4.1 Metrics that matter: yield, FPY, scrap, and OTD

A factory generates exactly two outputs: physical hardware and operational data. Ignoring the data means the physical product will eventually fail in the field. However, not all data is created equal. In electronics manufacturing, “Vanity Metrics” are sometimes misused by factory managers to make operations look good while deliberately hiding underlying operational issues.

Engineering leaders must consistently distinguish between metrics that measure activity (how busy people are) and metrics that measure health (how stable the process is). A world-class factory is not one that expertly diagnoses and fixes defects at high speed; it is one that systematically prevents those defects from being created in the first place.

First pass yield (FPY): the only truth

Definition: The exact percentage of units that pass the entire testing process on their very first attempt, without any human intervention, rework, or re-testing.

FPY = Units Passed First Time / Total Units Tested

The engineering reality

Most contract manufacturers will proudly report a “Final Yield” or “Ship Yield” of 99%+. This is often a manipulated metric. Final Yield includes units that failed test initially, were taken apart by a technician, had components manually replaced, and then finally managed to pass the test station on a subsequent try.

• The Trap: When FPY is a low 85% but Final Yield is reported as 100%, the factory is operating as a high-volume repair shop, not a controlled manufacturing line.
• The Warning: When FPY suddenly drops below 95% on a mature SMT line, it indicates a systematic process drift (e.g. clogged solder paste stencil, exhausted reflow oven filter) rather than just a “bad batch of parts,” requiring immediate engineering intervention.
• The Consequence: Hand-reworked units universally possess lower long-term reliability than units that flowed cleanly through the automated line. A chronically low FPY statistically predicts a future spike in field returns, even when every unit eventually “passed” the factory test.

Scrap rate: the true cost of waste

Definition: The financial percentage of materials that are permanently discarded because they cannot be economically or practically repaired.

Scrap Rate = Value of Scrapped Material / Total Production Value

The engineering reality

Scrap is never just the raw cost of the lost component. It is the cost of the component + the factory labor invested so far + the machine time wasted + the toxic disposal fees + the total Opportunity Cost (the profit that could have been made building a good unit in that exact same time slot).

• The Timing Rule: Defect cost grows exponentially as the board moves down the line. Scrapping a bare, unpopulated PCB at the start of the line incurs minimal loss. Scrapping a fully populated, conformal-coated PCBA at the very end of the line incurs the maximum possible financial loss.
• The Standard: Scrapping a locally damaged $50 microchip is unfortunate; scrapping a completely assembled $5,000 server blade because it cannot be reworked is a significant financial event.

OTD (On-Time Delivery): the supply chain promise

Definition: The exact percentage of shipments that leave the factory dock on or before the originally committed date.

The engineering reality

“On Time” is a binary state. A shipment is either on time, or it is late. “Almost on time” is recorded as late.

• The Measurement Trap: A common factory trick to artificially boost OTD is quietly moving the goalposts. When a factory misses a Friday shipment but silently updates the ERP “Target Date” to Monday, the system reports “100% OTD.” This masks the true delivery performance from the customer.
• The Root Cause: When true OTD consistently sits below 90%, it points to a systemic materials planning failure, a broken supply chain, or a chronically overbooked production schedule rather than a localized workforce issue.

The “Bone Pile” and WIP (Work In Progress)

Definition: WIP is the total inventory sitting exposed on the factory floor, partially finished. The informal term “Bone Pile” refers to the stack of failed boards sitting on a shelf waiting for a technician to diagnose and repair them.

The engineering reality

A visibly large Bone Pile is a physical symptom of a failing process. It represents trapped, burning cash.

• The Bottleneck: When the Bone Pile grows faster than repair technicians can diagnose and fix the boards, the main production line must be administratively halted. The line is actively producing defects faster than they can be recovered.
• The Metric: “Days of WIP” must always be measured. When a raw board typically takes 2 days to be boxed but currently takes 5 days to navigate the factory, the production flow is critically blocked.

The vanity metric trap: “testing into compliance”

The most dangerous localized failure on a manufacturing floor is the “Retest Loop.”

• The Scenario: A finished board fails the Functional Testing (FCT). The operator, needing to meet their hourly quota, simply hits the “Retest” button. It fails again. They wiggle the cables and hit “Retest” a third time. Miraculously, it passes. The operator applies a green sticker (pass) to it and ships it.
• The Truth: The board was not fixed. The board was merely subjected to enough vibration and thermal variation to find the exact millimeter where a fractured solder joint temporarily made contact. This is a highly defective board with an intermittent fault. It is statistically likely to fail in the customer’s hands.
• The Critical Control: Test software must be hardcoded to permanently lock out a Serial Number after two consecutive test failures. It must forcibly route the board to a Level-3 Technician for rigorous physical analysis. Infinite retests must never be authorized.

---

Recap: Manufacturing Quality Metrics

Metric	Target Value	Failure Threshold	Required Action
First Pass Yield (FPY)	≥ 95% (mature SMT line)	< 95%	Immediate engineering intervention for systematic process drift.
Scrap Rate	Minimize; cost grows exponentially with production stage.	Scrapping high-value assembled units (e.g., $5k server blade).	Analyze root cause; prioritize early-stage defect detection.
On-Time Delivery (OTD)	100% (binary, based on original committed date).	< 90% consistently.	Investigate systemic materials planning or supply chain failure.
Test Retry Policy	Zero infinite retests.	> 2 consecutive test failures for a unit.	Hard-lock serial number; route to Level-3 technician for physical analysis.
WIP / “Bone Pile”	Monitor “Days of WIP” vs. standard flow time (e.g., 2 days).	Pile grows faster than repair capacity; flow time critically extended (e.g., 5 days).	Halt main production line to address defect generation rate.