Market Research Methods Every VC Analyst Should Rethink
Summary
Market research methods in VC are often sequenced wrong: secondary data arrives last, primary research is used for validation rather than filtering, and the methods that feel rigorous rarely build conviction. This guide covers what secondary sources are worth running, how to map a sector before seeing individual companies, and why AI-assisted screening changes the order of operations without replacing judgment.
The first screen takes four minutes. The eighty-first takes two weeks. The problem is not the volume of dealflow. It is that most market research methods in early-stage VC are built for the wrong moment. They assume you have time, a research team, and a market that behaves like a structured case study.
Most European funds have two analysts, one associate, and a calendar full of calls. Market research methods need to fit that reality. What follows is how to research a deal fast enough to matter, and what to stop doing entirely.
Secondary research: what is worth your time and what is not
Secondary research means data someone else collected. The question is which data changes your conviction before a first meeting.
Worth running:
Sector activity on Dealroom or PitchBook: deal count, median valuations, and investor activity by geography over the last 18 months. EU-specific datasets differ meaningfully from US-centric alternatives, and for DACH or Swiss Seed deals, the delta matters.
Funding history of the company and its direct competitors: not to validate the market size, but to understand who else has looked at this space and passed. A competitor that raised at Seed and went quiet eighteen months later is more useful information than a market sizing slide.
Job board signals: headcount growth, engineering-to-sales ratio, job titles appearing in new geographies. These are harder to inflate than revenue projections and often more current than anything in a database.
Founder trajectory on LinkedIn: not the polished version in the deck, but the chronological gaps, the quieter stints between named roles, the reference-check-adjacent details visible in the timeline.
Not worth your time:
Generic TAM/SAM/SOM sizing from Gartner, IDC, or Statista. The analyst who presents an EUR 87 billion total addressable market has told you nothing about whether this specific company can hold its segment. You will triangulate market size better in thirty minutes of primary research than in a three-hundred-page industry report.
Attending every conference as a sourcing method. Go to conferences for relationship maintenance. The founders who pitch you in a hallway at an event are, in most cases, the ones who have not already been seen by the funds you are competing against. That is both a selection effect and a warning sign.
Company press releases as primary data. They exist to shape perception. Read them once to understand how the company wants to be understood. Stop there.
Thesis-driven mapping: the step most small funds skip
The funds that generate coverage before conviction run a sector map before they see individual companies. The map asks three questions: who are all the companies in this space, what stage are they at, and which ones have the founder profile worth investigating?
This is not a complex exercise. It is time-consuming, which is why a three-person fund typically skips it and relies on inbound flow.
A usable sector map has four columns: company name, last funding round and date, estimated headcount, and the one-sentence differentiation claim from their own website or pitch. Four columns, twenty to thirty companies, two to three hours of work. That is a Monday morning shortlist.
The map does two things. First, it shows whether you are entering a space already dominated by Series B companies while your mandate is at Seed. That is a mismatch worth knowing before you invest time in a first call. Second, it forces the thesis to become specific.
"We like B2B vertical SaaS in the DACH region" is a theme. "We are looking at construction procurement software in Germany, where there are currently eleven companies between EUR 1M and EUR 5M ARR and three of them have raised in the last six months" is a thesis you can act on.
AI-assisted sourcing tools have made this step materially faster. They refresh the sector map automatically and flag new entrants. The judgment about which thirty companies belong on the map, and what the clusters mean, still sits with the analyst. Automation handles coverage. The interpretation of coverage remains a human task.
Primary research as a filtering tool, not a validation ritual
Most primary research in early-stage VC is done too late and for the wrong purpose. The standard sequence is: see the deck, take the first call, do references after term sheet. By the time references are called, most analysts have already formed a view. The references become a box to check, not a source of new information.
Primary research run earlier, before the second founder meeting or during it, is different. The goal is to fail fast, not to confirm.
Three conversations that move a deal, run at the right time:
A customer or potential customer, focused on the buying process. Not to validate product-market fit. The question is structural: how long does the purchasing decision take, who signs, what budget line does it come from. A EUR 40,000 contract that requires CFO sign-off and a six-month procurement cycle is a different business than a EUR 40,000 contract a team lead approves in a week. That difference changes the unit economics, the sales motion, and the capital requirements.
