What each clause in a term sheet is actually worth

The situation

A term sheet is read for its headline price. The clauses behind it decide who is paid first, and what each shareholder takes home on a sale.

Founders compare offers on valuation, and lawyers negotiate clause by clause. Nobody puts the price and the clauses on the same scale.

Why it is hard

What a clause is worth depends on the sale price.

A preference matters when the sale is small. A cap matters only above a certain point, and participation at almost every price.

• Holders make choices. An investor with a preference converts to ordinary shares only when that pays more, and one series’ choice changes another’s.
• The staff option pool is usually counted before the new money comes in. It dilutes the founders, not the investor.
• Protection against a later, cheaper round changes the share count. How it is worked out moves real money.
• Documents are often silent on a point, such as how two series rank. The obvious reading is not always the right one.

The approach

We modeled every clause down to what each shareholder receives at each sale price. The same model ran across more than thirty transactions.

• The cap table is a record of events: the founding, each round, each convertible and the option pool. Holdings at any date come from replaying it.
• The waterfall pays debt first, then preferences in the order the documents set, then what is left.
• Every holder with a choice is tested: take the preference, convert or exercise. The model settles the choices until nobody gains by switching.
• It covers protection against cheaper rounds, both full and weighted-average. It also covers convertible notes, SAFEs, and pools created before or after the money.

Every result passes four checks. Payouts add up exactly to the sale price, and share counts reconcile.

Nobody receives less than zero, and nobody gains by switching.

Where the documents are silent, the model does not pick a reading. It shows the result under each one, so the silence becomes a point to negotiate.

Two versions of a deal are compared clause by clause, switching one clause at a time. Each holder’s difference is split across the clauses that caused it.

A worked example

This is a synthetic example, not client data. Two offers put in the same 10 million.

Offer A values the company at 40 million before the money. It has a pool of 10 percent and a plain preference.

Offer B values it at 45, 12.5 percent higher. But it has a pool of 15 percent, and the investor is paid twice on a sale.

Sale price	Founders, offer A	Founders, offer B	Investor, offer A	Investor, offer B	Founders lose
20	8.8	6.7	10.0	11.8	2.1
40	26.3	20.0	10.0	15.5	6.2
60	42.0	33.4	12.0	19.1	8.6
100	70.0	60.1	20.0	26.4	9.9
150	105.0	93.5	30.0	35.5	11.5

On a sale for 60 million, the founders receive 42.0 million under offer A and 33.4 under offer B. The higher offer leaves them 8.6 million worse off.

The higher price is worth 1.1 million to the founders. The bigger pool costs them 3.0, and paying the investor twice costs 7.0.

Under offer A, the investor takes its money back below a sale of 50 million and converts above it. Under offer B there is no such point: it takes its money back and a share of everything else.

The gap widens as the sale price rises, to 11.5 million on a sale for one hundred and fifty. Offer B never pays the founders more, at any price.

Where it breaks

• It assumes the company is sold. In a stock market listing, most of these clauses fall away.
• The split of a difference across clauses depends on the order they are switched. Where two clauses interact, the joint effect lands on the later one.
• Break points are found by stepping through sale prices, so a very narrow band can be missed.
• Some events are not modeled yet: secondary sales, buybacks, dividends, redemption rights and drag or tag rights.
• The sale price is taken as the value of the shares. Any debt has to be taken off first.

If you are comparing offers, compare what each leaves you at a realistic sale price.