Hypothetically, assume a company whose valuation increases roughly geometrically at first, but then levels off at an asymptote, and assume that during the geometric phase the valuation fluctuates by 50% around its trend. Assume that no one knows where the asymptote is. Periodically, there are funding events, during which the company sells stock at a 1x liquidation preference; assume it sells enough stock so that at each funding event, the sum of the liquidation preferences is greater than 50% of its valuation. Consider a funding event that occurs when the valuation happens to be 25% above trend; therefore the liquidation preferences will be greater than 0.625 of the on-trend valuation, and then the next time that the valuation falls to 50% below trend, the sum of those preferences will be greater than the company's current valuation. At this point, many of those with liquidation preferences (some of whom may have a short time horizon; or may be risk-averse; or may believe that the current valuation will not rise very much in the future) may strongly prefer selling or liquidating the company immediately, which is at odds with common-stock-holders, who would get nothing in such a liquidation and who would prefer to keep going.<p>Note that in this model, this result occurred not because of mismanagement, but due to the natural and expected fluctuation in market prices, combined with unavoidable uncertainty about which price changes are fluctuations and which are simply 'the new normal', combined with the sum of liquidation preferences being comparable in magnitude to the fluctuations in prices, combined with the bad luck of a funding event happening to occur when the valuation was above trend.<p>Is this model reasonable? If so, how should management of such a company approach fundraising? It seems to me that fluctuations are inevitable and that the only thing management could control is how much money they raise at once; if fluctuations are 50% around trend, then management needs to keep the amount of money raised small enough so as to keep the sum of liqudiation preferences well under 1/3 of the current valuation (because in that case even if the valuation falls from 50% over trend to 50% under trend, the valuation will still be strictly greater than the sum of liquidation preferences). More realistically, the size of fluctuations would not be known in advance to be 50% but must be estimated. In the article, the valuation fluctuated from $1bil to $250mil, so the fluctuation parameter would be at least (1-x)/(1+x) = 250/1000 = 4 -> x = 3/5 = 60%, and the sum of liquidation prefs "should have" been kept well less than 250/1000 = 1/4 of the current valuation.