A pitch deck describes the company as its founders see it. Due diligence must also test claims that can be checked beyond the data room. Useful external research is not a collection of links: it begins with a question that could change the fund's view, identifies the best available source, and records what that source can actually establish.
Start with testable claims
Rank the deck's claims by decision impact. “The company owns the trademark” calls for checking the holder and territory. “The market has 20,000 buyers” needs a buyer definition, a countable population, and a time period. “There are no competitors” requires a search for alternatives, including products that solve the same problem differently.
Before searching, write down the claim, the evidence you expect, the consequence of a different answer, and the date on which the claim must hold. The last item prevents you from comparing a September deck with an industry figure published two years earlier.
Match the source to the question
For a French company, the INPI company database provides entries from the national company register, some corporate acts, and non-confidential annual accounts. It can confirm legal identity or show that accounts are available. It cannot establish a startup's current revenue when recent accounts have not been published.
The INPI trademark database helps examine a published mark's holder, status, and territory. A search result is not a legal opinion on freedom to operate or ownership of the underlying technology. For insolvency proceedings, France's public-service guidance explains how to use BODACC notices. Failure to find a notice does not prove the absence of financial risk.
Industry sources answer different questions. An Eurostat count can frame the number of businesses in a sector, but it does not show how many can use the product or intend to buy it. A competitor's website establishes that an offering is advertised, not its adoption. Customer interviews can illuminate actual workflows and alternatives, provided the sample and recruitment method are recorded.
Keep a verification log, not a link dump
A useful research entry contains the claim, the opened document or URL, publisher, publication and access dates, the precise passage, and a status. Distinguish “confirmed within this scope,” “contradicted,” “unverified,” and “insufficient source.” A second website repeating the company's press release is not independent corroboration.
Record the investment implication and next step. If the deck says the startup owns a mark but the register lists a founder as holder, ask for the assignment or license and supporting documents. That observation alone does not prove the startup has no rights.
A fictional record can fit in five lines: claim, “the company owns the mark”; source, INPI entry accessed on September 23; finding, the displayed holder is the founder; limit, an assignment not reflected in the viewed result remains possible; next step, request the assignment or license and check its scope. Another reviewer can then pick up the work without repeating the entire search.
Fictional example: separate a signal from a verdict
Imagine “Atelier Delta,” a fictional industrial software company. Its deck claims “4,000 addressable factories in France” and “protected technology.” An analyst finds an official count of 4,000 establishments in a neighboring category, a published patent listing a cofounder, and two competing products.
Three checks remain. How many establishments meet the product's technical and budget constraints? Does the patent cover the sold feature, and what rights does the company hold? Do the competitors target the same job, buyer, and service level? Until these questions are answered, the findings are leads, not a validated market size, secured asset, or settled competitive thesis.
Bring decision-relevant findings back into the deal
At screening, a few focused checks may suffice: identity, obvious discrepancies, market order of magnitude, and existing alternatives. During diligence, investigate questions whose answers could change valuation, investment conditions, or the committee's recommendation. Bring each external finding into the deal record with its source, limitation, and a question the team or founders can resolve.
This discipline avoids treating the deck as its own proof or a web page as a final verdict. Continue with our guides to bottom-up market sizing and reviewing AI-generated investment analysis.