ON INSTITUTIONAL FAILURE

The Signal Was There. The Attention Was Not.

Three organizations. Three industries. Three decades apart. Each possessed the information needed to survive. None built the architecture to act on it in time.

On Thursday, March 9th, 2023, depositors withdrew $42 billion from Silicon Valley Bank (SVB) in a single day. The bank was seized by regulators the following morning. It was the fastest bank run in American history — and arguably the most foreseeable.

SVB's annual report, filed with the SEC just weeks earlier, disclosed that the bank held more than $91 billion in long-duration bonds purchased when interest rates were near zero. The Federal Reserve had been raising rates aggressively for a full year. The math was available to anyone who read the filing: SVB was sitting on unrealized losses exceeding $15 billion in a bond portfolio it had classified as "held to maturity"— a designation that allowed the losses to stay off the income statement, but not off the balance sheet. Their depositor base was 97% uninsured, overwhelmingly concentrated in VC-backed startups.

A correlated risk profile that meant the moment one founder pulled funds and sent a message to their network; the whole edifice would move together. The bank's Chief Risk Officer position had been vacant for eight months. Credit rating agencies were in active discussions about a downgrade. The signals were not hidden. They were in the public record, available to anyone paying attention.

Some investors were. Founders Fund had quietly withdrawn its deposits before the public run began. The signals reached some decision-makers. They did not reach the ones who could act on them before the window closed.

The most uncomfortable finding from decades of institutional failure analysis is not that organizations lacked foresight. It's that they possessed it and could not act. The signal was there. The attention was not.

RAJEEV RONANKI

Call it informed collapse: the specific failure mode in which an organization — or its regulators, its board, its risk committee — possesses accurate, timely information about an emerging threat and still cannot convert that information into action before the window closes. It is not a failure of intelligence. It is a failure of architecture, with a clear absence of systems designed to ensure that the right signals reach the right decision-makers while there is still time to act.

SVB is a recent and vivid example, but the pattern is neither new nor confined to banking. In 1975, a Kodak engineer built the world's first digital camera and brought it to management with a simple observation: this technology would eventually replace film. Kodak's strategic planning documents from the 1980s accurately forecast that digital would displace film by 2010. Their market intelligence operation conducted, by one account, some of the most comprehensive competitive analyses ever performed by a corporation of that era.

The signals were present for decades. The institutional capacity to act on them was not. By 2012, Kodak filed for bankruptcy.

The pattern recurs with such regularity that it begins to look less like coincidence and more like a structural property of how large organizations process information under conditions of uncertainty and competing priorities. The question is not whether your organization has the signals. It almost certainly does. The question is whether they are reaching the people who can act, and whether they are reaching them in time.

346 Deaths. Five Months of Visible Signals.

On October 29th, 2018, Lion Air Flight 610 crashed into the Java Sea minutes after takeoff from Jakarta, killing all 189 people aboard. Investigators quickly identified the cause: a newly installed flight-control system called MCAS had repeatedly pushed the nose of the aircraft down in response to a faulty sensor reading. The pilots, who had not been trained on MCAS, could not override it in time.

Over the following five months, the signals accumulated. International aviation regulators circulated safety bulletins. Pilot unions raised concerns about the inadequacy of MCAS training. Boeing's own internal communications, later obtained by Congressional investigators, showed engineers who were alarmed. One employee described the aircraft as "designed by clowns who in turn are supervised by monkeys." Another wrote that he would not put his family on the plane.

On March 10th, 2019, Ethiopian Airlines Flight 302 crashed under nearly identical circumstances, killing all 157 people on board. The 737 MAX was grounded two days later — five months and 346 deaths after the first crash had made the failure mode visible.

This is informed collapse at its most consequential. The information existed. Engineers had it. Regulators had it. Pilot representatives had it. What failed was not the intelligence function. What failed was the institutional architecture for converting distributed, cross-functional signal awareness into coordinated, timely action.

The Slow-Motion Version: Credit Suisse

Not every informed collapse happens at SVB speed. Credit Suisse took years to fall, and the signals were visible at every stage.

In March 2021, the bank disclosed a $5.5 billion loss from the collapse of Archegos Capital Management: a family office that had taken on extraordinary leverage through derivative positions that Credit Suisse's own risk systems had failed to flag as dangerous. In the same month, the bank's exposure to Greensill Capital, a supply-chain finance firm that collapsed amid allegations of fraud, became public. Two major risk failures in the same quarter.

What followed was two years of visible deterioration that any observer with access to public information could track. Client withdrawals accelerated, including CHF 84 billion in the fourth quarter of 2022 alone. The bank's own annual report acknowledged "material weaknesses" in its internal controls over financial reporting. Credit default swap spreads had been widening for months, the market's real-time assessment of default probability, visible to every institutional investor covering the sector. When the Saudi National Bank publicly stated it would not provide additional capital in March 2023, it was the match dropped into a room that had been filling with gas for two years.

UBS acquired Credit Suisse in an emergency government-brokered transaction on March 19th, 2023, at CHF 3 billion — a fraction of the bank's stated book value. The signals of this outcome had been accumulating, publicly, for 24 months.

What These Three Have in Common

SVB, Boeing, and Credit Suisse operated in different industries, at different speeds, with different failure modes. What they share is not the nature of the signal. It is the architecture that failed to process it.

In each case: signals existed across multiple organizational domains that, connected, would have demanded immediate action. In each case: those signals were processed in silos, by different teams, on different reporting cycles, in different dashboards, and the connection that would have revealed the systemic risk was never made before the window closed. In each case: the failure was not a lack of intelligence. It was a structural deficiency in how the organization converted distributed signal awareness into coordinated institutional response.

The Diagnosis Most Organizations Resist

The instinct, when confronted with these cases, is to identify the individual failures: the SVB executive who didn't hedge the bond portfolio, the Boeing manager who overrode the engineers, the Credit Suisse risk committee that cleared the Archegos positions. These individuals exist and their decisions mattered. But focusing on them is a way of avoiding the harder, more important diagnosis.

The deeper pattern in every case of informed collapse is not individual failure. It is the absence of a shared substrate for connecting signals across organizational boundaries, weighing them against each other, and surfacing the resulting picture to the people who need it — automatically, in time, with enough context to act.

The signals of the next institutional collapse are present somewhere right now. Perhaps in a financial filing, in an engineering report, in a customer behavior trend, or in a regulatory communication. They are mixed with noise, competing for attention, decaying as the decision window narrows. The organizations that build the infrastructure to connect those signals and route them to decision-makers before the window closes will avoid becoming the next case study.

The signal was there. The architecture to act on it was not.

That is the problem worth solving.