Skip to content

First-Pass Loss Audit

Audit — instantiates Effective-Input Delivery Assurance

Isolates and books the loss at the first boundary the input must cross — the earliest, largest, most easily hidden drop, before any downstream stage can even see it.

A retrospective audit that zeroes in on a single stage — the first boundary the supplied input must cross — and quantifies how much is lost right there, before the rest of the path even begins. Its defining move is refusing to spread attention across the whole funnel: it goes after the first pass specifically, because that is usually where the largest and most invisible fraction disappears (the drug metabolized before it reaches the bloodstream, the applicants filtered at the first screen, the units scrapped on the first operation), and because a loss booked before the first meter never shows up in any downstream number. It compares the supplied baseline to what actually survives the first crossing, books the difference as first-pass loss, and names the mechanism responsible.

Example

An electronics plant runs a surface-mount assembly line and reports "98% yield." A First-Pass Loss Audit distrusts that number, because it is measured at final test — after rework loops have quietly re-fed failed boards back through. The audit fixes the first boundary at the first reflow-and-inspection gate and asks a stricter question: of the 10,000 populated boards that entered, how many passed the first time, untouched? The answer is about 8,900 — a first-pass loss of ≈11%, almost all of it solder-bridging on one fine-pitch component, where the stencil aperture simply cannot deposit paste reliably at that pitch (an absorption-capacity limit at the entry boundary). The audit books ≈1,100 boards of first-pass loss to that cause.

That loss was invisible because rework eventually recovered most of the boards — the "hidden factory" of rework concealed an 11% first-pass loss behind a 98% final yield.[1] The audit's output is not "the line is fine"; it is "your biggest, most curable leak is one component's paste deposition, and it is costing you an eighth of your first pass."

How it works

  • Fix the first boundary and the supplied baseline. Everything downstream — and any rework or recovery that launders the number — is deliberately out of scope.
  • Count first-time survivors only. No second chances: the audit counts what crossed the first boundary untouched, not what eventually made it.
  • Book supplied − first-pass survivors as loss, attributed. The gap is charged to a mechanism: an absorption ceiling, degradation, or rejection at entry.
  • Report it against the nominal figure. The value is the contrast — the first-pass loss the rework-inclusive headline hides.

Tuning parameters

  • Boundary placement — exactly where "first pass" is drawn; move it and you capture a different loss. The most consequential dial, and the one most worth fixing before you look at the data.
  • Rework / recovery treatment — count only untouched first-time survivors, or allow one recovery pass; stricter exposes more hidden loss.
  • Attribution depth — book to a stage, or drill to root cause (which component, which limit).
  • Audit scope — full census versus sampled lots; census for rare high-cost loss, sampling for routine flow.

When it helps, and when it misleads

Its strength is surfacing the single biggest, most hidden leak — the one that both a whole-system average and a rework-laundered final number conceal — and pointing correction at the highest-yield fix.

Its failure mode is tunnel vision: obsessing over the first pass can miss compounding small losses spread across later stages, so a healthy first-pass number is not a healthy funnel. Its classic misuse is to run backwards to exonerate — draw the boundary after the messy first step so the first-pass figure looks clean, or count rework as first-pass. The discipline that keeps it honest is to fix the boundary and the "no recovery" rule before seeing the data, and to pair the audit with a full-funnel view so a good first pass is never mistaken for good end-to-end availability.

How it implements the components

First-Pass Loss Audit fills the loss-accounting side of the archetype, scoped to the first crossing:

  • loss_or_leakage_accounting — its core output: the loss at the first boundary, quantified and attributed to a cause.
  • supplied_input_baseline — the audit anchors on exactly what was supplied; first-pass loss is that baseline minus first-time survivors.
  • downstream_absorption_capacity — first-pass loss is usually an absorption ceiling at the entry boundary, and the audit names that limit as the mechanism.

It does not chart the whole stage-by-stage funnel (that is Stagewise Availability Assay and Availability Funnel Dashboard), model transport physics (Advection-Diffusion or Transport Modeling), or fix the loss it books (Route–Form–Timing Optimization); this mechanism accounts for the first-pass loss, it does not measure every stage or correct it.

Notes

First-pass loss is not total loss, and this audit is deliberately narrower than a whole-system Mass Balance (which reconciles all inputs and outputs) or a Sankey Loss Map (which draws every split). Its whole value is depth over breadth on the one stage that usually dominates and hides best. Run it when a headline number looks suspiciously good; run a funnel view alongside so the later stages are not forgotten.

References

[1] The "hidden factory" — the plant capacity consumed by rework and scrap that never appears in headline yield, a term associated with quality theorist Armand Feigenbaum. First-pass yield counts only units that pass without rework, which is exactly why it exposes losses a rework-inclusive final yield conceals. ↩