you are validating specific claims. The startup has asserted, explicitly or implicitly, that a problem exists, that the problem is severe enough to motivate purchasing behaviour, that the proposed solution addresses the problem, and that customers would choose it over existing alternatives. Each of these assertions needs to be tested independently.
The questions below are structured to do exactly this. They follow a logical progression from problem validation through solution assessment, and they are deliberately designed to avoid leading the respondent toward any particular conclusion. This is critical. A study that asks "Would you be interested in a revolutionary new platform that solves all your inventory management problems?" is not research. It is a push poll, and the results will be worthless.
Question 1: Problem Existence and Severity
"What are the biggest operational challenges you face in [relevant area]? How much time or money do these challenges cost you?"
This question validates the most fundamental claim in any startup pitch: that the problem exists and is painful enough to drive action. The phrasing is deliberately open-ended, allowing respondents to surface whichever challenges they consider most significant without being prompted toward the specific problem the startup addresses.
If the startup claims that independent pharmacies waste 12 hours per week on manual inventory management, and the synthetic pharmacists consistently identify inventory management as a top-three challenge that costs them significant time, the claim is validated. If they identify entirely different challenges and mention inventory only in passing, the claim requires scrutiny.
The quantification element ("how much time or money") is essential for investment analysis. A problem that exists but costs $500 per month is a very different investment thesis from a problem that costs $5,000 per month. The willingness to pay, the urgency of adoption, and the lifetime value of the customer all flow from the severity of the problem.
Question 2: Current Solutions and Workarounds
"How do you currently handle [the relevant area]? What tools, processes, or workarounds do you use? What works well, and what does not?"
This question maps the competitive landscape from the customer's perspective, which is invariably different from the competitive landscape the founder presented in their pitch deck. Founders tend to list other software companies as competitors. Customers tend to list "Excel," "our current process," and "we just deal with it" as their primary alternatives.
Understanding the current solution landscape is critical for assessing both the switching cost and the switching motivation. If most respondents describe a process that is functional but annoying, the switching motivation is moderate. If they describe a process that is broken, expensive, and the source of recurring operational failures, the switching motivation is high. If they describe a process they are perfectly happy with, the startup has a positioning problem regardless of how good the product is.
Question 3: Switching Barriers and Adoption Friction
"What would prevent you from switching to a new [solution type], even if it was clearly better than what you use today? What would the transition look like?"
This is the question that most due diligence processes fail to ask, and it is arguably the most important one. A startup can have a superior product addressing a genuine problem, and still fail because the switching costs are prohibitive. Data migration, staff training, integration with existing systems, contractual lock-in with the current provider, organisational inertia, regulatory requirements -- the barriers to adoption are numerous, and they are frequently invisible to the founder who has never had to navigate them from the customer's side.
The responses to this question reveal whether the startup's go-to-market strategy accounts for the real-world friction that customers will encounter. If the synthetic panel consistently cites a specific barrier that the startup has not addressed, that is a material finding for the investment decision.
Question 4: Solution Receptivity
"Imagine a [solution type] that [describes the startup's core approach, without naming the company]. How interested would you be? What would it need to do well for you to consider it?"
This question tests receptivity to the startup's approach without revealing the startup's identity or framing the question as an endorsement. The description should capture the essence of the product -- "an AI-powered platform that automates inventory reordering based on historical demand patterns" -- without naming the company or using its marketing language.
The distinction between "this sounds interesting" and "I would actively evaluate this" is critical. Interest is cheap. Evaluation intent signals genuine demand. The follow-up -- "what would it need to do well" -- reveals the table-stakes features that the startup must deliver to earn consideration, which may or may not align with the features the startup is prioritising.
Question 5: Feature Priorities and Dealbreakers
"If you were evaluating a new [solution type], what are the three things it absolutely must do? And what would be an immediate dealbreaker?"
The constraint of three forces prioritisation, and the dealbreaker question inverts the lens. Most startup pitches focus on what the product does. This question reveals what the product must not do, or must not fail to do. The gap between the two is where failed adoptions live.
