Skip to content

Correlated Capacity Demand

Prime #
754
Origin domain
Systems Thinking & Cybernetics
Subdomain
capacity planning and resilience → Systems Thinking & Cybernetics

Core Idea

When a finite shared resource serves multiple consumer processes, and the demand peaks across those consumers are tail-correlated rather than independent, the realized joint peak exceeds the capacity that was sized for independent peaks — by an amount that scales with the strength of the correlation in the tail. Capacity planning that assumed independent demands systematically under-provisions, and resilience strategies that relied on diversification — sharing the resource across consumers to smooth the load — collapse when the consumers' peak demands co-occur.

The structural commitments are five. A finite shared resource with a definite capacity — bandwidth, beds, megawatts, dollars of liquidity, responders. Multiple consumer processes drawing on the same resource, each with its own time-varying demand. Demand-correlation in the tail: the peaks of the consumer demands are statistically dependent under stress conditions, even if they appear uncorrelated under normal conditions. Capacity sized for non-correlated peaks: the planning assumption, often implicit, is independence, so the design load is a diversification-discounted sum rather than the joint exceedance. And joint-peak realization: when the correlated stress condition occurs, the resource is overwhelmed and the shortfall is allocated by some rationing rule, with downstream cascades.

The structural force is the tail-asymmetry between the planning distribution and the realized distribution: the planning model treats demands as independent and produces a capacity envelope much smaller than the joint-exceedance event realizes. The pattern is invisible during normal operation — demands genuinely do look uncorrelated — and reveals itself only during the rare conditions that drive the correlation. What looks like bad luck is in fact a structural under-provisioning whose realization is rare but whose occurrence is predictable from the correlation structure.

How would you explain it like I'm…

Everyone Showers at Once

Imagine one water pipe shared by every house on the street. Most of the time people use water at different moments, so the pipe is fine. But on a hot morning everyone turns on their shower at the exact same time, and suddenly the water slows to a trickle. The pipe was big enough only because we forgot everyone might want it at once.

When All the Peaks Line Up

When many users share one limited resource, like power, beds, or bandwidth, planners size it for how much everyone needs together. The trap is that they often assume the busy moments happen at random separate times, so the peaks cancel out. But sometimes the peaks line up: a heatwave makes everyone crank the air conditioning at once, and the shared supply was never built for that combined spike. Worse, sharing the resource was supposed to be the safety trick, smoothing demand across users, and that trick fails exactly when the peaks co-occur. It looks like bad luck, but it was actually predictable from the fact that the demands move together under stress.

Tail-Correlated Peaks

Correlated capacity demand is what happens when a finite shared resource serves several consumers whose demand peaks are tail-correlated rather than independent, so the real joint peak exceeds the capacity that was sized for independent peaks. The shortfall scales with how strongly the peaks are correlated in the tail. The deep problem is a mismatch between two distributions: the planning model treats demands as independent and produces a small capacity envelope, while the realized world makes the peaks co-occur and blows through it. Diversification, the usual resilience move of sharing one resource to smooth load, collapses precisely under the stress conditions that make the demands move together. The pattern is invisible in normal operation because demands genuinely do look uncorrelated, and reveals itself only in the rare stress event, so what looks like bad luck is structural under-provisioning whose timing is rare but whose occurrence is predictable from the correlation.

 

Correlated capacity demand: when a finite shared resource serves multiple consumer processes whose demand peaks are tail-correlated rather than independent, the realized joint peak exceeds the capacity sized for independent peaks, by an amount that scales with the strength of tail correlation. The structure has five commitments. A finite shared resource with definite capacity (bandwidth, beds, megawatts, liquidity, responders). Multiple consumer processes drawing on it, each with time-varying demand. Demand-correlation in the tail: the peaks are statistically dependent under stress even when they look independent in normal conditions. Capacity sized for non-correlated peaks: the planning assumption, often implicit, is independence, so the design load is a diversification-discounted sum rather than the joint exceedance. And joint-peak realization: when the correlated stress hits, the resource is overwhelmed and the shortfall is rationed with downstream cascades. The driving force is the tail-asymmetry between the planning distribution and the realized distribution; the model's independence assumption yields a capacity envelope far smaller than the joint-exceedance event realizes. The pattern is invisible during normal operation and surfaces only in the rare conditions that drive the correlation, so apparent bad luck is in fact structural under-provisioning that is predictable from the correlation structure.

