The Typical Sound

For years, the Swedish Navy operated on a piece of acoustic evidence they trusted completely. Hydrophones planted along the coastline occasionally picked up a specific noise, and when they did, the incident was classified as a probable Soviet submarine. Analysts called it the “typical sound.” It became the deciding factor in a national security assessment that ran for decades, shaping naval posture during some of the tensest years of the Cold War.

The typical sound was not a submarine. It was herring.

Specifically, it was the noise produced by massive schools of herring releasing bubbles, a phenomenon researchers Magnus Wahlberg and Håkan Westerberg eventually traced and confirmed. The sensor had been picking up something real the entire time. It just was not the thing anyone thought it was.

This is worth sitting with, because the failure here is not the one most organizations assume it is. The equipment worked. The recordings were accurate. Nobody fabricated data or ignored their instruments. The failure happened one layer up, in the gap between what the sensor detected and what the organization decided it meant.

One indicator became the whole answer

Here is how the reasoning collapsed:

An indicator (typical sound detected) got treated as identical to a conclusion (submarine present). Once that equivalence was accepted, no other input was required. Confidence did not build gradually as corroborating evidence accumulated. It arrived instantly, attached to a single acoustic signature.

That is the mechanism worth naming, because it shows up constantly in organizational life, usually with far less drama than a Cold War naval standoff. A spike in support tickets gets read as “the client is at risk.” A sudden burst of employee questions gets read as “morale is slipping.” A single missed deadline gets read as “this vendor is failing.” In each case, something real was detected. The question is whether the organization let that one detection stand in for the underlying judgment it was actually trying to make.

Wahlberg, describing the psychology behind the Navy’s error, put it plainly: once authorities tell you how things are, you start collecting evidence that confirms it. That is not a comment about faulty sonar. It is a comment about what happens once an indicator becomes a story. New observations stop being tested against the hypothesis and start being sorted to support it.

Why a dashboard is not just more metrics

The instinct after a failure like this is often to add more sensors, more hydrophones, more data points. But volume was never the problem. The Navy had plenty of data. What it lacked was a structure that required different signals to be interpreted against each other before a conclusion was reached.

A dashboard, used correctly, is not a scoreboard that tells you red, yellow, or green. It is a discipline that forces a different sequence of reasoning:

Signal, then context, then corroboration, then assessment. Not signal straight to conclusion.

If the Navy had applied that sequence to the typical sound, the acoustic signature would have prompted a question rather than a classification: given this signal, what else should we expect to see if this really is a submarine? Radar contact. Consistent vessel movement. Independent intelligence. A pattern that held up across multiple sources rather than one.

The same test applies to an account health dashboard, a workforce risk dashboard, or any governance tool built to answer a question as consequential as “is this healthy.” Utilization, service volume, escalation rate, response time, renewal signals, none of these are wrong to track. The risk is in letting any single one of them answer the question on its own.

The harder discipline is what to do when one indicator disagrees

The genuinely useful moment is not when five indicators move together. That is easy to interpret. The harder moment is when one indicator diverges sharply from the rest, and the organization has to decide what that divergence means.

Sometimes a single outlier indicator is the earliest warning available, arriving before the rest of the dashboard has caught up. Sometimes it is herring. The only way to tell the difference is to have already built the habit of asking what corroborating evidence would need to exist before the indicator is allowed to become a conclusion.

That habit is a governance question, not a technology question. It has to be decided in advance, as a standing rule for how the organization interprets its own data, rather than improvised in the moment by whoever is closest to the alarm. Organizations that wait until the signal fires to decide how much weight it deserves tend to get the answer the signal wants them to get.

The Swedish Navy did not fail because their instruments were bad. They failed because nobody had built a structure requiring the instrument to be corroborated before it was believed. That is the infrastructure question underneath the story, and it is the one worth asking before the next single indicator starts sounding very certain of itself.

The Swedish submarine hunt and the herring-bubble discovery are covered in Radiolab’s 2021 episode “Red Herring.”