Olga Magomedova
6 min readBy Olga Magomedova

The partner class that stopped counting

On 5 November Goldman Sachs named its largest partner class since 2010 and declined to publish a gender breakdown. Olga Magomedova on what disappears when a firm stops counting the structure that produces its traders, and why the household ledger still has to.

A larger class and a smaller disclosure

On 5 November Goldman Sachs elevated 109 bankers to its partner ranks, the largest class since 2010. The announcement included a figure on how many new partners began at the firm as interns and a figure on how long the typical route to partner takes. It did not include a figure on how many are women. Bloomberg's own count of the names put the share at fewer than one in five, the lowest since 2014. In 2024 the equivalent share was 27 per cent.

An executive cited a changed legal environment as the reason the firm no longer publishes the demographic breakdown it had published in prior years. The bank released other metrics about the class. It released the share working in revenue-producing divisions. It released the share who started as interns. The variable that went missing is the one the rest of the industry has spent a decade tracking.

What a missing variable actually means

In an engineering file, a variable that stops being reported is not treated as unchanged. It is treated as unknown. The honest response to a missing reading is to widen the tolerance until the reading returns. The dishonest response is to assume the last known value still holds.

The same logic applies to a promotion pipeline. If a firm counted its intake by gender for a decade and then stops counting, the reasonable prior is not that the number held steady. The reasonable prior is that the direction of travel is the reason the measurement stopped. The outside counts suggest exactly that.

For Magomedova, this is a familiar failure mode. A system under stress drops the measurement that most reveals the stress. The measurement is then restored only after the behaviour it was tracking has already changed.

Why the structural question is a performance question

The same week produced a counterweight worth sitting with. The sixth annual Investing in Women Code report in the UK found that signatories directed 32 per cent of their venture capital investment value to teams with at least one female founder, against a wider market rate of 15 per cent. Angel investor groups with more than 30 per cent women investors directed 98 per cent of their investments to women-led or mixed-gender founding teams. Code membership has grown from 12 organisations in 2019 to more than 330.

Those are not fairness numbers. They are allocation numbers. They describe where capital actually went when the people deciding were not all the same. The firms that kept counting ended up making different trades.

A desk that rewards adding risk after a winning streak will keep producing traders who add risk after a winning streak. A firm that stops counting who gets promoted will keep producing the composition the counting was designed to change.

What the household ledger still has to do

None of this changes the arithmetic facing a woman building a trading practice at home. The institutional pipeline will do what it does. The household ledger is still the ledger she controls. The account balance at the end of the quarter is still the honest report card.

The practical response to a firm that stops counting is to count more carefully at the desk. Position size written down before the trade. Drawdowns logged against the plan they were supposed to stay inside. A short post-mortem after any unexpected behaviour. The artefacts of discipline do not depend on anyone else's disclosure policy.

They are also the only artefacts that travel. An institutional title is a loan from an institution. A documented process is an asset that belongs to the trader.

Life can change at any moment. Your independence and skills are the only foundation you can truly rely on.

The principle that does not move

The partner class will be announced again in two years. The number may be higher. The number may be lower. The disclosure may return or it may not. None of those outcomes should alter the work of a trader who has decided that her independence is not contingent on an institution's reporting choices.

The structural argument for more women in markets remains the performance argument it has always been. The Code data in the UK is one more piece of evidence that diverse decision rooms allocate capital differently, and measurably. The partner class story is one more reminder that the institutions holding the scoreboard can decide to stop showing the score.

The response is not outrage. It is a ledger that does not depend on their scoreboard. Build the process before you need it. Keep the measurements the firm declines to keep. The desk is the one place where the counting cannot be quietly dropped, because the account at the end of the year will have counted anyway.