Structural Signature

a finite shared resource with definite capacity — multiple consumer processes drawing on it — demand peaks correlated in the tail under stress — a correlating stressor as common cause — capacity sized under an independence assumption — a joint-exceedance event — the tail-asymmetry between planning and realized distributions as the load-bearing invariant

The pattern is present when each of the following holds:

  • A finite shared resource. A pooled capacity with a definite ceiling — bandwidth, beds, megawatts, liquidity, responders — serving several consumers.
  • Multiple consumer processes. Distinct demands draw on the same resource, each with its own time-varying load.
  • Tail-correlated demand. The consumer peaks are statistically dependent under stress conditions, even when they appear uncorrelated under normal operation — tail correlation, not centre correlation, is what matters.
  • A correlating stressor. A common-cause condition — climate, market regime, social event, infrastructure failure — couples the peaks, distinguishing this from independent demand.
  • Independence-assumption sizing. Capacity is planned, often implicitly, against a diversification-discounted sum of marginal peaks rather than the joint exceedance.
  • A joint-exceedance event. When the stressor occurs, the correlated peaks co-occur, demand overwhelms the resource, and the shortfall is allocated by some rationing rule with downstream cascades.
  • Planning-vs-realized tail asymmetry. The planning model produces an envelope much smaller than the realized joint-exceedance event — the invisible-until-rare structural under-provisioning.

These compose into a foreseeable rather than unlucky failure: name the stressor that correlates the consumers, size against the joint exceedance, and hold reserves not exposed to that same stressor — nominal redundancy sharing the common mode being worthless.

What It Is Not

  • Not adaptive capacity. adaptive_capacity (the embedding nearest neighbor) is the ability to reconfigure in response to change. This prime is a statistical fact about provisioning: when demands are tail-correlated, capacity sized for independent peaks fails at the joint exceedance. It concerns the sizing distribution, not the system's flexibility to adapt.
  • Not risk pooling. risk_pooling is the variance-reduction benefit of aggregating independent risks. This prime is precisely the case where pooling's promise fails — under tail correlation the diversification discount reverses, and pooling can make the joint peak worse. It names the boundary condition that voids risk pooling.
  • Not a margin of safety. margin_of_safety is headroom held against worst-case load. This prime explains why a margin sized under an independence assumption is the wrong size: the correlating stressor makes the realized joint peak far exceed the marginal peaks the margin was built on. Margin policy depends on getting this distribution right.
  • Not load balancing. load_balancing distributes work across resources to smooth utilization. This prime concerns a shared finite resource whose consumers' peaks co-occur under a common stressor; balancing across resources that share the stressor provides no relief, because the correlation is in the demand, not the distribution.
  • Not scalability. scalability is how cost or performance changes with size. This prime is orthogonal to scale: a perfectly scalable system still fails at the joint exceedance if its capacity was sized for independent rather than correlated peaks.
  • Common misclassification. Reading a joint-exceedance collapse as "we got unlucky." The failure is rare but structurally predictable from the correlation. The tell: ask what stressor correlates the consumers' peaks, and whether the reserves share exposure to that same stressor. Nominal redundancy that shares the common mode is worthless; the failure was sized in, not unlucky.

Broad Use

In healthcare, ICU capacity sized for independent admission peaks is overwhelmed when a pandemic wave, a heatwave, and a high-trauma weekend correlate, because the stress conditions couple admission types that are normally independent. In electricity grids, heatwaves drive air- conditioning demand up while simultaneously reducing solar, wind, and hydro generation, so demand-up and supply-down correlate as a structural feature of the climate substrate. In banking, correlated default scenarios have many borrowers failing to roll over funding at the same window, overwhelming capacity sized on diversified default risk. In reinsurance, hurricane seasons with multiple correlated events overwhelm capacity sized for independent storm losses. In incident response, on-call rotations sized for independent incident probability saturate when a deploy, a holiday, and a dependency outage correlate. In mutual-aid agreements among utilities, a wide-region storm has every utility requesting help at once, destroying the assumption that storms hit one utility at a time. In supply chains, redundant suppliers sharing an upstream input make diversification illusory, because the correlation is at the upstream bottleneck. In cloud infrastructure, redundant replicas across zones all depending on the same upstream storage layer fail together during a regional outage. Across substrates the structural pattern is constant: shared finite capacity plus correlated tail demand plus planning that assumed independence equals systematic under-provisioning that surfaces only under the rare joint-exceedance event.