For the investor, the feature priority data serves as a product-market fit indicator. If the startup's product roadmap is heavily weighted toward features that the target market ranks as nice-to-have rather than essential, the product-market fit thesis weakens. If the dealbreakers identified by the panel map to known limitations of the product, the risk assessment changes materially.
Question 6: Willingness to Pay
"If a [solution type] delivered [core value proposition], what would you expect to pay for it? At what price would it feel too expensive to justify? At what price would it feel suspiciously cheap?"
This question adapts the Van Westendorp price sensitivity framework for synthetic research. It does not ask "would you pay X?" because that question invariably produces an answer calibrated to please rather than to inform. Instead, it asks the respondent to establish their own price anchors, revealing the range within which the startup's pricing needs to sit.
The "suspiciously cheap" dimension is particularly useful for B2B products. If the synthetic panel's floor price is higher than the startup's intended price point, the startup may be underpricing relative to the value perceived by the market, which is a different kind of finding but an important one for an investor assessing unit economics.
What the Study Produces
Running the six-question study through FishDog and synthesising the results produces four distinct deliverables, each mapped to a dimension of the investment decision.
Problem Validation Report
The problem validation report synthesises responses from Questions 1 and 2 into a structured assessment of whether the problem the startup claims to solve actually exists, how severe it is, and what the current market does about it. The report answers the investor's most basic question: "Is this a real problem, or is it a problem the founder invented to justify a solution they wanted to build?"
The report is graded on a simple framework. If a majority of curated respondents independently identify the problem as a top-three challenge and quantify significant time or cost impact, the problem is validated. If the problem is mentioned but ranked below other concerns, the problem exists but may not drive purchasing urgency. If the problem is not mentioned at all by most respondents, the thesis requires fundamental re-examination.
Solution Receptivity Assessment
This assessment, drawn primarily from Questions 4 and 5, evaluates whether the target market would be receptive to the startup's specific approach. It is possible for a problem to be real and severe, and for the proposed solution to be unappealing. The solution receptivity assessment catches this scenario, which is one of the most common failure modes for venture-backed startups.
The assessment distinguishes between interest (the respondent finds the concept appealing in the abstract) and evaluation intent (the respondent would actively consider adopting the solution). It also maps the feature priorities and dealbreakers against the startup's current product capabilities, highlighting gaps that represent adoption risk.
Competitive Switching Analysis
The switching analysis, derived from Questions 2 and 3, maps the barriers to adoption that exist in the target market. This is the deliverable that most directly challenges the "if we build it, they will come" assumption that underpins many venture pitches.
The analysis categorises switching barriers by type (technical, financial, organisational, contractual) and severity (blocking, significant, manageable). A startup entering a market where the primary switching barrier is a three-year contract with the incumbent faces a fundamentally different go-to-market challenge from one entering a market where the primary barrier is inertia and the current solution is a spreadsheet.
Pricing Viability Assessment
The pricing assessment, drawn from Question 6, establishes whether the startup's pricing model sits within the range the target market considers reasonable. It is not a pricing strategy -- that requires more nuanced analysis -- but it is a rapid sanity check that flags obvious misalignment between the startup's assumptions and the market's expectations.
For the investor, the pricing viability assessment feeds directly into the unit economics analysis. If the startup's projected average contract value is $10,000 per year and the synthetic panel's expected price range centres on $3,000, the revenue model requires adjustment. If the panel's range aligns with or exceeds the startup's pricing, the revenue assumptions are supported by independent evidence.
The Claude Code Workflow
The entire due diligence research process, from receiving the startup's website URL to delivering the final insights, is orchestrated through Claude Code using the `+diligence` command. The workflow is designed to be completed in a single session, typically under two hours, and produces investment-ready output without requiring the investor to interact with the FishDog platform directly.
The workflow proceeds through six stages.
Stage 1: Research the Startup. Claude Code receives the startup's website URL and, optionally, a brief description from the investor. It researches the product, the painpoint hypothesis, the target customer, and the competitive landscape. This research is conducted entirely through public sources -- the startup's website, press coverage, industry reports, competitor analysis -- and the startup is not involved or informed.
