The Spreadsheet That Costs the Round.

Almost no seed round dies because a cap table was wrong. What happens instead is slower and more expensive: the process loses two weeks, the founder loses the initiative in the negotiation, and the round closes on terms that were set while everyone was waiting for a clean answer. In a US market where capital has never been more concentrated, that is a bad time to be the company that could not produce its own ownership numbers.

Why the tolerance for error has collapsed

The headline numbers from the first half of 2026 read like a boom. US startups took in $412.7 billion in H1, already ahead of the full-year 2025 total, per the PitchBook-NVCA Venture Monitor. Beneath the headline, the market is doing something else entirely. Megadeals of $100 million or more captured 87.5% of that capital. AI absorbed 86% of all venture dollars. Three firms took in nearly half of everything raised by managers, and first-time fund formation is tracking its weakest year since 2016.

Translate that into what it feels like at the seed stage. The pool of capital chasing a company outside the handful of favoured categories is smaller, the investors deploying it are fewer, and every one of them has more optionality than they did three years ago. When an investor has a full pipeline, they do not fight through a messy cap table. They deprioritise it, work the next opportunity, and come back if it is still open. Nothing is said out loud. The process simply cools.

The underlying issue is that a cap table is the only document in the raise that is a matter of fact rather than a matter of judgment. A projection can be defended. A market size can be argued. Ownership either reconciles or it does not, and an investor who finds it does not reconcile now has a question about everything else you have told them.

Where manual cap tables actually break

Spreadsheets are not the problem in themselves. Plenty of clean companies run on one for the first year. The problem is that a manual table has no memory and no enforcement: it records what someone remembered to type, in the format they chose, on the day they got to it. Six recurring failures account for most of what we find in a pre-raise review.

• Grants promised but never papered. An offer letter or a Slack message says 0.5%. No board consent was signed, no grant was issued, no vesting start date exists. The employee believes they hold equity. The cap table says otherwise. Both are right, and a lawyer now has to reconcile them.
• The ledger and the spreadsheet disagree. The stock ledger is the legal record. The founder's file is a working document. Once they diverge, every downstream number, including the 409A and the option pool, inherits the discrepancy.
• Vesting tracked by memory. Cliffs, early exercises, 83(b) elections, leavers who kept vested shares, leavers who did not. A departure eighteen months ago that was never processed is still sitting in the table as fully outstanding.
• Option pool arithmetic on the wrong base. A pool expressed as a percentage of issued shares is a different pool than one expressed on a fully diluted post-money base. That difference is worth points, and the confusion normally shows up at exactly the wrong moment.
• Instruments outside the file. A convertible note from an angel two years ago, an advisor warrant, a side letter granting pro rata rights. Each is enforceable and each was in someone's email rather than in the model.
• Version drift. Three files with similar names, two of them shared with investors, none of them marked as current.

None of these are exotic. Every one of them is discoverable in a morning by someone who knows what to look for, and every one of them is materially harder to fix once a term sheet is on the table and the clock is running.

The SAFE stack and the number founders get wrong

The single most consequential modeling failure at seed stage is the treatment of convertible instruments, because it is the one that changes what the founder owns rather than merely what the file looks like.

A typical US pre-seed company arrives at its priced round carrying several SAFEs signed at different moments, at different valuation caps, with different discounts, and often in a mix of pre-money and post-money forms. Each instrument converts on its own terms. Pre-money and post-money SAFEs allocate dilution differently, and the post-money form in particular fixes the investor's percentage in a way founders consistently under-model. Solving the stack correctly is circular: the conversion price depends on the share count, which depends on the conversion. Doing it by hand, under time pressure, at midnight, is how the error enters.

Layer the option pool on top and the gap widens. Carta's data puts median founding-team ownership at roughly 56% on a fully diluted basis by the time a company raises seed, falling to about 36% by Series A. A meaningful part of that second drop is not the new money at all: it is the pool top-up the incoming investor requires as a condition of signing, sized on the post-money base, funded entirely out of the pre-round holders. Founders who model the round without the top-up are modeling a transaction that is not on offer.

The practical consequence is a founder negotiating a valuation while holding the wrong ownership number. They optimise the headline figure, concede the pool, and discover the outcome after signing. The instrument that caused it was signed, in good faith, eighteen months earlier.

