Repair Mistakes

Where Electronics Depot Repair Numbers Go Wrong: 8 Costly Mistakes

The eight measurement and process errors that make depot repair numbers lie, each with the symptom that gives it away and a numeric fix.

Symptom: your first-pass yield reads 94 percent but customers keep returning units. Root cause: you are counting a board as passed the moment it clears functional test, before burn-in. Many latent faults, cold solder joints and marginal electrolytics, only surface after 4 to 8 hours at 55C. Fix: measure First-Pass Repair Yield at final gate, after burn-in, not at the bench. A depot running 96 percent bench yield often drops to 88 percent true yield once burn-in escapes are counted. That 8 point gap is the number quietly feeding your return stream.

Symptom: No-Fault-Found units pile up and techs swear the boards are fine. Root cause: you are treating NFF as a diagnostic failure when 40 to 60 percent of NFF cases are actually intermittent faults that need vibration or thermal cycling to reproduce, plus incomplete failure descriptions from the field. Fix: require a symptom code on every RMA and re-test suspected NFF units under thermal cycle from 0C to 60C. Track No-Fault-Found Rate separately for reproduced versus truly clean. Cutting NFF from 25 percent to 12 percent on a 500 unit monthly stream recovers roughly 65 units of avoidable teardown labor.

Symptom: quoted repair time is 30 minutes but the floor averages 55. Root cause: your Board Rework Time estimate captures iron-on-pad time only and ignores setup, reflow preheat, inspection, and rework of adjacent joints disturbed during the pull. A single BGA reball is rarely under 45 minutes once you count 8 to 12 minutes of profile ramp and 10 minutes of X-ray verification. Fix: time-stamp the full cycle from kit pull to final AOI, not the soldering window. Depots that recalibrate this way typically find real rework time runs 1.6 to 2.0 times the raw bench estimate.

Symptom: cost per unit swings 30 percent month to month with no process change. Root cause: you are loading Diagnostic Labor Cost against repaired units only, so a bad batch with many scrapped boards inflates the survivors. Diagnostic time is spent whether the unit ships or scraps. Fix: spread diagnostic labor across all units touched, then track scrap separately. If techs diagnose 600 units at 18 dollars per 20 minute session and only 480 ship, charging the full 10,800 dollars across 480 overstates unit cost by 25 percent versus spreading it correctly across 600.

Symptom: Parts Salvage Value looks great on paper but recovered parts rarely get reused. Root cause: you value salvaged connectors, ICs and modules at list price without derating for removal stress, unknown hours, and no traceability. A reflowed QFP pulled at 245C has lost usable life and cannot go into a warranty repair. Fix: derate salvage to 20 to 40 percent of new for passives and mechanicals, and to zero for heat-sensitive silicon unless retested. A board carrying 60 dollars of nominal salvage often yields 15 to 22 dollars of genuinely reusable value.

Symptom: warranty reserve runs dry two quarters after a product launch. Root cause: you set Electronics Warranty Reserve from a flat failure rate instead of the bathtub curve, so early-life returns at 2 to 4 percent per month swamp a reserve built for a 0.8 percent steady-state rate. Fix: reserve against the infant-mortality hump for the first 90 days, then taper. If field returns run 3 percent monthly for the first quarter on a 10,000 unit shipment, that is 900 units to fund before the curve flattens, roughly triple a naive steady-state estimate.

Symptom: RMA queue lead time is quoted at 5 days but units routinely take 12. Root cause: you measure only active repair touch time and ignore queue dwell, parts backorder wait, and the return leg. In most depots, hands-on repair is under 20 percent of total RMA Queue Lead Time; the rest is waiting. Fix: measure wall-clock from receipt scan to ship scan and segment the dwell. A queue of 300 units clearing 40 per day carries a built-in 7.5 day wait before a tech ever opens the box, no matter how fast the bench is.

Symptom: you keep repairing units that should have been replaced, and repeat-repair units burn margin. Root cause: no refurbish-versus-replace threshold, so techs default to fixing everything because that is what benches do. When cumulative repair cost passes 55 to 65 percent of replacement cost, or a unit returns for a third repair, it is a loser. Fix: run a Refurbish vs Replace Decision Score at intake and gate on it. Scrapping a unit whose repair estimate is 48 dollars against a 70 dollar replacement, with a 40 percent recurrence risk, protects both Cost Per Repaired Unit and your yield figures.

Published 2026-07-01.