Clarity

The prime sharpens a distinction often elided as "we got unlucky": capacity that was adequate by the planning model was structurally inadequate by the realized model, and the gap is the correlation in the tail. The pattern looks like bad luck because the failure is rare; it is in fact a structural under-provisioning whose realization is rare but whose occurrence is structurally predictable from the correlation structure. Naming it converts "an unlucky confluence" into "a known correlation we failed to size for."

It also distinguishes correlated-demand failure from neighboring shapes. A plain bottleneck is one process loading a single resource at steady-state, handled by classical queueing. A thundering herd is many consumers released by a shared triggering signal — correlated in time because of a common release event, a subspecies of correlated demand. A tail-risk property is a feature of the loss distribution; the prime is the operational consequence of tail risk in the demand structure on a finite shared resource. The diagnostic move is uniform: for any shared finite resource, ask what stress conditions correlate the consumer demands, and size the resource against the joint exceedance under those conditions, not the marginal exceedance of any one consumer. The clarifying force is to make the correlating stressor the object of analysis rather than the marginal demand of each consumer.

Manages Complexity

A wide class of "the resource was overwhelmed and we don't know why" failures collapses into a four-part accounting: enumerate the consumer processes drawing on the shared resource; identify the stress conditions that correlate the consumer demands — climate, market regime, social event, infrastructure failure; estimate the joint-exceedance demand under those conditions rather than the marginal expectations; and choose among three interventions — increase capacity to cover the joint peak, decouple the consumers with truly independent reserves, or ration by policy in advance. The same accounting transfers from ICU surge planning to electricity grids to banking liquidity to supply chains to incident-response staffing.

The compression is sharpened by the recognition that the only structural defense is reserves not exposed to the correlating stressor. "Distributed across geographies" reserves that share a climate or regulatory exposure are not independent, which is why offshore reinsurance, cross-asset-class capital, and cross-substrate redundancy are valued in resilience planning. Recognizing a capacity failure as a correlated-demand failure thus directs the analyst to the right intervention: not more nominal redundancy, which may share the common mode, but genuinely independent reserves plus rationing-by-policy for the joint event. This is far more compact and accurate than reasoning about each resource collapse separately, because it names the binding quantity — the tail correlation and its driving stressor.

Abstract Reasoning

The prime supports several inferences. The diversification illusion: pooling consumers reduces variance under independence, but under correlation the variance reduction collapses and the pooled load can be worse than the sum of marginal loads under certain rationing rules. Tail-correlation versus centre-correlation: two demand processes can be near-uncorrelated under normal conditions and strongly correlated in the tail, and the relevant correlation for capacity planning is the tail correlation, often much higher than linear-correlation estimates suggest — which is why copula models exist to capture it. Common-cause versus common-effect correlation: tail correlation may arise from a common-cause stressor driving multiple consumers or from common-effect coupling where consumers respond to each other, the former being the prime's typical case and the latter overlapping with cascade.

Two further inferences concern defense and design. Genuinely independent reserves: the only structural defense is reserves not exposed to the correlated stressor, so reserves sharing a climate or regulatory exposure are not independent regardless of their count. Joint-peak design: capacity sized for the joint peak is much larger than capacity for marginal peaks and often prohibitively expensive, so operational practice combines some joint-peak provisioning, policy-driven rationing during joint events, and emergency-decoupling protocols. These inferences follow from the capacity-statistics structure alone, so they apply to a hospital, a grid, and a cloud region alike, and they tell a planner that nominal redundancy is worthless against a stressor it shares, and that the design target must be the joint exceedance rather than the marginal one.