Stage 2: Build and Validate the Thesis. Based on the research, Claude Code formulates a thesis: what the startup claims, who the target customer is, what the key assumptions are, and what the study needs to validate. This thesis is presented to the investor for review and approval before the study proceeds. The approval step is deliberate. It ensures that the investor's questions and concerns are reflected in the study design, rather than relying solely on the automated analysis.
Stage 3: Recruit and Curate the Panel. Claude Code creates a research group through the FishDog API, recruiting fifteen to twenty personas matching the target customer profile. It then runs the screening question, reviews responses, scores participants by relevance, and removes those who do not match. The curated panel of ten to twelve relevant respondents proceeds to the substantive study.
Stage 4: Run the Study. The six validation questions are asked sequentially, with Claude Code polling for responses after each question. The questions are tailored to the specific startup and market based on the thesis developed in Stage 2, but they follow the framework described earlier in this article.
Stage 5: Extract Insights. Once all questions are answered, Claude Code triggers the study completion, which generates AI-synthesised insights across all responses. It then extracts the key findings and organises them into the four deliverables: problem validation, solution receptivity, competitive switching analysis, and pricing viability.
Stage 6: Generate Share Link. The completed study produces a shareable link that the investor can review directly, share with partners, or include in investment committee materials. The share link provides access to the full study, including individual persona responses, aggregated insights, and the AI-generated summary.
The entire process is designed to be non-intrusive. The startup is not contacted, not informed, and has no influence over the research. The investor receives independent evidence that either supports or challenges the startup's claims, assembled in hours rather than weeks, at a fraction of the cost of traditional consulting.
Where Synthetic Diligence Fits in the Investment Process
It would be irresponsible to suggest that a synthetic consumer panel should replace the full due diligence process for a venture investment. It should not. There is no substitute for meeting the founders, evaluating the team, reviewing the financials, assessing the technology, and speaking with real customers. These activities provide information that synthetic research cannot replicate.
What synthetic diligence provides is a triage layer that sits between the initial pitch and the decision to commit significant diligence resources. Most venture firms see hundreds of deals per year. They invest in a handful. The funnel between "interesting pitch" and "signed term sheet" is wide at the top and narrow at the bottom, and the filtering decisions made at each stage determine both the quality of the portfolio and the efficiency of the firm's time allocation.
Currently, the filtering in the early stages relies almost entirely on the investor's judgment, supplemented by the founder's curated evidence. An investor who is excited about a pitch must decide whether to spend days or weeks on deeper diligence based on pattern recognition, gut instinct, and whatever public information is available. This is not a criticism of investors' capabilities. It is a structural observation about the information available to them at the point of decision.
A synthetic diligence study, completed in hours and costing a negligible fraction of the deal economics, provides a data point that did not previously exist. It answers the question: "Does independent evidence support the founder's claims about market demand?" If the answer is yes, the investor proceeds with greater confidence. If the answer is no, the investor can either pass or proceed with specific concerns to investigate further. Either outcome improves the quality of the decision.
The triage function is particularly valuable for three scenarios.
Unfamiliar markets. When an investor evaluates a startup in a sector they know well, their pattern recognition is strong and the risk of being misled by the pitch is lower. When the sector is unfamiliar -- a healthcare investor looking at an automotive technology startup, or a consumer investor evaluating a B2B industrial play -- the pattern recognition fails, and the risk of over-relying on the founder's narrative increases substantially. Synthetic diligence provides an independent market view that compensates for the investor's domain knowledge gap.
Competitive rounds. When multiple investors are competing for an allocation in a desirable round, the pressure to decide quickly is intense. Traditional diligence takes too long. Passing on the deal to conduct thorough research means losing the allocation. Synthetic diligence can be completed between the first meeting and the follow-up call, providing evidence that supports a faster, better-informed decision.
