Researchers quantify the board-strategy gap Elliott alleged at Norwegian Cruise Line

Contextualized word embeddings compare two bodies of text: the language of a director's accumulated professional experience and the language a company uses about its own strategy ^[2]. Rather than asking whether a director belongs in an "Operations" box, the measure asks how closely that person's professional identity sits to what this particular company says it is doing ^[2]. DesJardine and Mertens report two readings on NCLH's pre-campaign board: it sat further from its own strategic identity than the boards of its closest peers did ^[3], and its directors sat unusually close to one another ^[4]. A skills matrix can show a dozen filled categories and register neither result ^[15].

Both readings run on material that was already public. The skills matrix the authors use as their worked example comes from NCLH's own 2025 proxy statement ^[12], and director tenures and titles are what search firm databases already catalog ^[16]. Thirty-seven days separated Elliott's February launch from the March settlement ^[18]. That is the window in which a board answers an argument about fit, and the answer has to be built from work done well before the window opens.

The trade-off is between two kinds of measurement, not measurement against none. Skills matrices are cheap, comparable across filers and legible to the readers who grade them, with New York Stock Exchange listing standards nudging companies toward the disclosure and ISS and Glass Lewis scrutinizing what gets filed ^[11][13]. A fit score answers the question the matrix cannot ^[15], and once computed it exists as a number an opponent can quote back. A board that runs the diagnostic buys information and creates discoverable evidence in the same act; a board that does not leaves the first computation to whoever is buying the stock.

The measure improves after a settlement whose incoming directors an activist selected, so it risks scoring Elliott's judgment against Elliott's own critique. The reported post-settlement narrowing ^[5] is the weakest part of this record for exactly that reason. The load-bearing finding is the pre-campaign peer comparison ^[3], which was computable from filings before anyone knew who the replacements would be. The summary reports the two gaps separately and does not put them on a shared scale, so the tidy formulation that the board resembled itself more than its strategy is not something these results, as described, establish ^[3][4].

The record shows no indication that Elliott ran anything of this kind. Its 59-page document argued in words that the board lacked industry expertise, strategic alignment and fresh perspective, and that the stock had trailed both the market and close peers, with Royal Caribbean outpacing NCLH's growth by roughly 400% over five years on Elliott's account ^[6][7][8]. The embedding analysis carries a September 8, 2026 dateline ^[21], 166 days after the settlement was struck ^[20]. One campaign at one company, measured by the method's own authors ^[1], is a demonstration rather than a validated screen, and the authors' claim that directors and investors can now build better boards with it ^[17] is a claim about what the tool permits, not evidence that it predicts anything.

That still changes the calculation for a nominating committee this quarter. The cost of computing your own fit score is a number you may not want; the cost of not computing it is that the first version of that number belongs to someone accumulating a stake above 10% ^[6]. If the score is poor and the refresh gets deferred anyway, the deferral is the thing the score documents next year.