Knowledge Transfer

The transferable content is the four-part accounting — enumerate consumers, identify the correlating stressor, estimate the joint exceedance, choose among capacity, decoupling, and rationing — together with the diversification-illusion and genuinely-independent-reserves inferences. Because the pattern is pure capacity statistics with substrate-neutral vocabulary, the moves carry across domains. Copula-based correlation modeling ports from credit-risk analysis to surge planning, the structural insight that independence underestimates joint-peak risk transferring intact. Grid-level joint-peak planning — correlated demand plus correlated supply failure — ports to hospital surge planning — correlated admissions plus correlated staff illness — with the same four-part accounting. Catastrophe-bond logic about correlated peril exposure ports to cloud-tenant capacity planning. The portable diagnostic — what stressor correlates the consumers? — names the failure mode once the stressor is named, and counsels conservative design when it cannot be.

These transfers work because the structural roles are stable: a shared finite resource, multiple consumer processes, a planning independence assumption, a tail correlation, a correlating stressor, a joint exceedance, and a rationing rule. A surge planner, a grid operator, a bank treasurer, and a cloud architect are all running the same move: ask what stressor correlates the demands, size against the joint exceedance, and build reserves not exposed to that stressor. The portable lesson is that diversification is an illusion when the diversified parts share a common stressor, so the right question is never only "is each consumer's peak covered?" but "what correlates their peaks, and is our reserve exposed to the same thing?" — a lesson that travels intact from an ICU to a power grid to a cloud region, and that, once held, turns a rare and bewildering collapse into a foreseeable consequence of a correlation that the planning model assumed away.

Examples

Formal/abstract

The structure shows cleanly in the statistics of pooling. Take a finite shared resource of capacity C serving multiple consumers, each drawing a demand D_i with the same marginal distribution. Under the independence-assumption sizing, a planner reasons by diversification: the sum of n independent demands has variance that grows like n while the mean grows like n, so the coefficient of variation shrinks like 1/√n, and the diversification-discounted peak the planner sizes for is far below the naive sum of individual peaks. This is the plain bottleneck intuition, and it is correct only under independence. Now introduce a correlating stressor — a common-cause variable S that, in its tail, pushes every D_i up together. The relevant correlation is tail correlation, not centre correlation: the demands can be near-uncorrelated in normal conditions (so the planner's data look reassuring) yet strongly dependent in the tail, which is exactly why copula models exist to capture dependence structure that a single linear-correlation number hides. The joint-exceedance event is the realized distribution's tail: when S occurs, the variance-reduction-from-pooling collapses, the demands co-peak, and realized demand far exceeds C — the planning-vs- realized tail asymmetry that is the load-bearing invariant. The diversification illusion is provable here: pooling that reduces variance under independence can, under strong tail dependence and certain rationing rules, leave the pooled load worse than the sum of marginal loads. The defense the math forces is genuinely independent reserves — reserves whose own availability is not a function of S — because any reserve exposed to the same stressor adds count without adding true diversification.

Mapped back: the pooling model instantiates every role — shared capacity, multiple consumers, an independence sizing assumption, a common-cause stressor inducing tail correlation, and a joint-exceedance event — making the diversification illusion and the demand for stressor-independent reserves theorems rather than slogans.

Applied/industry

An electricity grid during a heatwave is the prime's archetype, and it adds a second correlation the naive planner misses. The finite shared resource is generation-plus-transmission capacity; the consumers are millions of loads, dominated in summer by air conditioning. A planner sizing for independent peaks assumes that not everyone's demand peaks at once. But a heatwave is a correlating stressor that couples the demands: extreme heat drives air-conditioning load up across the entire region simultaneously — demand correlates in the tail even though, on a mild day, individual households' usage looks independent. Worse, the same stressor correlates the supply side: heat reduces thermal-plant efficiency and transmission ampacity, drought cuts hydro, and still air cuts wind, so demand-up and supply-down correlate as a structural feature of the climate substrate — a joint exceedance on both sides at once. The planning-vs-realized asymmetry is the rolling blackout: a grid "adequate" by independent-peak planning is structurally inadequate by the realized joint-exceedance model, and the gap is the tail correlation, not bad luck. The prime's intervention catalogue applies directly: size against the joint exceedance (the coincident-peak heatwave scenario, not the marginal household peak); hold genuinely independent reserves — interconnection to a grid in a different weather system, since reserves sharing the same climate exposure are not independent regardless of their nameplate count; and ration by policy in advance (demand-response programs, planned curtailment) for the joint event. The identical four-part accounting governs hospital surge (correlated admissions plus correlated staff illness in a pandemic) and cloud capacity (redundant zones all depending on one upstream storage layer), where "distributed across regions" reserves that share a common dependency are the same diversification illusion.