Pro forma is the document you negotiate with

There is a distinction worth being precise about, because it determines who controls the conversation. The current cap table describes what is outstanding today. The pro forma cap table describes what ownership becomes if a specific proposed round closes: every SAFE and note converted at its own terms, the new money in, the pool created or refreshed, and each holder's resulting percentage on a fully diluted basis.

Investors negotiate against the pro forma. If the founder cannot produce one, the investor's version becomes the only version in the room, and every subsequent point is discussed on their arithmetic. That is not adversarial behaviour on their part. It is simply what happens when one side has the model and the other does not.

A pro forma table built properly should let you answer, inside the meeting rather than a week later:

1. What do I own the day after close, fully diluted? Not on the current table. After conversion, after the new money, after the pool.
2. What does a 2% larger pool actually cost me? Expressed in points of founder ownership, not as an abstraction.
3. What changes if the round is upsized by $1 million? A question that gets asked in almost every process and answered well in very few.
4. Which SAFE holder ends up with the largest position, and is that what everyone expects? Caps signed a year apart produce outcomes that surprise both sides.
5. What does the table look like two rounds from now? Sophisticated seed investors are underwriting their own dilution path, not just today's entry.

Building that model is not a technology purchase. Cap table software maintains the record accurately and is worth having; it does not tell you which structure to accept. The judgment sits in the capital and conversion modeling work behind the file.

Waterfall modeling: the question nobody asks until it matters

The cap table tells you who owns what. The waterfall tells you who gets paid what, and the two answers diverge more often than founders expect.

Liquidation preferences, participation rights, seniority between series, conversion elections and, in some structures, dividends all sit between an exit price and a founder's proceeds. A 1x non-participating preference behaves very differently from a 1x participating one at a modest exit, and the gap between them is invisible on an ownership percentage. At a $40 million outcome, two companies with identical cap tables and different preference stacks can produce common-holder proceeds that differ by millions.

Waterfall modeling answers three questions that ownership percentages cannot: at what exit value does common stock begin to receive anything meaningful; where does the preference stack overhang the entire common base; and at what price does a preferred holder convert to common rather than take its preference. Those thresholds should be understood before the term is agreed, because after signing they are simply facts about your company. They also matter well before an exit, for anyone building a retention story around option grants: an employee holding options struck above the point where common begins to participate holds a number on a page, not compensation.

This is the same discipline we apply to financial planning and modeling more broadly, and it is worth doing at seed rather than discovering it during a sale process.

A pre-raise cap table checklist

What a founder should be able to hand over in one folder, on the day an investor asks:

• A current cap table reconciled to the stock ledger, with a date on it and a single owner responsible for it.
• Executed copies of every instrument: stock purchase agreements, SAFEs, convertible notes, warrants, side letters. Not summaries. Signatures.
• Board consents for every option grant, matched to the grant records, with vesting start dates and exercise prices that agree with the plan documents.
• The equity incentive plan and the authorised reserve, with remaining unallocated shares stated clearly.
• 409A valuations, current and historical, aligned to the grant dates they were used to support.
• A pro forma table at the target raise, with the conversion mechanics of each instrument shown rather than assumed.
• A one-page exit waterfall at two or three outcome levels, so you understand the structure you are being offered.

Companies that can produce that folder are not simply better organised. They are signalling something an investor cannot verify any other way: that the operating discipline behind the numbers is real. In a market where the capital available outside the top of the distribution is genuinely tighter, that signal is doing more work than it did three years ago.

The cost of getting this wrong is rarely a dramatic rejection. It is a slower process, a weaker negotiating position, and a founder who ends up owning several points less than they thought. Every point of it was decided by a spreadsheet nobody had time to check.

Questions we get asked

What is the difference between a cap table and a pro forma cap table?

The cap table records what is issued today. The pro forma models what ownership becomes after a proposed round closes, including every conversion, the new money, and the option pool the investor requires. Investors negotiate against the pro forma, so a founder who only has the current table is negotiating without the relevant number.

Why do SAFEs cause problems on a spreadsheet cap table?

Each SAFE converts on its own terms, and a stack signed over eighteen months usually carries different caps, different discounts, and a mix of pre-money and post-money forms that dilute existing holders differently. The maths is circular and unforgiving, and hand-built models are where the error enters. It surfaces at the priced round, when it is expensive.

Should we buy cap table software?

Generally yes, for record-keeping accuracy and audit trail. But software maintains the table; it does not decide what structure to accept, model the trade between valuation and pool, or tell you where the waterfall leaves common stock. Those are judgment calls, and they are the ones that move outcomes.