The Pre-Wire Diligence Standard for Seed Deals

You have decided you want this seed deal. This standard tells you whether your confirmatory diligence is complete enough to commit capital and wire, and it is written so two partners looking at the same deal reach the same verdict. It is for early-stage investors, platform and talent partners, and angels who need an acceptance test for the moment before the money moves, not another memo template or a checklist of things to skim.

Most published guidance describes what to write in a memo or what to review in a data room. This is different: it is the definition of done for the confirmatory gate. Every load-bearing claim about the founder, the traction, and the team must be cleared and evidenced against named criteria before you grade a deal ready-to-wire.

What "ready-to-wire" means at seed

Ready-to-wire means every load-bearing claim in the deal is marked corroborated by an outside source, with its evidence recorded, and no unresolved negative remains on founder identity, traction, cap table, or IP. If any of those carries an unverified headline claim, the grade is not-yet.

The distinction that makes this gradeable is corroborated versus unverified. A claim is corroborated when a paying customer, a public filing, or a signed instrument confirms it. A claim is unverified when your only source is a founder dashboard, a pitch deck slide, or the founder's own word. Two people applying that rule to the same fact will tag it the same way, which is the entire point of a standard.

Seed diligence is short. Typical seed confirmatory work runs 1 to 2 weeks, against 4 to 8 weeks for a Series B or later round. That tight box is the real constraint on completeness. You cannot invent thoroughness in the final days, so the standard assumes a pre-built evidence log where each fact is tagged as you clear it.

The Frank case is the reason this standard exists in the shape it does. JPMorgan acquired Frank for $175 million on a claim of more than four million users when the real figure was near 300,000. The fraud unraveled only when the buyer tried to contact the customer base after acquisition. The lesson is not that fraud happens; it is that a document survives review, and only live contact with a real customer defeats a synthetic number. Charlie Javice was later sentenced to 85 months and ordered to pay over $300,000,000.

The four workstreams you must clear

Four workstreams carry the grade: founder identity and background, traction corroboration, cap table and IP, and references. Each has a named done state. A deal is ready-to-wire only when all four clear; a fail in any one is disqualifying, not something you price around with a discount.

• Founder identity and background. Verify LinkedIn and CV claims against employer records, confirm degrees and specific accomplishments cited in pitch materials, and run an open-source search of the founder's internet presence. For larger checks or newer relationships, add a vendor report covering criminal, litigation, and financial history. Done means no identity or history claim stands unverified.
• Traction corroboration. Cross-reference every headline number against independent data and call paying customers. Done means each metric is confirmed by a customer or a filing, with the vanity-versus-active distinction settled.
• Cap table and IP. Re-model every instrument from scratch and trace every IP asset to a signed assignment. Done means the table reconciles to signed documents and the entity owns its core technology.
• References. Complete a mix of on-list and off-list calls including one struggled founder. Done means coverage across categories with no unresolved negative.

Column three of every workstream is the evidence. If a workstream produces no written record of what confirmed the claim, treat it as unstarted.

The confirmatory procedure, in order

The procedure below is the sequence a deal lead runs from desk screen to commit gate. It maps one-to-one to the acceptance steps you grade against, and it is built to fit the 1-to-2-week seed window.

Sources disagree on where references sit in this order. CRV recommends reference calls after receiving a term sheet and before signing it. Top firms run them earlier, between the partner meeting and the term sheet. Diligence often formally starts once the term sheet is agreed. Pick one order and make it team policy so your grading stays consistent across deals; the standard requires that references clear before you wire, not that they happen on a specific day.

How many reference calls, and which kind

For a seed deal, target five to seven reference conversations, mixing founder-provided on-list references, self-sourced off-list back-channel references, and at least one founder from a company that struggled or failed. Volume alone does not grade a deal; the mix does.

On-list references are the three to five hand-selected people the founder provides. By default they come back positive, so a positive on-list call proves almost nothing and a mixed one is a bad signal. Off-list references are the people you find on your own. These are the calls that move decisions: investors rarely walk on a curated reference but frequently walk on an independently sourced one when a serious new issue about the team surfaces.

The struggled-founder call is non-negotiable in this standard because it is the one conversation that surfaces how a founder behaves when things go wrong, which curated references are selected to hide. If your reference set contains no off-list call and no struggled founder, the reference workstream is not done, regardless of how many total calls you logged.

Positive curated references prove almost nothing; the off-list call is where a deal actually breaks or holds. </pull> Finding off-list references is the friction most funds run out of time for inside a two-week window. Independently sourcing former colleagues, churned employees, and struggled founders by hand is slow, and in a busy process it is the step that quietly gets skipped. This is exactly where I remove friction: describe the person you need in plain English and get a bench of names to call, without waiting on the founder's list.

refolk prompt: Founders whose startups shut down or were acqui-hired after raising a seed round note: Returns a bench of "struggled founder" references the standard requires, sourced independently of the deal you are grading. slug: 3d2yrpw51c

The cost of these calls is asymmetric and cumulative. A twenty-fund process turns into the same three customers taking six investor calls each, and by close those customers have drawn their own conclusions about the round. Run tight, shared reference lists rather than sending every partner to the same well.

## Traction: corroborated versus vanity

Confirm a traction claim only when a paying customer or a public filing confirms it; a founder dashboard is unverified. The single most common inflation is the gap between total and active users. "10,000 users" may be technically true while active users in the last 30 days number 147, and those are not the same claim.