Mapped back: the heatwave grid is correlated capacity demand — shared generation capacity, millions of consumers, an independence sizing assumption, a heat stressor correlating demand-up with supply-down, and a blackout as joint exceedance — so the fix (size for the coincident peak, hold cross-climate reserves) is the same structural move as hospital surge and cloud-region planning.

Structural Tensions

T1 — Tail Correlation versus Centre Correlation (measurement). The prime's binding quantity is correlation in the tail under stress, but the data a planner usually has is centre correlation under normal operation — and the two can differ wildly, with demands near-independent on ordinary days yet tightly coupled in the extreme. The failure mode is sizing on a reassuring linear-correlation estimate that vastly understates the tail dependence, so the planning model looks sound on all the available data and fails exactly where no data existed. Diagnostic: ask whether the correlation was measured in the tail (rare joint-stress events) or extrapolated from the bulk. The prime warns that copula structure, not a single correlation number, governs the joint exceedance; trusting centre correlation is trusting the calm to predict the storm.

T2 — Independence Assumption versus Realized Coupling (sign/direction). Capacity sized under independence produces a diversification discount; under correlation that discount reverses and the pooled load can exceed the sum of marginal loads. The failure mode is the diversification illusion: pooling consumers "to smooth the load" actively worsens the joint peak when a common stressor couples them, so the resilience strategy becomes the vulnerability. Diagnostic: ask whether the consumers share a common-cause stressor before crediting any variance reduction from pooling. The prime's hardest inversion is that diversification's benefit has the wrong sign under tail dependence; a plan that counts on pooling for safety is, against a shared stressor, counting on the exact mechanism that concentrates the failure.

T3 — Nominal Redundancy versus Stressor-Independent Reserves (scopal). The only structural defense is reserves not exposed to the correlating stressor — but redundancy is usually counted by nameplate, and "distributed across regions" reserves that share a climate, a regulator, or an upstream dependency are not independent regardless of their count. The failure mode is buying redundancy that shares the common mode: backup generators that fail in the same flood, replica zones depending on the same storage layer, reinsurance exposed to the same peril. Diagnostic: for each reserve, ask whether its own availability is a function of the stressor that drives the demand peak. The prime's leverage is that redundancy count is worthless against a shared stressor; only genuine stressor-independence adds resilience, and that must be verified, not assumed from quantity.

T4 — Joint-Peak Provisioning versus Cost (boundary). Sizing capacity for the joint exceedance is the structurally correct target, but joint-peak capacity is much larger than marginal-peak capacity and often prohibitively expensive — so full joint-peak provisioning is rarely affordable. The failure mode is two opposite errors: under-provisioning to the marginal peak (the prime's named failure) or over-building to a joint peak so rare the capacity sits idle at ruinous cost. Diagnostic: ask what blend of partial joint-peak capacity, policy-driven rationing during joint events, and emergency decoupling protocols covers the tail at acceptable cost. The prime hands off here to risk-budgeting; it identifies the joint exceedance as the real load but does not claim the entire joint peak must be built — the design is a portfolio, not a single capacity number.

T5 — Common-Cause versus Common-Effect Correlation (coupling). Tail correlation can arise from a common-cause stressor driving multiple consumers independently, or from common-effect coupling where consumers respond to each other — and the prime's typical case is the former, while the latter overlaps with cascade. The failure mode is misattributing the coupling: defending against an exogenous stressor (size for the heatwave) when the real mechanism is consumers amplifying one another (a bank run, a retry storm), which feeds back and demands a cascade-breaking remedy instead. Diagnostic: ask whether the demands rise together because of a shared external cause or because each consumer's surge triggers the next. The prime addresses common-cause correlation; reactive common-effect coupling is a different prime with a different intervention, and conflating them defends the wrong mechanism.

