Due diligence
The due diligence questions worth sending a founder
Skip the 80-line questionnaire. The questions worth sending come from the company's own record: what's missing, where two sources disagree, and what the founder promised on the call. Usually three of them decide whether you go on.

Short answer: the due diligence questions worth sending come from the company's own record. Read everything they sent, check the claims that matter against independent sources, and then ask about three things: what's missing, where two sources disagree, and what the founder promised and hasn't sent yet. Rank those by whether the answer would change your decision. I think most deals turn on three questions or fewer.
Most funds also have the other kind. An 80-line questionnaire about market, team, product, cap table, churn and culture. The founder loses a week to it. Half the answers are already in the deck. And the one question that decides the deal sits on page four, between "describe your hiring plan" and "list your competitors".
Read before you ask
Before you write a single question, read the deck, the data room, the intake answers and the call notes. Then check the claims that carry the deal against something the company didn't write.
A question you could have answered yourself costs you twice. The founder spends a day on it, and they learn you didn't read what they sent.
Three kinds of question worth sending
What's missing. A seed SaaS company that states revenue and says nothing about gross margin or churn. A medtech company with a launch date and no regulatory pathway. The gap is the question. Ask for the figure or the document, and give a date.
Where two sources disagree. Say the deck says $1.2M ARR, the bank export shows about $70k a month coming in, and on the call the founder said "we're at about a million". That's three numbers from three sources. Asking "what's your ARR?" gets you a fourth.
Quote what they said and ask how the pieces relate:
On the call you said "we're at about a million". The deck states $1.2M ARR. If the two measure different things, for example contracted against live revenue, how do they relate?
There's often an innocent answer. A signed contract that starts in March. A pilot counted as a customer. You still want it in writing before the committee meets.
What they promised. Every first call has a line like "I'll send you the cohort data this week". Nobody writes it down, and three weeks later nobody remembers who owes what. A promise is a question that already has a deadline. Write it down in their words, with the day they gave.
Three before you proceed
Now rank them. Which answers would change what you do? Usually one or two. Sometimes three. Those go first. The rest can wait for confirmatory diligence.
For each one, write two lines for yourself: why it decides the deal, and what answer would settle it. Send the founder only the second line. They need to know what a good answer looks like. Your reasoning stays inside the fund.
How Aventro builds the list
Aventro is an AI workspace for VC due diligence. It reads the record, checks each claim against independent sources, and builds this list from what it finds.
- After a full analysis it draws at most three questions before you proceed. Each one is tied to a specific finding in the record, with why it decides and what would settle it.
- "Ask the company" sends the question with what would settle it. Why it decides stays with the fund. Ask again and Aventro finds the request already out, so the founder never gets it twice.
- When a founder says a figure aloud that differs from the record, the request is drafted around their own words, quoted, and asks how the two relate.
- From a call transcript, it pairs what your team asked with what the company said back. A promise becomes a request you can send in one step, quoting their reply, with the day they gave as its due date.
- A figure the company didn't state shows as not stated, so the gaps are visible before you write a word.
It drafts. A person decides what goes out. The rest of the diligence workflow is here, and every source behind the checks is on the verification page.
A test for any question
Before a question goes out, ask:
- Is the answer already in what they sent?
- Would the answer change what we do?
- Does the founder know what a good answer looks like?
Yes to the first: delete it. No to the second: it can wait. No to the third: rewrite it.
If you want to see the questions Aventro draws from your own deal flow, book a demo.
Related
- How to verify a startup’s claims before investingSeven steps to verify what a startup claims before investing: list the claims, go to the register that holds each, confirm the match, rank the sources, compare figures, ask, and label what stays unconfirmed.
- Can AI find contradictions between a pitch deck and the data room?How AI finds contradictions between a pitch deck, the data room and calls: every document read on arrival, each disagreement shown side by side, outside records compared too, and the question sent to the company.
- DiligenceDocuments read on arrival, claims checked at the source, gaps asked of the company.
