Contrary to popular belief, stock exchanges only handle the execution of a trade. The subsequent critical steps of clearing (agreeing on the trade details) and settlement (exchanging money for assets) are handled by separate back-office functions between the involved parties.
Traders executing large "block" trades deliberately hide their full intent. They break up orders across time, venues, and brokers to prevent the market from detecting their strategy, which would cause prices to move against them before the full trade is complete.
Beyond regulatory requirements, some financial firms enforce extreme information silos between trading desks to protect proprietary strategies. This can extend to physical separation and even pumping in pink noise to prevent teams from overhearing each other's conversations and leaking "alpha."
For a large investment bank, knowing its net position in a stock isn't simple. It involves aggregating direct holdings, client positions, and synthetic exposures from derivatives like swaps and structured notes, all of which react differently to price changes.
Systems are often designed for a single, automated workflow. When a human deviates—like two traders working the same order from different windows—it can trigger unexpected errors. This can lead to massive unintended positions, as the system rejects one part of the trade while another proceeds incorrectly.
Despite its zero-sum reputation, finance operates on high-trust relationships. When trade details are disputed, the first step isn't a formal process but an informal negotiation ("split the difference and shake on it") to preserve the professional relationship, which is more valuable than winning a single transaction.
Core back-office processes for reconciling trades between firms often depend on archaic technology like FTPing files. These systems are fragile; file formats can unexpectedly change based on the number of asset classes traded, requiring teams of people to manually verify and fix data pipelines daily.
When two banks can't agree on a final number, it's not a math error, but a data integrity problem. With numerous system hops (exchange, gateway, trading system, booking system), a software bug can flip stock symbols or corrupt data at any point, leading to mismatched realities that can take years to resolve.
During the 2008 crisis, Goldman Sachs could quickly assess its total net exposure to any asset because it had a single, monolithic database for all positions. Competitors, cobbled together from acquisitions, had disparate systems, making it slow and difficult to get a clear picture of their risk during a fast-moving crisis.