Case Study 1: Soviet Factories and Body Counts — When Numbers Replace Reality

"Not everything that counts can be counted, and not everything that can be counted counts." — attributed to William Bruce Cameron

Two Systems, One Pathology

This case study places two of the most consequential manifestations of Goodhart's Law side by side: the Soviet planned economy's metric-driven production system and the American military's body count strategy in Vietnam. These systems operated on different continents, in different decades, in different institutional contexts, and in pursuit of different objectives. Yet they broke in structurally identical ways, for structurally identical reasons. The parallel is not a metaphor. It is a diagnosis.

The Soviet Planning System: Metric as Master

The Problem of Coordination at Scale

The Soviet Union's centrally planned economy faced a genuine and enormously difficult coordination problem. In a market economy, prices serve as distributed signals that coordinate production and consumption without any central authority needing to understand the whole system. In a planned economy, that coordination mechanism is absent. Someone -- specifically, Gosplan, the State Planning Committee -- must decide how much of everything to produce, where to produce it, how to distribute it, and what it should cost.

This is an information problem of staggering proportions. The Soviet economy at its peak encompassed some 200,000 enterprises producing millions of distinct products. No committee, no matter how large or talented, could directly observe the quality and quantity of output across this vast system. The planners needed metrics -- standardized, quantifiable, reportable indicators that could travel from factory floor to Moscow and back.

The metrics they chose were the metrics available: physical quantities (tons of steel, meters of fabric, units produced), value aggregates (rubles of output), and input ratios (labor per unit, material per unit). Each factory received a plan -- a set of target numbers it was expected to hit. Fulfilling the plan brought bonuses, promotions, and political safety. Failing the plan brought scrutiny, reassignment, and worse.

The Taxonomy of Gaming

What followed was not a single failure but an entire ecosystem of gaming strategies, each a predictable response to the specific metric being used.

When output was measured by weight: The nail factory story from the main chapter is the most famous example, but the pattern was ubiquitous. Chandelier factories, measured by weight, produced chandeliers so heavy they tore the ceilings from apartment buildings. Glass factories, measured by tons, produced glass so thick it was nearly opaque. Sheet metal factories rolled metal so thick it could not be shaped into the products it was intended for.

When output was measured by quantity: Factories that had been producing too-heavy products swung to the opposite extreme. The nail factory's tiny nails are the canonical example, but similar dynamics appeared across industries. Textile factories measured by meters of fabric produced fabric so narrow it was useless for clothing. Shoe factories measured by pairs produced shoes in only the easiest sizes to manufacture, leaving consumers with the wrong sizes.

When output was measured by value: Factories discovered that the ruble-value of their output could be maximized by using expensive inputs. A furniture factory measured by ruble output would use expensive hardwoods rather than cheaper, perfectly adequate materials. The metric rewarded waste.

When output was measured by number of product types: Some directives required factories to produce a specified number of different products. Factories responded by producing trivially different variants -- the same product in a different color, or with a minor cosmetic change -- to check the "variety" box without the expense of genuine diversification.

In every case, the pattern was the same. The metric captured one dimension of what the planners actually wanted. The factory optimized along that dimension, ignoring -- and often actively sacrificing -- every other dimension of quality, usefulness, and efficiency.

The Ratchet Effect

The gaming was compounded by a second pathology known as the ratchet effect. Each year's plan was based on the previous year's performance. If a factory exceeded its quota in year one, year two's quota would be set higher. Factory managers quickly learned to avoid exceeding their quotas by too much -- to produce just enough to meet the plan but not so much as to raise next year's expectations. A factory capable of producing 120 percent of its quota would deliberately produce only 102 percent, hiding its excess capacity from the planners.

The ratchet effect is Goodhart's Law applied to the metric of year-over-year improvement itself. The planners wanted steady improvement. The metric (last year's output as the baseline for this year's target) was supposed to drive improvement. Instead, it drove concealment.

