To verify a startup's traction, ask for the source record behind each number that drives your decision and check it yourself: signed contracts for customers, invoices and bank receipts for revenue, and cohort data for retention. At seed this takes an afternoon. Most gaps you find will be definitions rather than fraud, and either way you need to know before you wire.
Key takeaways
- Tie each claim to a record: customers to contracts, ARR to invoices, cash to bank data.
- Most overstatement comes from definitions, such as pilots counted as customers or one-time fees counted as recurring revenue.
- Recompute the burn multiple on verified ARR. It can move a lot.
- Bank statements sent as PDFs can be forged. Read-only access or a direct bank confirmation is stronger.
- Ask for definitions in writing, then hold the company to them in its updates after you close.
Why this check matters more now
The Joonko case shows the extreme version. Its founder pleaded guilty to securities fraud after raising about $27M, with prosecutors saying she overstated customers and falsely claimed some of the world's largest companies as clients. When an investor questioned performance in 2023, she allegedly provided falsified bank statements and fictitious purchase orders.
Few founders are anywhere near that. The everyday problem is softer. A Crunchbase guest piece on the case lists the usual places where traction drifts: LOIs and pilots described as signed revenue, customer counts that include non-payers, one-time fees counted as recurring, and churn left out of the story.
Two market facts raise the stakes for early investors. First-time financings reached 8,625 through Q3, above the full-year 2021 record, so there are more young companies to evaluate. And only 16% to 24% of recent $1M-plus seed cohorts have moved past seed so far, while Series A investors reportedly want $2M to $3M in ARR. If you are underwriting a path to that bar, the starting number has to be right.
The four numbers to check, and what to ask for
Customers. Ask for the customer list with contract start date, term and status. Count only those paying today. Then call two or three, chosen by you, not by the founder.
ARR. Ask for the definition in writing and a month-by-month schedule by customer. Match a sample to invoices. Remove pilots, services, implementation fees and anything cancelable without a contract behind it.
Retention. Ask for cohort data, not a single percentage. A blended figure can hide a weak recent cohort behind a strong early one. If the company is too young for cohorts, say so and price that uncertainty rather than accepting a projected rate.
Cash. Compare bank receipts with invoiced revenue for the last six months. A persistent gap means slow collections, or revenue that is not what it seems. Ask for read-only bank access or a direct confirmation, because the Joonko case shows a PDF statement is not proof on its own.
Definitions vary even among professionals. Value Add VC's comparison of funding trackers shows H1 2026 totals from $412.7B to $560.3B, largely because each counts differently. Nobody is lying there. It is a useful reminder that a number without its definition tells you little.
A worked example, with illustrative numbers
Say a seed company reports $1.2M in ARR, with $900K of net new ARR over the past year and $1.5M of net burn. That is a burn multiple of about 1.7, which looks efficient.
You ask for the customer schedule and find two things:
- $300K is from paid pilots with no commitment past the pilot term.
- $180K is one-time implementation and services fees.
Verified recurring revenue is $720K, or $60K a month. If all $480K of removed revenue was added this year, net new ARR was $420K, and the burn multiple becomes about 3.6.
Then you compare the bank data. Receipts average $48K a month against the $60K invoiced. The $12K monthly gap may be slow payers, which is a question for the founder and not an accusation. Ask for the aged receivables.
Nothing here has to be dishonest. The founder may have used a loose definition that nobody challenged. But your decision changes: a company at $720K with a 3.6 multiple is a different investment at a given price than one at $1.2M with a 1.7 multiple, and it is further from a $2M to $3M Series A bar.
What to do when the numbers do not tie
A mismatch is a starting point, not a verdict. Young companies keep imperfect books, and a founder may have built the deck from a spreadsheet nobody has reconciled. Send the differences back in writing and ask the founder to explain each one. Then sort what you hear into three groups.
- Definition differences. The company counted something you would not. Restate the figures and carry on, adjusting your view of price if the restated number changes the case.
- Timing differences. Revenue was invoiced in one month and collected in another, or a contract was signed but has not started. These usually reconcile with a small amount of work.
- Unexplained differences. The founder cannot show a record, or the explanation changes each time you ask. Pause here. A forged document is rare, but a story that keeps shifting is reason enough to stop, whatever the cause.
Keep your own notes on what you asked, what you received and when. If a syndicate is relying on your work, say plainly what you checked and what you did not.
How to ask without damaging the relationship
Good founders expect these questions, and the way they respond tells you something. A few things I have found help:
- Ask early, and ask everyone. "This is what I request from every company" removes any implication.
- Request records, not explanations. A schedule and a sample of invoices is less awkward than debating a slide.
- Separate forecast from fact. Optimistic projections are different from false claims about existing customers. Judge the first on reasoning and the second on records.
- Watch for reluctance over simple things. Slow answers are normal at a small company. Repeated refusal to share a customer list under NDA is a signal.
If the company has a data room, this should take little time. Our data room checklist shows what a prepared seed company tends to have ready.
A checklist before you wire
- Get written definitions of customer, ARR, churn and pilot.
- Reconcile the ARR schedule to invoices on a sample of at least five customers, including the largest.
- Remove pilots, services and one-time fees, then restate ARR.
- Recompute growth and the burn multiple on the restated figure.
- Compare six months of bank receipts with invoiced revenue, using read-only access or a direct confirmation.
- Make two reference calls to customers you selected.
- Review the cap table and outstanding SAFEs with the SAFE modeler, so the price you are paying is the one you think it is.
- Agree on the monthly metrics and definitions the company will report after you close.
After the check closes
Verification does not end at the wire. The written definitions become the baseline for every update you receive, which makes later decisions easier. If the company changes how it counts ARR, you will notice. That also feeds the decision I wrote about in how to decide whether to follow on in a seed company.
The takeaway
At the earliest stage, you are often the only person checking. The test is not whether a founder is honest, which you cannot verify anyway. It is whether each number you rely on traces to a record you have seen, under a definition you both agreed on.
Frequently asked questions
What documents should I ask for to verify startup revenue?
Ask for a month-by-month ARR schedule by customer, a sample of contracts and invoices, and six months of bank data. Reconcile them against each other. If they tie out and the definitions are clear, you can rely on the headline number much more.
Should pilots count toward a startup's ARR?
Generally not, unless the customer has committed to continue past the pilot term. Paid pilots are real revenue and worth reporting, but they should be shown on a separate line. Mixing them into ARR overstates what is likely to recur.
How much diligence is reasonable for a small angel check?
A small check does not need a full audit, but it should still cover the basics. Getting written definitions, one ARR schedule, a bank comparison and two customer calls can take an afternoon. Scale up from there as the size of the check and the price rise.
Is it rude to ask to speak with a startup's customers?
No. It is a normal request, and experienced founders expect it, usually late in the process and under an agreed approach. Pick customers yourself from the full list where you can. A founder who repeatedly refuses is giving you useful information.
Sources
- AI founder admits securities fraud
- When Traction Becomes Personal Exposure: What The Joonko Guilty Plea Means For Founders
- Jumbo Series A Rounds Hiding The Real Squeeze On Seed Startups
- Seed Got Smaller: Q3 2026 PitchBook-NVCA Venture Monitor Data
- Venture Numbers, Reconciled: Why Every Tracker Says Something Different About 2026