Kill flags: the eleven we see most often.

A taxonomy built from 1,142 dossiers. Eleven flags account for 78% of all material kills — and four are recoverable if the founder addresses them in the first call.

01The taxonomy

A kill flag is an observation that, when confirmed, terminates the dossier before it reaches a partner meeting. Not a concern, not a question to revisit — a position the rest of the diligence cannot recover. Across 1,142 dossiers analysed in private beta, we tagged 247 distinct issues as material enough to surface in the IC summary. Eleven flag types account for 78% of those. The remaining 22% are idiosyncratic — vertical-specific, country-specific, or the kind of one-off pattern that does not repeat often enough to taxonomise.

The eleven split cleanly into two cohorts. Seven are structural: the issue is wired into the business model, the cap table, or the co-founder relationship, and recovering from it requires re-architecting the company. The other four are recoverable in the first call. The founder either reframes the claim honestly, or they don't — and the answer tells the IC more about the founder than the original claim did.

The distinction matters because most kill flags are not, on inspection, evidence of bad founders. They are evidence of founders who have absorbed a fundraising script that treats every pitch as adversarial. The script teaches them to fortify positions that, with a better framing, were never actually weak. Removing the fortification is sometimes all the diligence requires. Other times the fortification is the company.

This note inventories all eleven. We use the same taxonomy across every dossier — same names, same severity scoring, same recommended IC conditions. Consistency is the discipline. The IC partner reading their fortieth dossier should not have to learn a new vocabulary on the fortieth.

02The seven that kill the round

These are the seven we cannot diligence around. Each requires a structural change to the business before the round can price.

1. Founder–market mismatch. The founder has the conviction but not the lived knowledge. A second-time founder pivoting from consumer to healthcare; a domain expert who has never built a product; an operator who has scaled a team but never written a strategy memo. The test is to ask the founder to describe the structural failure mode of their three most plausible competitors in detail. Not the elevator pitch — the specific reason each competitor is underequipped for the customer the founder is targeting. Founders with real domain knowledge answer in three minutes per competitor. Founders without it talk around the question.
2. Cap-table damage. Pre-seed angels holding more than 25% of the company; common stock below 50% of issued; a strategic investor with right-of-first-refusal on the next round. Each alone is recoverable through a reset, but the reset is a six-month negotiation no founder enjoys. We flag the combination when it appears, and the combination appears more often than the deck implies.
3. Single-channel CAC. One acquisition channel responsible for more than 70% of paying customers, with no measurable diversification effort over the prior nine months. Founders treat this as a feature (“we found product–market fit on LinkedIn ads, why diversify?”) and it is, until the channel saturates or the platform changes its algorithm. The most expensive lesson in growth-stage SaaS is paying for distribution risk you did not price.
4. Margin structurally below the cohort floor. A SaaS company with sub-40% gross margin and no plausible path to 60% inside three years. A vertical-AI company with a model-cost line that cannot fall below a fixed fraction of revenue. The flag is not the absolute number; the flag is the structure that prevents the number from improving. A founder who can articulate the path — and produce the engineering or pricing milestones that move the line — clears the flag. A founder who insists the current margin is the steady-state does not.
5. TAM that fails the four-cut rebuild by an order of magnitude. See Note 01 for the rebuild. The kill flag is not the gap — every deck has a gap. The kill flag is when the rebuilt TAM, defensibly constructed against the founder's own ICP, returns a number under $250M at full capture. Below that line the round economics do not pencil for either side.
6. Platform dependency without a mitigation plan. See Note 03. We mark this kill flag when two conditions hold simultaneously: the product is inference-dependent in a single-provider configuration, and the founder cannot describe a thirty-day mitigation plan. Roughly one in four AI decks satisfies both. It is not always fatal, but it is always a term-sheet condition.
7. Co-founder structure that won't survive Series A. Two founders with overlapping role definitions and equal equity, no operating agreement, no resolved authority for hiring decisions. Three founders where two have CEO ambitions. A solo founder with a half-time CTO who has not committed to going full-time after the round closes. The IC does not solve interpersonal questions; the IC prices them in, and below a certain threshold the price is too high.

These seven sit upstream of the round and require pre-round resolution. We do not refuse to write a dossier on a company with one of these flags — we surface it, score it, and recommend the IC condition. Sometimes the condition is met; the company restructures, the cap table resets, the co-founders take the difficult conversation. More often the founder withdraws, finds a less rigorous investor, and the company encounters the same flag eighteen months later under worse leverage.

03The four that don't have to be

The remaining four kill flags share a property: they look the same on the deck as the structural seven, but the cause is framing, not structure. The founder dropped a defensive framing during fundraising, and the framing concealed a number that, presented honestly, would not have killed the round.