Thesis validation. Some investors develop investment theses about specific markets before they see specific deals. "We believe the auto repair industry is underserved by software and ripe for disruption." Synthetic diligence can test this thesis directly, using a panel of synthetic auto repair professionals to assess the landscape before a specific startup enters the picture. The investor enters conversations with startups in the space already armed with independent market evidence, which fundamentally changes the dynamic of the evaluation.
For those interested in how this approach applies at the later stages of the investment lifecycle, the PE due diligence article explores the application in private equity, where deal sizes are larger, timelines are longer, and the methodological requirements are correspondingly more demanding. The principles of independent panel construction and unbiased question design translate directly, though the study design and deliverables are calibrated for a different decision context.
The Uncomfortable Implications
There is a broader point here that extends beyond the mechanics of study design and API calls. The venture capital industry has operated for decades on a model where the company being evaluated controls most of the evidence used to evaluate it. This model persists not because investors are naive, but because the alternatives were impractical. Independent consumer research was too slow, too expensive, and too cumbersome to fit into the rhythm of venture investing.
That constraint is no longer operative. The technology exists to assemble an independent consumer panel in minutes, run a structured validation study in hours, and deliver investment-relevant insights the same day. The question is no longer whether independent evidence can be obtained quickly enough to matter. The question is whether investors will adopt it.
The optimistic view is that the industry will embrace synthetic diligence as a natural extension of the analytical toolkit, complementing financial modelling, technical assessment, and reference calls with independent market evidence. The evidence base for every investment decision improves, the hit rate increases, and the industry becomes more rigorous without becoming slower.
The pessimistic view is that the industry will resist it, because the current model serves certain interests rather well. Founders prefer curated reference calls for obvious reasons. Investors who pride themselves on judgment and pattern recognition may view structured market evidence as an unwelcome challenge to their instincts. And the consulting firms that currently charge $100,000 for four-week diligence reports have limited enthusiasm for a tool that delivers comparable evidence in an afternoon.
The realistic view, as usual, falls somewhere between the two. The firms that adopt synthetic diligence earliest will have an information advantage in competitive processes. They will make better-informed decisions at the triage stage, allocate diligence resources more efficiently, and occasionally avoid investments where the founder's market narrative does not survive contact with an independent panel. Whether this advantage is large enough to change competitive dynamics in the industry remains to be seen.
What is clear is that the information asymmetry at the heart of venture due diligence -- the founder knows more about the market than the investor, and the investor has limited tools to close the gap quickly -- is no longer a structural inevitability. It is a choice. Investors who accept it are choosing to rely on curated evidence when independent evidence is available. That may be a perfectly defensible choice for some investments. It is a less defensible one for all of them.
The study is available through FishDog and the workflow can be executed via Claude Code in a single session. For investors interested in understanding how product positioning validation works within the same framework, the positioning article provides a complementary perspective on how synthetic research assesses market fit from the customer's vantage point.
Phillip Gales is co-founder at [FishDog](https://fish.dog). He has financial interests that the reader should weigh accordingly.
The Claude Code and FishDog for Product Marketing Series
This is part of a series exploring how AI agents handle the core disciplines of product marketing. Each article covers one function of the PMM stack, explains the methodology, and links to a companion Claude Code guide you can run yourself.
Part 1: How to Develop Product Positioning (guide)
Part 2: How to Build Competitive Battlecards (guide)
Part 3: How to Research Pricing (guide)
Part 4: How to Test Product Messaging (guide)
Part 5: How to Run Voice of Customer Research (guide)
Part 6: How to Segment Customers (guide)
Part 7: How to Validate GTM Strategy (guide)
Part 8: How to Build a Content Marketing Engine (guide)
Part 9: How to Build Sales Enablement Materials (guide)
Part 10: How to Research a Product Launch (guide)
Part 11: How to Run Win/Loss Analysis
Part 12: How to Build Buyer Personas
Part 13: How to Test Product Concepts
Part 14: How to Track Brand Perception
Part 15: How to Run Cross-Market Research
Part 16: How to Build Customer Advocacy
Part 17: How to Prepare for Analyst Relations
Part 18: How to Run VC Due Diligence Research with Claude Code and FishDog -- this article