A PCB/PCBA yield analytics workflow connects serialized units, process steps, inspections, tests, defects and rework into a controlled dataset that teams can use to contain escapes and improve the process. A dashboard is useful only when definitions, denominators and defect dispositions are consistent.
Key takeaways
- Define unit, opportunity, pass, rework, retest, scrap and false call before calculating yield.
- First-pass yield (FPY) should exclude rework/retest recovery; final yield should be reported separately.
- Rolled throughput yield multiplies step yields and exposes cumulative loss hidden by final pass rate.
- Separate process defect, design issue, component issue, test-system issue and handling damage.
- Link every record to product revision, lot/panel, machine/program, material and timestamp.
- Use Pareto analysis for prioritization, then verify root cause with evidence and a control plan.
Core metrics and definitions
| Metric | Definition | Decision it supports |
|---|---|---|
| FPY | Units passing a step first time / units entering the step | Process performance without recovery |
| Final yield | Accepted units / started units after allowed rework/retest | Delivery output |
| Rolled throughput yield | Product of FPY across sequential steps | Total hidden factory impact |
| Defects per unit | Confirmed defects / inspected units | Defect density trend |
| DPMO | Defects / total defined opportunities × 1,000,000 | Opportunity-normalized comparison |
| False-call rate | Inspection/test calls not confirmed as defects / total calls | Inspection-program effectiveness |
| Escape rate | Defects found downstream / units or opportunities shipped from prior gate | Detection/control effectiveness |
Never compare DPMO between products unless “opportunity” is defined consistently.
Minimum traceability model
Capture:
- product, PCB and BOM revision;
- serial/lot/panel and position;
- material/component lot and approved substitution;
- machine, line, recipe/program and software version;
- operator where process requires it;
- timestamp and environmental/process parameters;
- inspection/test result, measured values and limit revision;
- defect code, location, evidence, disposition and rework history.
Practical analytics workflow
- Validate data completeness and remove duplicate/retest ambiguity.
- Calculate FPY by process, product, revision, line and time window.
- Build a Pareto of confirmed defects, not raw machine calls.
- Stratify the top defect by location, package, material lot, machine, program and panel position.
- Contain affected WIP and shipped risk using traceability.
- Confirm root cause through inspection, measurement or controlled experiment.
- Implement corrective action and update PFMEA/control plan/work instructions.
- Verify effectiveness over a defined volume/time and watch for side effects.
Defect taxonomy that supports action
| Level | Example |
|---|---|
| Process | Printing, placement, reflow, soldering, cleaning, test |
| Defect | Insufficient paste, offset, open, bridge, void, polarity, no fault found |
| Location | Refdes/pin/net, panel position and board region |
| Cause class | Design, material, machine, method, measurement, environment, handling |
| Disposition | Use-as-is, rework, repair, scrap, return or engineering review |
Free-text comments can supplement but should not replace controlled codes.
Common analytics failures
| Failure | Consequence | Better control |
|---|---|---|
| Retest counted as a new unit | Inflated denominator and yield | Serialized unit genealogy |
| AOI false calls treated as defects | Wrong Pareto priorities | Confirmed-defect field |
| Revision mixed in one trend | Design change looks like process drift | Revision filter/change marker |
| Only final test tracked | Upstream rework cost is hidden | Step-level FPY and RTY |
| Dashboard without containment | Insight arrives after more escapes | Alert and ownership workflow |
Yield-data RFQ and quality checklist
- Unit/lot/panel serialization and genealogy requirement
- Process gates and required measured-value retention
- Defect taxonomy and customer-specific codes
- FPY/final-yield/rework/scrap reporting definitions
- Traceability fields and record-retention period
- Alert, containment and notification thresholds
- First-article, control plan, PFMEA and change-notification scope
- Data export/API/portal format and access control
- Corrective-action format and effectiveness evidence
Review PCB testing and quality and submit program requirements through the PCB quotation page.
Standards and references
- IPC-CFX / IPC-2591 — Connected Factory Exchange
- IPC-HERMES-9852 — Machine-to-machine communication in electronics assembly
- IPC J-STD-001 — Requirements for soldered assemblies
- IPC-A-610 — Acceptability of electronic assemblies
- IPC-7912 — End-item DPMO and manufacturing quality data
Confirm current revisions and customer-specific metric definitions.
Yield analytics FAQ
What is the difference between first-pass and final yield?
First-pass yield counts units passing without rework or retest. Final yield includes units recovered through the allowed process.
Should AOI calls count as defects?
Only confirmed defects should drive defect yield. Track false calls separately to improve the inspection program.
What is rolled throughput yield?
It is the product of first-pass yields across sequential process steps and shows cumulative process loss.
Can two products' DPMO be compared directly?
Only if defect opportunities and counting rules are defined consistently.
What data is essential for root cause?
Product/revision, unit/panel position, process/machine/program, material lots, timestamp, measured result, defect location and disposition are core fields.