8. Burn understated. The founder quotes adjusted burn without flagging it as adjusted, and the variance to net burn exceeds 15%. See Note 04. Recoverable because the underlying business is sound — the issue is presentation. The founder reframes the number in the first call, the IC notes the original framing as a one-off, and the conversation continues. If the founder refuses to reframe, the flag escalates from recoverable to structural and we treat the unwillingness as the signal, not the burn number.
9. Cohort retention undisclosed. The deck shows aggregate retention or a triangle chart and no cohort table. See Note 08. Recoverable because most seed-stage companies have the cohort data — they have not been asked to present it in cohort form. The recovery path is to ask, accept the table in any reasonable format, and read it together. If the founder produces the table and it looks healthy, the flag clears. If the founder cannot produce the table at all, the flag escalates.
10. NDR mis-defined. The founder claims net dollar retention above 120% and the definition does not survive thirty seconds of reconciliation. See Note 09. Recoverable because the actual retention number is usually still defensible — it is just smaller than the headline. The founder accepts the corrected number, the IC discounts the original claim, the conversation continues. If the founder defends the original claim past the reconciliation, the flag escalates.
11. Pre-screened reference list. The founder offers four references, all of whom are champions, none of whom have churned, none of whom were almost lost. See Note 06 for the field guide. Recoverable in one ask: name two customers who have churned, downgraded, or expressed material dissatisfaction, and route us to them. Founders with the relationship to ask for the introduction recover quickly. Founders who cannot — or will not — flag a relationship-with-customers problem that goes beyond reference selection.

The pattern across all four: the underlying business is fine. The flag is the founder's choice of framing, and the choice can be undone in a single conversation if the founder is willing. The diligence question is not whether the framing was used; it is whether the founder will drop it once the IC has the real number.

04How we score severity

Every flag we surface is scored on three axes, each on a 1–5 scale.

• Confidence. How certain we are, given the evidence in the deck and the public record, that the flag is real. A 5 is unambiguous from primary documents. A 1 is inferential — a pattern that fits but has not been directly confirmed.
• Reversibility. Whether the founder can neutralise the flag in the thirty days between dossier delivery and IC vote. A 5 is a one-call reframing. A 1 is a structural change requiring a cap-table reset or a co-founder transition.
• Materiality. Whether, if the flag is real and unresolved, it changes the round price. A 5 alters the equity split materially. A 1 is a footnote that adds nothing to the term sheet.

We sum the three. A composite above nine is a kill — the round does not price without resolution. Between six and nine is an IC condition; the round can price subject to the founder demonstrating the path. Below six is noted but does not block.

We publish the three component scores, not just the composite, because the components are arguable in different ways. A founder can dispute the confidence score by producing evidence. They can dispute the reversibility score by demonstrating progress. They cannot easily dispute the materiality score — that one is the IC's to make — but they can argue against the conclusion the IC draws from it. Publishing the components keeps the conversation honest, and gives the founder a specific thing to push back on rather than a verdict to argue with.

05The pattern behind the eleven

The eleven flags do not share a topic. They share a structure.

In each case, the founder has chosen a framing that protects the round at the cost of the relationship with the next-stage investor. The TAM gets inflated to clear a perceived bar. The burn gets adjusted to look like a different burn. The references get pre-screened to remove discordant voices. The platform dependency gets minimised to make the moat look real. None of these choices is unethical in isolation. Each is what a coached founder does to clear the first round.

The kill flag is rarely the framing itself. It is the founder's unwillingness to drop the framing when the IC challenges it. Recoverable flags are the ones where the founder drops the framing in the first call. Structural flags are the ones where the framing is the company.

This is why we do not score founders on intent. We score them on the response to the challenge. A founder who inflated the TAM and reconstructs it honestly in the second call is a different founder from one who restates the original claim more emphatically. The number on the slide is the same. The diligence outcome is opposite.

The reason eleven flags cover seventy-eight percent of material kills is that the fundraising script that produces them is small. Founders are reading the same handful of playbooks, encountering the same handful of fortifications. Naming the fortifications is most of the work.

06How the dossier presents them

Every dossier closes with a kill-flag section. Each surfaced flag receives a single line — flag name, severity composite, recommended IC condition — followed by a paragraph of evidence. The highest-severity flag appears first. The IC partner who reads only the closing section gets the worst news, the score, and the recommendation, in that order.

We do not hide flags in service of the founder. The founder benefits from knowing which framings the IC found unconvincing, even on dossiers that ultimately pass. The most common feedback we get from founders after a dossier-driven pass is that the surfaced flag was something they themselves were uncertain about and had not yet had a forcing function to address. The dossier becomes the forcing function.

For the four recoverable flags, the dossier includes the recovery path. Not the full script — the founder writes their own — but the structural ask. Reframe the burn number against net rather than adjusted. Produce a cohort table for the prior eight quarters. Restate the NDR with the standard formula and disclose the gap to the headline. Introduce two references who have churned or downgraded. The recovery path is the conversation we recommend the IC have in the next meeting, not after the term sheet.

The discipline of naming the flag is what separates the dossiers that close cleanly from the ones that drag for months. Every IC has been in the room where everyone agreed the deal was fine and no one could articulate what was wrong. The taxonomy is a tool against that meeting.