The verification move is direct customer contact, and Hustle Fund's published floor is the cleanest bar to adopt: talk to at least three real customers who pay money, and at least two investors not suggested by the founder. Paying customers defeat synthetic user lists, and unsuggested investors defeat curated enthusiasm.

Each traction signal proves something specific, and each has a tell when it lies:

- **Active-user count.** Proves live demand. It lies when reported as a cumulative sign-up total; demand a 30-day active definition and a source export.
- **MRR and retention.** Prove durable revenue. They lie when pulled from a dashboard the founder controls; reconcile to bank data or a filing.
- **Logo list.** Proves commercial reach. It lies when the "customers" are pilots or free users; confirm each is paying and in production.
- **TAM.** Proves market size. It lies when it cannot be sourced to a credible third party; if the number traces only to the deck, treat it as unsupported.

Cap table, IP, and the data-room minimum

At seed, the acceptance state for legal foundations is narrow but hard: the cap table reconciles to signed instruments, and IP assignments are signed by every founder and contractor. Full audited financials, deep cohort history, and SOC 2 are deferred to later rounds and are not part of this gate.

The consensus seed data-room minimum is a pitch deck, basic financials, cap table, certificate of incorporation, SAFE documents, IP assignments, and founder bios, typically 40 to 50 documents. VCs almost always request the cap table, financial model, articles of incorporation, and IP assignments first, because those are the load-bearing documents. Re-model every SAFE, option grant, and convertible note from scratch rather than trusting the founder's summary table.

Unclear IP ownership is the single most common deal-killer at seed. If a founder or contractor built core technology before incorporation and left without assigning it, the startup owns nothing, and no term-sheet adjustment fixes that. A tidy-looking cap table can also hide unsigned SAFEs, unconverted notes, or handshake equity promises; reconcile every line to a signed document before you accept the table as clean.

Note that unresolved hygiene reprices the deal through delay, not headline terms. Founders who have not cleaned the cap table, IP, or contracts add 4 to 8 weeks during diligence. If you find yourself in that stretch, the question is no longer price; it is whether the deal is quietly dying.

Where this goes wrong

The failure modes below are where a diligence process passes on paper and fails in reality. Each is a named trap with a check that neutralizes it. Treat this section as the most valuable part of the standard, because a gate that misses these grades a bad deal ready-to-wire.

• Vanity-metric pass. "10,000 users" is true but 30-day actives are ~147. Check: demand an active-user definition and a source export, not a dashboard screenshot.
• On-list-only references. Curated referees always return positive, so a mixed one is already a bad signal. Check: require self-sourced off-list calls before grading the reference workstream done.
• Synthetic customer data. Fabricated lists survive document review, as the Frank case shows. Check: contact real paying customers directly before you close, not after.
• Cap-table false clean. A tidy table hides unsigned SAFEs or handshake equity. Check: reconcile every line to a signed instrument.
• IP mirage. Product built pre-incorporation or by an unassigned contractor. Check: signed PIIA from every founder and contractor, no exceptions.
• "Advisor from Google" inflation. A three-month contract or one coffee reported as senior tenure. Check: verify tenure and title against employer records.
• Slow-diligence self-deception. Treating a stalling process as rigor. Check: repeated same-information requests and a six-week drift signal dying conviction, not thoroughness.
• Underrepresented-founder false negative. A thin back-channel network read as a red flag. Check: do not let the absence of a reference network automatically count against a founder.

That last one cuts the other way from every check above, and it matters. The absence of a dense reference bench is a fact about the founder's network, not about their integrity. Geography sharpens this. In Refolk's index of professional profiles, the UK founder pool is roughly a quarter the size of the US pool, at 153,989 against 605,718, so back-channel coverage is denser and harder to escape in the UK and thinner by default in smaller markets. Read a thin network in context, not as guilt.

The Frank case sits in the top-right of that matrix as a claim treated as corroborated when it was not. The buyer accepted a document instead of contacting the customer base, and the four-million-user figure held through review because nobody made the call that would have broken it. The structural lesson: the most expensive failures happen when a buyer accepts a document in place of live contact.

The pre-wire checklist

Run this checklist at the commit gate. Every item must be true to grade ready-to-wire. A single unchecked load-bearing item means not-yet, and the correct response is to resolve it or walk, not to discount the price.

Keeping the standard current and honest

Adopt this as team policy, then keep three things under review so the standard does not drift. First, the evidence-log format: there is no single published named template for it, so build your own with the documented components - confirmed facts, open issues, quantified risks, and recommended deal-term changes - and hold every deal to the same skeleton so grades stay comparable.

Claim: <the exact assertion being graded>
Materiality: high | low
Status: corroborated | unverified
Source: <paying customer, filing, signed instrument, or "founder dashboard only">
Verified by: <name and date>
Open issue: <what, if anything, remains>

Second, re-check the time box against your own deal flow. The market is stretching: median seed-to-Series-A time has moved to roughly 774 days, up from about 420 in the recent past, which means a seed founder must survive far longer on this round. That raises, not lowers, the bar on traction durability and cash discipline, so revisit what "enough proof" means as the interval moves.

Third, keep your off-list sourcing sharp. The reference bench is the workstream that erodes first under time pressure, and it is the one that moves decisions. Building an independent bench of former colleagues, churned employees, and struggled founders by hand is the bottleneck; Refolk collapses that search into a plain-English query so the off-list call, not just the curated one, actually happens inside the seed window. Re-run those searches on every deal rather than reusing a stale list, because the graph changes and the point is independence from the founder's selection.