The Diligence Red-Flag Reference: Walk Away, Price It In, or Ignore

You found a warning sign during diligence on a company or its founders, and now you need to know what to do with it. This reference grades the flags an early-stage investor actually stumbles onto - across the company's public footprint and the founders' history - into three verdicts: walk away, price it in, or ignore. It is built for platform and talent partners, deal-team analysts, and angels who want to jump to one row and leave with an answer.

Most published checklists list red flags as undifferentiated prose that mixes a fatal cap-table dispute with a cosmetic pitch tic and never says how each flag misleads. This one is organized by the signal you find, not by workstream, and every entry tells you what the signal proves and what it looks like when it lies.

The three verdicts, defined

Every flag resolves to one of three verdicts. Walk-away means the flag, if confirmed, blocks the investment regardless of price. Price-it-in means the flag is real but survivable at a lower valuation, more structure, or a longer diligence loop. Ignore means the flag is noise: a false positive that will kill good deals if you treat it as evidence.

The hard part is that the same surface signal can land in any of the three, depending on what it turns out to prove. A departed co-founder is cosmetic until you learn they still hold 25 percent of the company. A clean PACER report is reassuring until you realize the founder's history is state-based and older than the federal index reaches. The verdict is never the signal itself. It is what you confirm underneath it.

A useful baseline before you start: a VC can count on spending at least 20 hours of diligence per deal, and a study of 700 firms found that deals once took 83 days to complete. That budget is what buys you the verdict. Rushing it is how cosmetic flags get read as fatal and fatal ones get waved through.

The graded red-flag lookup

Jump to the row that matches what you found. Each verdict below assumes the flag is confirmed; the "what it proves" column is what you are actually buying or avoiding, and the failure-mode section that follows tells you how each one lies.

The equity, IP, and runway rows are where money moves. Unclear IP ownership is the one flag on this list that can silently convert from price-it-in to walk-away: a single GPL violation can force relicensing of an entire codebase, and missing founder 83(b) elections are a frequent, preventable legal-review flag. Confirm the entity provably owns all its code before you decide.

The bottom three rows are where deals die by mistake. Each is a false positive dressed as a signal, and each has a specific test that flips it.

What each flag proves, and what it looks like when it lies

Here the flags recur across CRV, Affinity, AngelSchool, MicroVentures, and lender checklists: messy or opaque cap tables, unclear IP ownership, weak founder credibility or commitment, unresolved legal issues, and a small or undefined market. Affinity frames the whole thing across nine areas including a dedicated founder-background workstream: finance, tax, legal, HR, assets, IT, products and services, marketing and sales, and founder background.

Cap-table and equity flags

A messy cap table proves nothing on its own; the register just needs cleaning. The dangerous version is dormant equity. A co-founder who left after six months but still holds 25 percent of the company is not awkward, it is a live obstacle, because investors will not want to negotiate around that uncertainty at the next round. This is the flag that looks cosmetic and is not. The test: audit vesting and the full cap register, not the current ownership summary the founder shows you.

Runway and commitment flags

The one repeatedly cited numeric threshold is runway. Investors treat runway below the expected range as a serious red flag, and founders should aim for at least 18 to 24 months at the current burn rate. Burn itself is described qualitatively - high burn without matching growth - rather than with a hard multiple, so read burn against validation, not against a fixed number. A part-time or hedging founder is a genuine commitment flag; a quiet or non-charismatic one is not, and the difference matters enough to get its own failure mode below.

Legal and prior-venture flags

Unresolved legal issues and key-person risk are consistently named. Lenders specifically flag founder disputes, unclear equity ownership, lack of industry experience, and over-reliance on one person. The load-bearing statistic behind all of this is that 65 percent of high-potential startups reportedly fail due to co-founder conflict, from Noam Wasserman's work. Use it, but know its floor: it traces to a 1989 survey of investor perceptions, not a modern audit. It justifies scrutinizing team dynamics. It does not settle a base rate.

The verdict is never the signal itself. It is what you confirm underneath it.

Where to verify: public founder-history sources and their limits

The primary public source for litigation, bankruptcy, and prior-venture disputes is PACER, the federal court index. It is reliable within its scope and misleading at its edges, so treat every source below as partial and stack them.

PACER charges $0.10 per page with a $3.00 cap per document, updates once daily at midnight, and keeps most cases created before 1999 in paper form only. It is federal only, so state-court litigation, most contract and employment disputes, and equity fights settled privately never appear. The practical consequence: the older and more state-based a founder's history, the more a clean PACER report is a false negative. Named company-screening platforms that practitioners use alongside it exist, but they cover company signals, not the state-court and private-settlement gaps.

The on-list references you get from the founder are the third source here, and they carry a bias you must correct for. Because references default positive, a merely mixed on-list reference is usually a bad signal. That structural skew is the whole reason back-channels exist.

