Per-Unit Invariance Check¶
Comparison method — instantiates Scale-Invariance Testing
Takes a per-unit rate as given and tests whether it stays flat as the number of units grows, exposing fixed costs, saturation, and coordination overhead.
Per-Unit Invariance Check takes an already-normalized per-unit figure — cost per delivery, latency per request, minutes per case — as given and asks one sharp question: does it stay flat as the number of units rises? Its defining idea is the flat-line test against count. A per-unit quantity that is genuinely scale-invariant traces a horizontal line as N grows; the moment it bends upward (or downward), something has switched — a fixed resource is saturating, a coordination cost is appearing, or hidden coupling among units has turned on. Where a metric check builds the denominator, this check consumes one and hunts for the count at which per-unit behavior breaks.
Example¶
A last-mile delivery depot believes its cost per delivery is constant, so growth is "just more volume." The check tests that belief directly by plotting cost per delivery against daily order count. At 200 deliveries a day the cost sits near $4.20; at 400 it dips slightly to $4.10 as route density improves — a mild economy of scale, the line still roughly flat. But at 900 a day it climbs to $5.60, and the plotted curve reveals the shape: flat, then rising, with the knee near 700. Attribution is immediate — around 700 a single dispatcher and the depot's finite dock doors saturate, and overflow vans idle waiting to load. The assumed per-unit invariance holds only below the knee; past it the depot suffers a diseconomy of scale.[n1] The action this implies is structural — add a second dispatch pod before 700 rather than pour in volume and expect the rate to hold.
How it works¶
- Hold the per-unit metric fixed and vary the count. The unit rate is the y-axis; the number of units is the x-axis. Nothing else moves if it can be helped.
- Look for the bend, not the average. A stable mean across the whole range can hide a flat-then-rising curve; the check is about the shape as count grows.
- Attribute the knee to a saturating factor. Name the fixed resource or coordination cost that switches on at the breakpoint — that is what makes the finding actionable.
Tuning parameters¶
- Count-axis resolution near the knee — dense sampling around the suspected breakpoint locates it precisely; coarse sampling can straddle and miss it entirely.
- Which fixed factors to instrument — the more candidate bottlenecks you watch (staff, equipment, locks, dependencies), the likelier you catch the true cause, at more measurement cost.
- Smoothing / noise handling — how much you filter before declaring a bend real, trading a hair-trigger against a late alarm.
When it helps, and when it misleads¶
Its strength is that the per-unit lens is exactly where fixed costs and coordination overhead first become visible — totals hide them, and the point where the flat line turns up is often the first honest warning that a system is leaving its invariant regime. Its failure mode is attributing a bend to scale when the real cause is a context change (a new depot layout, a different customer mix) that merely coincided with higher volume, or sampling count so coarsely that the knee is stepped over. The classic misuse is assuming flatness and extrapolating a per-unit rate straight past the count where it breaks. The guarding discipline is to sample densely near the knee and to separate genuine context shifts from the effect of sheer count before blaming scale.
How it implements the components¶
Per-Unit Invariance Check realizes the count-driven comparison slice of the archetype — the part that tests whether a per-unit rate survives more units:
scaling_ratio— its expected relation is a flat per-unit line (slope near zero); deviations from it are the finding.comparison_scale_set— the range of unit counts it sweeps is its comparison set.breakpoint_detection— its whole payoff is locating the count where the per-unit line bends as a fixed factor saturates.
It consumes the comparable rate rather than constructing it: the normalized_metric_or_ratio denominator and the acceptable_deviation_band drawn around it are Normalized Metric Check's work, and it reads a per-unit figure it does not itself build.
Related¶
- Instantiates: Scale-Invariance Testing — it is the archetype's saturation test, aimed at whether per-unit behavior holds as counts grow.
- Consumes: Normalized Metric Check supplies the per-unit rate it tests for flatness.
- Sibling mechanisms: Pilot-to-Scale Validation · Normalized Metric Check · Dimensional Scaling Test · Simulation Rescaling Sweep · Stratified Scale Sampling · Breakpoint Review Table · Log-Log Scaling Check
Editorial Notes¶
Form Classification¶
Form family: Experiment, Test & Rehearsal
Rationale: The mechanism deliberately varies unit count while holding the per-unit measure and other conditions fixed to learn where scaling bends or saturates.
Nearest alternative: Analysis, Modeling & Optimization — Curve analysis interprets the results, but controlled variation of scale is the defining probe.
Review outcome: Adjudicated after independent review; high confidence.
Origin Attribution¶
Primary origin: Mathematics
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Testing whether a normalized quantity remains invariant under scaling is a mathematical modeling practice.
Related originating lineages:
- Economics & Finance — Per-Unit Invariance Check is rooted in economics and finance: Economics of scale asks whether per-unit cost or performance stays invariant as system size changes.
- Statistics & Experimental Design — Experimental design and statistics materially shaped Per-Unit Invariance Check through randomization, inference, sensitivity analysis, and validation. Empirical scaling tests and residual checks supply the method for detecting deviations from flat per-unit behavior.
Review resolution: Light authoritative-source research resolves the primary-origin disagreement in favor of mathematics. Encyclopedia of Mathematics: Dimensional Analysis and Invariance directly documents the defining practice or theory described in the selected origin rationale. Other listed domains are retained only where the blind reviews identify material co-development or translation; broader adoption remains separate as domain_reach=multi_domain.
Attribution caveat: The boundary with economics and finance is real because that field materially developed or translated the practice, but the cited provenance places the defining form in mathematics.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
Notes¶
[n1] Diseconomies of scale — the region in which the per-unit cost of a system rises with size, typically because a fixed resource saturates or coordination overhead grows faster than output. It is the concrete phenomenon a flat-then-rising per-unit curve reveals. ↩