T6 — Static Correlation versus Regime-Dependent Coupling (temporal). The prime treats the correlating stressor as identifiable and the correlation as a structural feature, but the coupling itself can change with regime — demands uncorrelated under one market or climate state become correlated under another, and new common dependencies form as systems integrate. The failure mode is a one-time correlation audit that ages: a reserve verified stressor-independent at design time shares a newly-introduced common dependency (a consolidated supplier, a shared cloud provider) the original analysis never saw. Diagnostic: re-audit the correlation structure as the system and its environment evolve, asking what new common modes have formed since the last assessment. The prime's stressor is not a constant; tail dependence is itself non-stationary, and yesterday's independent reserves can quietly become today's correlated ones.

Structural–Framed Character

Correlated capacity demand sits at the structural pole of the structural–framed spectrum — aggregate 0.0, every diagnostic structural. It is a pure capacity-statistics pattern: when demands on a shared finite resource are tail-correlated rather than independent, capacity sized for independent peaks fails at the rare joint exceedance. Nothing about it depends on a particular substrate's vocabulary or values.

Every diagnostic points one way. The pattern carries no home vocabulary that must travel — "demand," "capacity," "correlation," "tail" are substrate-neutral statistical terms, and the same structure describes a hospital ICU surge, a power-grid stress event, a bank liquidity run, a reinsurance catastrophe, and a supply-chain shock, each in its own field's words. It carries no evaluative weight: tail-correlation is neither good nor bad in itself; it is a statistical fact about joint exceedance that becomes a hazard only relative to a capacity decision. Its origin is formal — the statistics of correlated extremes on a finite resource, statable with no institutional content. It is not human-practice-bound: the pattern governs any shared finite resource under correlated load, including physical and ecological ones, with no human role required for the joint-exceedance failure to occur. And to invoke it is to recognize a tail-dependence already present in the demand structure — a correlation the data exhibit, not an interpretation imposed — and specifically to recognize where the independence assumption fails. On every diagnostic it reads structural, matching the all-zero aggregate.

Substrate Independence

Correlated capacity-demand is a maximally substrate-independent prime — composite 5 / 5 on the substrate-independence scale. The signature — demand (or supply shortfall) on a shared finite resource that is correlated in its tail, defeating capacity sized on an independence assumption — is recognized, not translated, across substrates that share no other vocabulary: healthcare (ICU capacity sized for independent admission peaks, overwhelmed when a pandemic wave, heatwave, and trauma weekend correlate), electricity grids (heatwaves driving demand up while cutting solar, wind, and hydro, so demand-up and supply-down correlate structurally), banking (correlated funding-rollover failure overwhelming capacity sized on diversified default risk), reinsurance (multiple correlated catastrophe events), incident response (on-call rotations saturating when a deploy, holiday, and dependency outage coincide), mutual-aid utility agreements (a wide-region storm making every utility request help at once), supply chains (redundant suppliers sharing an upstream input, making diversification illusory), and cloud infrastructure (replicas across zones all depending on one storage layer). That breadth earns the full domain score. Structural abstraction is maximal because the load-bearing element — tail-correlation of draws on a shared finite resource — is a pure statistical-structural property carrying no domain-specific commitments. Transfer evidence is the strongest kind: the same capacity-planning failure (sizing on independence, breaking on correlation) and the same correlation-at-the-shared-bottleneck diagnosis are documented identically across these substrates, so the transfer is one shared structural model rather than analogy.

  • Composite substrate independence — 5 / 5
  • Domain breadth — 5 / 5
  • Structural abstraction — 5 / 5
  • Transfer evidence — 5 / 5

Relationships to Other Primes

One-hop neighborhood: parents above, mutual partners to the right, children below.CorrelatedCapacity Demandsubsumption: Thundering HerdThundering Herd

Foundational — no parent edges in the catalog.