Building the back-channel set

Off-list, or back-channel, references are the people you find entirely on your own, outside the three to five individuals the founder hand-selects. This is where the informative signal lives, because on-list calls are coached and default positive. CRV recommends four to five conversations mixing firm-list and independently found people; top firms cited as Sequoia, Andreessen Horowitz, and Benchmark run 10 to 20 or more reference calls per deal, deliberately targeting people the founder does not name.

The method for finding them: trace the specific partner's career history and the founder's prior firms to surface former colleagues who are entirely outside the fund's current network. This is the friction point in diligence, and it is a sourcing problem. Naming the people who worked directly with a founder at a prior venture, and are now elsewhere, is exactly what Refolk does from a plain-English request.

One structural caveat on back-channels, from Refolk's index: the US-to-UK founder pool runs about 4:1 (665,330 versus 162,765 profiles carrying a founder title). A back-channel map that works easily in the US thins fast in the UK, where there are fewer independent former-colleague hops per founder. Plan for more effort per call there.

The diligence procedure, in order

Run these steps in sequence. The order matters: cheap filters first, expensive independent verification last, and the verdict only at the end when you can see patterns rather than anecdotes. One ordering disagreement to know: CRV and reverse-diligence guides place reference calls after signals mature and a term sheet is out; investor-side diligence on a company typically runs references before the term sheet. Pick the order your process demands, but do not skip the back-channel stage.

How this goes wrong: the false positives that kill good deals

This is the most valuable section in any red-flag reference, because most diligence errors are not missed fatal flags. They are cosmetic flags misread as fatal, and clean reports misread as clearances. Each failure mode below names the mistake and the test that corrects it.

"No competition," mis-read. Treating a founder's "we have no competition" as proof of a virgin market is the classic mistake. Every business has competitors; a founder claiming none is either unaware or hiding it. That damages credibility, but it proves a research or honesty gap, not market absence. Test: ask how customers solve the problem today. Verdict: ignore as a market signal, note as a credibility ding.

Glowing references, read as strength. Uniformly enthusiastic references with no texture and no acknowledgment of friction may indicate coached or pre-screened conversations. Reading uniform praise as conviction builds confidence on curated calls. Test: dig for the mixed signal; a reference set with zero friction is itself suspect. Verdict: ignore the glow, weight the hesitations.

Single negative anecdote, read as fatal. No single answer has to be disqualifying. The value comes from patterns across multiple conversations, and a repeated concern carries more weight than one outlier complaint. Test: require the concern to recur across independent calls before you downgrade. Verdict: ignore until it becomes a pattern.

"PACER clean" read as "clean." Assuming no federal record means no history misses state-court and privately settled equity disputes entirely. Test: add state records and back-channel before you conclude there is no history. Verdict: a clean federal search is a partial search.

Quiet founder, read as weak founder. Penalizing a non-charismatic operator for pitch style is a real and common error, because skeletons in closets do not show up in pitch decks. Integrity shows in history, not in delivery. Test: judge the record, not the room. Verdict: ignore.

Stealth label, read as hiding. In Refolk's index, about 6,930 of 665,330 US founder profiles reference stealth, roughly 1 in 96. A label that common cannot be a hiding signal. Test: verify prior-venture outcomes through records and back-channel before pricing anything in. Verdict: noise until proven otherwise.

Dormant equity, ignored. A departed founder holding large equity looks cosmetic until the next round negotiates around it. Test: audit vesting and the full cap register. Verdict: price-it-in at least, walk if unfixable.

A verdict memo you can paste into the IC deck

Write one of these per material flag. It forces you to name the source, what it proves, and how you checked it lies, which is the whole discipline of this reference in a few lines.

FLAG: <the signal as found, one line>
SOURCE: <where it surfaced: PACER / cap table / on-list ref / back-channel>
WHAT IT PROVES: <the underlying fact, not the surface signal>
HOW IT COULD LIE: <the false-positive test you ran>
CROSS-CHECK: <second independent source and result>
VERDICT: WALK-AWAY / PRICE-IT-IN / IGNORE
IF PRICE-IT-IN: <valuation, structure, or condition that offsets it>

Before you call diligence done

Run this checklist before the verdict goes to the investment committee. It maps to the failure modes above: every item exists because skipping it is how a good deal dies or a bad one gets through.

Keeping this reference current

Nothing in this document is time-sensitive except PACER's fee schedule and coverage cutoff, and both are re-checkable at the source. Confirm the per-page and per-document caps and the pre-1999 paper cutoff each time you rely on them, because those are the numbers that decide whether a clean search is a real clearance. The founder-index footprint figures come from Refolk's index and shift as the index grows; treat the 4:1 US-to-UK ratio and the roughly 1 percent stealth prevalence as orders of magnitude, not constants.

The one number to distrust on purpose is the 65 percent co-founder-conflict statistic. It anchors most founder-risk arguments and derives from a decades-old perception survey, so the honest move is to pair it with the verifiable public-record and back-channel process in the procedure above. A red-flag reference is only as good as its willingness to say where the evidence is thin. This one is thin on base rates and firm on process, which is the right way around.