Why the Planners Could Not Fix It

The most instructive aspect of the Soviet planning system is not that gaming occurred -- that is predictable -- but that the planners could not stop it despite knowing it was happening. Soviet economic journals were filled with discussions of these problems. Planners experimented with different metrics, different combinations of metrics, different enforcement mechanisms. They added quality inspectors. They introduced customer satisfaction measures. They tried multi-metric plans with dozens of simultaneous targets.

None of it worked permanently. Each new metric was gamed. Each new enforcement mechanism was circumvented. The fundamental problem was structural: the distance between the planners in Moscow and the factories in Novosibirsk was too great, the information asymmetry too severe, and the incentive to game too powerful. No metric -- and no combination of metrics -- could fully substitute for the thing the planners actually wanted: useful products, efficiently produced, that met the needs of Soviet citizens.

The Body Count: Vietnam's Metric of Doom

The Same Structure, Higher Stakes

The American military in Vietnam faced a coordination problem with the same structure as the Soviet planners' problem, but with lethal stakes. Senior commanders and civilian leadership in Washington needed to know: are we winning?

In a conventional war, the answer is visible on a map. Territory advances, strategic objectives are captured, enemy forces surrender. Vietnam offered none of these legible indicators. The war was a counterinsurgency fought in jungles, rice paddies, and villages, against an enemy that blended into the civilian population, abandoned territory willingly, and measured success by endurance rather than by conquest.

Robert McNamara, Secretary of Defense from 1961 to 1968, approached the problem the way he had approached management at Ford Motor Company: with data. If you cannot see progress directly, measure it. If you cannot measure it directly, find a proxy. The proxy McNamara and the military command structure settled on was the body count.

The Corruption Cascade

The body count incentive structure created a cascade of corruptions, each flowing logically from the metric's deployment as a target.

Inflation. Units under pressure to report high kills reported generously. A patrol that encountered enemy fire and found no bodies afterward might estimate enemy casualties based on the volume of fire exchanged. A commander reviewing these estimates had an incentive to accept the highest plausible number, because high numbers reflected well on his unit. The estimates cascaded upward through the chain of command, inflated at each level.

Reclassification. Civilian casualties were reported as enemy combatants. In a war where the enemy wore no uniforms and operated among civilians, the distinction between combatant and civilian was often genuinely ambiguous. The body count metric resolved this ambiguity in a predictable direction: when in doubt, count the body as an enemy kill. This practice was so widespread that the phrase "If it's dead and Vietnamese, it's VC" entered the soldiers' lexicon.

Tactical distortion. The most dangerous consequence of the body count metric was its effect on tactical decision-making. Because body counts determined officer evaluations, promotions, and unit reputations, some commanders chose tactics that maximized kills rather than tactics that achieved strategic objectives. Search-and-destroy missions were favored over the less dramatic but strategically more important work of building relationships with local populations, securing infrastructure, and disrupting supply lines. Air and artillery strikes were called in aggressively, because destroyed areas could be claimed as enemy casualties, even when the strikes killed civilians and generated precisely the popular resentment that fueled the insurgency.

Strategic delusion. The cumulative effect of inflated body counts was that the American military and civilian leadership believed they were winning a war they were losing. The body count numbers told a story of overwhelming American military superiority. The Tet Offensive -- a coordinated attack by enemy forces across the entire country -- shattered this narrative not because it was a military success (the attacking forces suffered enormous casualties) but because it demonstrated that the enemy retained the capacity for large-scale operations despite years of body counts that suggested they should have been depleted. The metric had built an alternate reality, and the real reality broke through it.

The Structural Isomorphism

Place the two systems side by side, and the shared structure is unmistakable.

The key structural features are identical:

1. Distance between principal and agent creates information asymmetry.
2. A proxy metric substitutes for the unobservable reality.
3. Incentive attachment ties the agent's welfare to the proxy.
4. Gaming exploits the gap between proxy and reality.
5. The proxy decouples from the underlying reality under optimization pressure.
6. The principal is deceived by the gamed metric and makes poor decisions as a result.