Children (1) — more specific cases that build on this

  • Thundering Herd is a kind of Correlated Capacity Demand

    The file explicitly names thundering_herd "a subspecies of correlated capacity demand" twice (Clarity + Not-to-be-Confused-With): both are shared-finite-resource + correlated-tail-demand, differing only in what makes the correlation (a shared release event in thundering_herd vs a common-cause stressor in the general prime). Direction verified: the general prime subsumes the timing-artifact special case. thundering_herd is a valid candidate slug. (Distinct from adaptive_capacity, risk_pooling, margin_of_safety per Phase-C — those stay severed.)

Neighborhood in Abstraction Space

Correlated Capacity Demand sits among the more crowded primes in the catalog (27th percentile for distinctiveness): several abstractions describe nearly the same structure, so a description that fits it will tend to fit its neighbors too — transporting it usually means disambiguating within this family rather than landing on it exactly.

Family — Finite Capacity & Contention (18 primes)

Nearest neighbors

Computed from structural-signature embeddings · 2026-06-14

Not to Be Confused With

The most consequential confusion is with risk_pooling, because this prime is in a real sense the negation of risk pooling's central promise, and conflating them inverts the correct advice. risk_pooling is the principle that aggregating many independent risks reduces relative variance — the coefficient of variation of a sum of n independent demands shrinks like 1/√n, so a pooled resource can be sized below the sum of individual worst cases and still be safe. Its whole leverage rests on independence. Correlated capacity demand is precisely the structure where that independence assumption fails in the tail: a common-cause stressor couples the consumers' peaks, the variance-reduction-from-pooling collapses, and the pooled load can be worse than the sum of marginal loads under certain rationing rules. The two are not competing heuristics but a principle and its boundary condition. risk_pooling tells you "pool to smooth the load"; this prime tells you "pooling is an illusion when the pooled parts share a stressor, and the smoothing you counted on becomes the concentration that sinks you." The practical danger is that a planner who has internalized risk pooling will actively pursue aggregation as a resilience strategy — combining ICUs, interconnecting grids, pooling liquidity — and thereby build the exact correlated-exposure structure this prime warns against, mistaking a vulnerability for a safeguard. The discriminating question is always whether the pooled risks share a common-cause stressor in the tail; where they do, risk pooling's sign flips and this prime governs.

A second genuine confusion is with thundering_herd, which is in fact a subspecies of correlated capacity demand and so must be distinguished by what makes the correlation. In the thundering herd, the consumers' demands correlate in time because a shared release event — a cache expiry, a market open, a breaker close — synchronizes them: they were held back together and discharge together onto a finite resource. Correlated capacity demand is the broader class in which the correlation arises from a common-cause stressor (a heatwave, a market regime, a pandemic) that pushes independent consumers' peaks up together, without any synchronizing release. The difference dictates the remedy. The thundering herd is cured by decorrelating arrivals in time — jitter, randomized backoff, staggered admission — because the correlation is a synchronization artifact of a shared trigger. Correlated capacity demand is cured by sizing against the joint exceedance and holding stressor-independent reserves, because the correlation is a standing feature of the demand structure under stress, not a timing artifact that can be jittered away. Jittering a heatwave does nothing — every air conditioner is hot at once regardless of when each turns on; conversely, sizing a cache origin for the "joint exceedance" misframes a problem that staggered expiry would dissolve. Treating the general correlated-demand failure as a thundering herd sends the engineer hunting for a shared release event and a decorrelation knob that do not exist; treating a thundering herd as general correlated demand over-builds capacity for a burst that should simply have been spread in time.

For a practitioner the distinctions order the analysis of any shared-resource collapse. First ask what kind of correlation overwhelmed the resource: a shared release event synchronizing arrivals (thundering_herd — decorrelate in time) or a common-cause stressor coupling peaks (this prime — size for the joint exceedance, hold independent reserves). Then beware the risk_pooling instinct: under a shared stressor, the aggregation that promised variance reduction is the mechanism of concentration. The prime's unique contribution is the insistence that diversification is an illusion when the diversified parts share a stressor, so the binding question is never only "is each peak covered?" but "what correlates the peaks, and is our reserve exposed to the same thing?"

Solution Archetypes

No catalogued solution archetypes reference this prime yet.