Taxonomy · Flow & Queuing · To the Bedrock

Four dimensions

Load pressure

How close arrival rate is to sustainable service capacity at the constraint. Formally, utilization ρ = λ / μ. Breathing is roughly ρ < 0.70; strained is 0.70–0.90; near saturation is above 0.90.

Breathing: average load leaves meaningful absorption room. Strained: capacity is still sufficient, but waiting sensitivity rises. Near saturation: small shocks produce large queue effects.

Variability profile

How uneven arrivals and service times are in practice. Formally captured by the coefficients of variation C_a (arrivals) and C_s (service times): values near zero are smooth, values near 1 are exponential, and values above 1 are bursty or fat-tailed.

Smooth: low burstiness, narrow work-size spread. Demand-bursty: arrivals clump in waves or interrupts. Work-size variable: item effort is fat-tailed or uncertain. Compound variability: both arrival and service variation are high.

Release discipline

How aggressively work is started relative to what can be completed.

Push-heavy: work starts early to keep everyone visibly busy. Pull-limited: WIP is capped; starts are constrained by finish capacity. Constraint-protective: release decisions subordinate to the limiting step.

Constraint behavior

What happens around the bottleneck in daily operations.

Starved: the constraint waits for inputs or decisions. Protected: priority and buffers keep constrained flow stable. Overfed: non-constraints inject work faster than conversion capacity. Moving: the bottleneck shifts between stages over time.

Flow regimes

These labels are interpretive lenses. They help teams reason about mechanism and intervention, not assign blame.

Regime	Typical Signature	Primary Risk	Intervention Bias
Breathing system low–moderate load · smooth variability	Moderate load, low-to-moderate variability, stable backlog.	Complacency about rising variation.	Protect feedback speed; keep WIP discipline before strain appears.
Strained but stable high load · moderate variability · pull-limited	High utilization with bounded queue, frequent local spill.	Tip into saturation after small shocks.	Reduce arrival bursts and service variability before adding load.
Burst trap moderate load · high variability	Average load looks safe, but queue appears intermittently.	Underestimating waiting because averages stay unchanged.	Smooth intake cadence; split large items; tighten handoffs.
Saturation trap near-capacity load · variability amplified	Utilization near 1, steep waiting curve, chronic backlog.	Runaway lead-time inflation and expedite churn.	Lower effective arrival rate, protect constraint, cap starts.
Activity-decoupled any load · push-heavy discipline · constraint moving	High local busyness metrics with weak end-to-end flow.	Optimizing activity while degrading throughput.	Shift KPIs to flow, waiting, and completion quality.
Overfed constraint variable load · constraint starved or overfed	Non-constraints run hot; bottleneck queue dominates cycle time.	WIP ballooning disguised as productivity.	Subordinate release to constraint rhythm; keep non-constraint slack.

Use it

Treat taxonomy as a diagnosis aid: identify your current regime, then test whether your intervention targets the mechanism that created waiting. Estimate load from cycle time or throughput data, observe variability from arrival intervals or effort distributions, and note whether intake is pull-limited or push-heavy. If you recognize your system in one of these regimes, open Explore and set load and variability to match—the readouts will show the quantitative difference between what the average suggests and what the system actually produces.

Explore Test each regime interactively with load and variability controls Concept Map Trace assumptions, observations, and surprises across theories