Where the Parallel Deepens

Both systems also exhibited the same secondary pathologies.

Ratchet effects in both. In the Soviet system, exceeding the quota raised next year's target, so managers hid capacity. In Vietnam, a unit that reported high body counts was assigned more aggressive missions, exposing soldiers to greater danger. Officers learned to report numbers that were high enough to look good but not so high as to attract assignments that would get their soldiers killed.

Punished honesty in both. Soviet factory managers who reported production difficulties were penalized, creating incentives to conceal problems until they became catastrophic. Officers in Vietnam who reported low body counts or questioned the metric's validity were sidelined. Colonel David Hackworth, one of the war's most decorated officers, became persona non grata after publicly criticizing the body count system. Honest reporting was not just unrewarded -- it was punished.

Systematic self-deception in both. The most corrosive consequence in both cases was not the gaming itself but the way gamed metrics fed back into decision-making. Soviet planners, reading reports of quota fulfillment, believed the economy was healthier than it was. American leaders, reading reports of enemy casualties, believed the war was going better than it was. In both cases, the gamed metric became the basis for further decisions, creating a feedback loop of escalating delusion.

Connection to Chapter 2 (Feedback Loops): Both systems exhibit the same positive feedback loop: optimistic metrics lead to optimistic decisions, which lead to continued commitment to the failing strategy, which leads to continued pressure for optimistic metrics. The gamed metric does not just fail to detect the problem. It actively conceals the problem from the people who could address it, locking them into a course of action that the uncorrupted data would have shown to be failing.

The Lesson

The Soviet factory system and the Vietnam body count represent Goodhart's Law at its most consequential. In one case, the result was an economy that produced things nobody wanted. In the other, the result was a war strategy built on fiction, prolonging a conflict that killed millions.

But the lesson is not "those people were foolish." The lesson is structural. The planners in Moscow were not stupid. Robert McNamara was not stupid. Both were operating under the same constraint: they needed to coordinate action at a scale that exceeded their ability to observe directly, so they used metrics as proxies for observation. And in both cases, the metrics were corrupted by the very optimization pressure they were designed to harness.

The question is not how to avoid metrics. Modern organizations cannot function without them. The question is how to use metrics while remaining aware that every metric, under sufficient optimization pressure, will begin to diverge from the reality it represents. The answer lies in the solutions discussed in the main chapter: multiple independent metrics, qualitative assessment, rotation and unpredictability, monitoring for gaming, and governance structures that keep the measurer and the measured in close, ongoing relationship.

The nail factory and the body count are not historical curiosities. They are blueprints for a failure mode that is as alive today as it was in 1960s Moscow or 1960s Saigon. Every time an organization ties high-stakes incentives to a single metric and monitors from a distance, the same pathology awaits.

Questions for Reflection

1. Both the Soviet planning system and the Vietnam body count were implemented by people who understood the risk of metric gaming. Why were they unable to prevent it? What does this suggest about the limits of awareness as a defense against Goodhart's Law?

2. The text describes a "ratchet effect" in both systems. How does the ratchet effect compound the basic Goodhart's Law problem? Can you identify a ratchet effect in a system you are familiar with?

3. In both cases, gamed metrics fed back into decision-making, creating what the text calls "systematic self-deception." How does this connect to the feedback loop concepts from Chapter 2? Is there a point at which the feedback loop becomes self-correcting, or can it persist indefinitely?

4. The Soviet Union eventually collapsed; the United States eventually withdrew from Vietnam. In both cases, reality eventually overrode the metrics. What determines how long a system can sustain the gap between metric and reality before the gap becomes unsustainable?

5. Consider a modern organization you are familiar with. Identify the principal, the agent, the proxy metric, and the gap between the metric and the underlying reality. Using the structural analysis from this case study, predict what forms of gaming are most likely to emerge. Then investigate whether your prediction is correct.