Yields are on target, yet field failures and RMAs are rising
Question
Our fab yield reports are green and final test pass rates are stable. But early-life field failures are up, RMAs bunch after big ramps, and a few lots show corner-case issues at temperature/voltage extremes. Suppliers say their outgoing quality is fine. How can factory numbers look good while the field feels fragile?
Answer ✅
Factory yields can mask margin erosion. If corner parametrics drift, guardbands compress, burn-in escapes rise, or assembly/packaging variation creeps in, you can “hit yield” while reliability stability slips. SSM-Audit adds a stability band beside the KPIs you already produce so you see whether quality is calm and repeatable across lots, corners, and time—or just “passing” with thinning safety.
What the bands would have shown 📊
• Lot-to-lot yield stability sliding A+ -> A0 (same average, higher variance)
• Parametric margin stability degrading to A- (corners at temp/voltage thinning)
• Burn-in escape stability weakening A0 -> A- (more early-life failures slip past)
• Test-escape rate tilting A0 -> A- (fails appear in system test despite final pass)
• Package/assembly reliability stability dipping A0 -> A- (reflow/voiding clusters)
• Field RMA cadence softening A+ -> A0 (DPPM increases, returns bunch after ramps)
What to do now 🛠️
- Band end-to-end: fab → probe → final test → burn-in → assembly → system/field. Publish a weekly band panel.
- Guard corners: when parametric margin band < A0, widen guardbands or hold ship; add corner-focused screens.
- Target high-variance lots: if lot yield stability < A0, elevate sampling, extend burn-in, and segregate shipments.
- Tighten escapes: raise test depth selectively where test-escape band slips; add canary tests on weak paths.
- Close the package loop: tie package reliability band to supplier/process windows; trigger requalification on dips.
- Field-first feedback: when RMA cadence band degrades, launch rapid 8D; ship-gate expansions until band recovers.
How SSM-Audit helps (practicalities) 🌟
• No additional infrastructure: sits beside MES, probe/final test logs, burn-in, and RMA systems.
• Numbers unchanged: yield, DPPM, and test results stay the same; stability is a read-only overlay.
• Easy to use: spreadsheet/BI friendly; a lightweight weekly quality ritual.
• Universal language: A++ / A+ / A0 / A- / A– aligns fab, test, assembly, quality, and customer teams fast.
CLI 💻 — try our mini Calculator to identify the drift
(Mini CLI Download Page)
Feed your CSVs and see bands and drift at a glance (numbers unchanged).
# Lot-to-lot yield stability (variance at constant mean)
ssm_audit_mini_calc semi.csv --kpi "Lot Yield Stability" \
--out bands_lotyield.csv --plot_kpi "Lot Yield Stability" --build_id semi
# Parametric margin stability (corner guardband health)
ssm_audit_mini_calc semi.csv --kpi "Parametric Margin Stability" \
--out bands_margin.csv --plot_kpi "Parametric Margin Stability" --build_id semi
# Burn-in escape stability (early-life failures post burn-in)
ssm_audit_mini_calc semi.csv --kpi "Burn-In Escape Stability" \
--out bands_biesc.csv --plot_kpi "Burn-In Escape Stability" --build_id semi
# Test-escape rate (system test failures after final pass)
ssm_audit_mini_calc semi.csv --kpi "Test-Escape Rate" \
--out bands_esc.csv --plot_kpi "Test-Escape Rate" --build_id semi
# Package/assembly reliability stability (reflow/void clustering)
ssm_audit_mini_calc semi.csv --kpi "Package Reliability Stability" \
--out bands_pkg.csv --plot_kpi "Package Reliability Stability" --build_id semi
# Field RMA cadence (DPPM over time, bunching)
ssm_audit_mini_calc semi.csv --kpi "Field RMA Cadence" \
--out bands_rma.csv --plot_kpi "Field RMA Cadence" --build_id semi
Outputs you will get:
• CSVs with stability bands for each timestamp (e.g., bands_margin.csv).
• Drift charts per KPI (--plot_kpi) showing exactly where reliability thins.
• Optional alerts if you enable thresholds in your setup.
Technical notes
Representation: x = (m, a) with a in (-1, +1)
Collapse parity: phi((m,a)) = m
Order-invariant pooling:
U = sum(w_i * atanh(a_i))
W = sum(w_i)
a_out = tanh( U / max(W, eps_w) )
Typical bands (example):
A++: a >= 0.75
A+: 0.50 - 0.75
A0: 0.25 - 0.50
A-: 0.10 - 0.25
A--: a < 0.10
Navigation
Back: SSM-Audit Q&A Series – Telecom QoE (Question 28)
Next: SSM-Audit Q&A Series – Housing Early-Warning (Question 30)
Page disclaimer
Illustrative scenario for research and education. Observation-only; do not use for critical decisions without independent validation.
SHUNYAYA 