The sale was fair. The common stockholders got nothing.

Marc Christen owned about 5% of the common stock of Trados, a software company that took venture capital in 2000 on a plan that could end in a public offering. It ended in a sale instead. In July 2005, SDL plc bought Trados for $60 million in cash and stock. The investors who held preferred stock received $52.2 million. The common stockholders, Christen among them, received nothing.

Not everyone who worked there left empty-handed. A management incentive plan took the first 13% of the price, $7.8 million, and paid it to managers and other employees before either class of stock saw a cent.

Christen sued the directors who approved the deal. The case ran until August 2013, when Vice Chancellor Laster of the Delaware Court of Chancery ruled. You might expect one of two endings: a squeeze punished, or a clean deal waved through. He wrote something stranger than either.

On how the deal was run, the judge leaned toward Christen. The evidence on fair dealing, he wrote, "decidedly favored the plaintiff," and the structure of the incentive plan was itself "evidence of unfair dealing towards the common stock." Because a majority of disinterested, independent directors had not approved the merger, the directors carried the burden of proving it entirely fair. On price, the evidence was "mixed."

They proved it anyway, and the reason was a single number in Trados's charter. Under it, the sale counted as a liquidation, and the preferred stockholders were owed a liquidation preference of $57.9 million. Against a $60 million price, the common had almost nothing to lose. In the court's words, the common stock "had no economic value before the Merger," and the directors "breached no duty to the common stock by agreeing to a Merger in which the common stock received nothing." On Christen's appraisal claim, the fair value of the common was zero.

Even the incentive plan, the part the court held against the directors, came mostly out of the investors' pockets. By the court's arithmetic, the common effectively paid $2.1 million of it and the preferred $5.7 million.

So the fight over who got paid was not settled at the closing table in 2005, or in court in 2013. It was settled when the preference was written. That is the idea worth carrying into any financing: every preferred stock with a preference draws a line, a sale price below which the common is worth nothing. At Trados the line was $57.9 million, and the company sold just above it. The price of a round is the number everyone repeats. The line is the number that decides whether common stock ever pays.

Paul Graham, who co-founded Y Combinator, has told founders: "The most important thing to understand about valuation is that it's not that important." Trados shows one way that can be true. A valuation says what the company was worth on the day of the round. The preference says who is paid first on the day it is sold.

The basic version is a 1x preference: the investors get back what they put in before the common gets anything. Brad Feld, a venture investor, noted that for many years 1x was the standard. What decides the rest is one word, participating. A non-participating investor chooses: take the money back, or convert to common and share like everyone else, whichever pays more. A participating investor takes both, the money back first and then a share of what remains.

To see the line move, take a company that raised $10 million for 40% of itself, a $25 million value after the money went in. Leave out any option pool, debt, fees and dividends, and sell it three times: for $5 million, $12 million and $50 million.

At $5 million, the price is below the line, and the common gets nothing under either version. That is Trados's position in miniature: a company that found a buyer and still paid one class of owner zero. At $12 million, just above the line, the common gets $2 million, or $1.2 million if the preference participates. At $50 million, the non-participating investors convert and take $20 million, leaving $30 million to the common. With participation they take $26 million, and the common's share falls to $24 million. The single word moved $6 million.

There is a second line hidden in the table. The non-participating investors convert only when 40% of the price beats their $10 million, which happens above $25 million, exactly the value they bought in at. Below that, they take their money back. Above it, they are shareholders like everyone else. A participating preference never makes that switch: the investors stand at the front of the line even in the big win.

The line is not fixed at signing. Every new round of preferred stock with a preference adds its own amount to it. In the example, a second round as large as the first, with its own 1x preference, would double the line, and the $12 million sale would pay the common nothing. That is also why the model charter lets existing preferred holders veto any new stock that would rank ahead of them. The investors already in line get a say over who may step in front.

Anti-dilution decides the split above the line when a later round is priced lower than an earlier one. The common version, broad-based weighted average, lowers the investors' conversion price by a formula that weighs the cheaper new shares against everything already outstanding, options included, so a small down round moves it a little. The full ratchet resets the conversion price to the new, lower price outright. Sell even a sliver of stock at half the old price, and the earlier investors convert as if they had paid half too. The common holders absorb the difference.

Neither clause shows up in a good year. Both decide, in a bad one, how much of the company the common still owns when a buyer arrives.

Near the line, a sale now returns the preferred's money and leaves the common nothing; only a higher price later would pay the common. On the same day, the two classes can want different things. That is why the board matters. The directors approve a sale, its price and its buyer. At Trados, the absence of a disinterested, independent majority is what put the burden on the directors to prove the deal fair.

Brad Feld wrote, months before the Trados sale, a sentence that describes the pressure in this case: "The greater the liquidation preference ahead of management and employees, the lower the potential value of the management / employee equity." Set beside Trados, it describes what happened. With the common worth nothing, the incentive plan was how the managers were paid in the sale. The court did not condemn the price. It did find that the plan's structure was evidence of unfair dealing toward the people it left out.

Every term sheet can be tested before anyone signs it, and the test takes an afternoon. Find the line: the total preference ahead of the common. Pick three sale prices, one below it, one just above it, one far above. Work out what each class receives. Then do it again with participation, with a ratchet, and with the next round's preference stacked on top. The gaps between the answers are the price of the clauses, and they are visible long before there is a buyer. Watch the margin above the line, too, because anything paid off the top comes out of it. At Trados the margin between the preference and the price was thin, and the incentive plan alone was bigger than it.

It took Christen's case until 2013 to explain what the charter had said all along. The preference was the price at which his 5% became worth something, and Trados sold too close to it. If your company sold tomorrow for a number that sounds like success, would it clear your line, and by how much?