A failed competitor's customer. Someone who tried a comparable product and stopped using it. Their objections are more useful than testimonials from active users. These conversations are harder to arrange. They are also faster to analyze because they surface the structural limits of the category, not just the specific product.
An operator in the space, not an investor. Someone who has built or run a company in the vertical. They know the operational constraints that do not appear in decks. Hiring difficulty, regulatory friction, supplier concentration, seasonality in B2B sales cycles. These are the details that a financial model won't surface until post-investment.
The risk in primary research is selection bias in who you call. If every customer conversation is with someone the founder introduced you to, you are running a validation ritual, not research. A reference list provided by the company is a starting point. The conversations that matter are the ones you found yourself.
AI-assisted screening changes the sequencing, not the judgment
A 2025 survey of nearly 300 venture dealmakers found 92% of VC firms using AI somewhere in the investment workflow, with 64% applying it specifically to company research. What the data does not capture is how those tools change the research sequence rather than just the speed.
The pattern in practice: AI-assisted platforms pre-process the secondary research layer and surface the three or four signals that warrant human attention. Instead of spending ninety minutes mapping a company before a first call, the analyst arrives with a pre-built profile and can spend that time on primary research instead.
This is not a productivity story. It is a quality story. The analyst who gets to primary research faster, with better secondary context already assembled, asks better questions in the first founder conversation. The judgment is still human. The no re-entry between Affinity and your inbox is a workflow benefit. It does not replace the call.
What AI tools do not do: tell you whether the founder will hold conviction through a down round. Whether the market is moving in the direction the deck suggests. Whether the reference who called the founder brilliant has a financial relationship with the company. Those remain the work.
What the IC memo reveals about how you researched
From the LP side of the table, the investment committee memo is the only research artifact we see. And research quality, or its absence, is visible in the memo structure.
A memo that cites the same three secondary sources the deck already referenced has not done research. It has repackaged the founder's narrative in different formatting.
A memo that includes a customer quote, an observation from a competitor's churned user, and one data point that contradicts the deck's claims has done research. That one contradicting observation, handled transparently in the memo rather than omitted, is the most trust-building element for an LP reviewing a GP's process.
The best early-stage memos we have seen at the LP level include a short section called something like "what we did not find" or "open questions." Not because the analyst did not have enough time to answer those questions. Because naming the limits of the research shows that the limits were deliberate and acknowledged, not accidental.
A conversation with a Geneva-based GP made the point directly: their investors do not expect certainty at Seed. They expect rigor. The fastest way to lose credibility with an LP committee is to present a deal where no one has talked to any customers outside the reference list.
Three methods worth removing from your weekly routine
Not all research methods are equal. These three are common in early-stage VC, feel structured, and rarely change the outcome:
Social proof stacking. The deck lists ten well-known advisors. Two of them are on the reference call list. Skip the call with those two and find a customer the founder did not mention. The advisor network is relevant after conviction is formed, not before.
LinkedIn survey outreach as market validation. "We sent a 10-question survey to 200 founders" proves that 200 founders responded to an email. It does not prove that 200 founders would pay for the product, use it consistently, or renew.
Reference checks as a final step. By the time you are checking references after a term sheet is on the table, you have already committed psychologically. Run the most important reference conversations before your second founder meeting. Early references calibrate your primary interview questions. Late references confirm what you already believe.
A practical research sequence for a small fund
For a two-to-four person fund, a workable market research sequence per deal:
Automated secondary profile via AI tooling: 30 minutes or under
Sector map check, where does this company sit among the 25 you already track: 15 minutes
One customer conversation before the second founder meeting: 45 minutes
One operator reference, can be async by email: 30 minutes
Competitive funding history, who passed and approximately when: 20 minutes via Crunchbase
Total active research time: under three hours for a first-pass conviction position. That is not a shortcut. That is less typing and more thinking.
What this sequence produces is not certainty. It produces a better IC memo first draft and a clearer set of questions for the second founder meeting. It also produces a defensible record of how the decision was made, which matters more than most analysts expect once an LP begins asking questions about a deal that did not perform.
The IC memo does not write itself. But the research that feeds it can be more systematic than it typically is in a fund that ran forty calls last month and wrote twelve memos in the same window.
What the analyst actually does on a Tuesday is not the dealflow sprint. It is the twenty minutes of structured secondary research before a call that turns a generic conversation into a diagnostic one.