Outbound defect rate 2.8% → 1.6% — main-line OEE up to 74%.
A trillion-VND manufacturing group with 2 plants standardized in-process quality control and runtime-based maintenance — defects are stopped where they occur instead of being caught at final inspection.
Figures supplied by the customers themselves and published with their consent, measured within the scope and period stated on each metric. Results vary with company size, industry and how standardized the starting data was.
TASA Group is a trillion-VND-revenue industrial manufacturer with 2 plants, supplying both the domestic market and export customers that run scheduled quality audits.
When the two sides first sat down, quality control was still running on paper forms and spreadsheets: defects were mostly caught at final inspection (2.8% of shipped volume) and tracing the originating step took days; OEE wasn't measured continuously, so unplanned downtime sat around 9% of machine hours with hand-written, unclassified causes; and the 40-person QC team spent roughly a third of its time on data entry and report assembly.
The group went in two phases, entirely on its own infrastructure. Phase one: quality control step by step on the line, with in-process quality gates and one standardized defect-code set shared by both plants. Phase two: maintenance moved from breakdown response to schedules based on real running hours, OEE measured automatically from the machines, all of it rolled up on a BI dashboard.
Before and after — with timeframe and method.
| Metric | Before | After | Timeframe & method |
|---|---|---|---|
| Outbound defect rate | 2.8% | 1.6% | After 12 months · every shipped lot, both plants |
| Main-line OEE | 61% | 74% | Measured automatically from machines · 3-month average |
| Unplanned downtime | ~9% of machine hours | 6.2% of machine hours | EAM: maintenance by actual runtime instead of a fixed calendar |
| QC team productivity | — | +35% | Inspected volume per person · data entry & reporting automated |
Figures supplied by the customers themselves and published with their consent, measured within the scope and period stated on each metric. Results vary with company size, industry and how standardized the starting data was.
From discovery to go-live.
The same defect had two names and two thresholds depending on the plant — turning on automatic alerts at that point would only have produced false alarms. The project deliberately paused the rollout, sat both plants' QA teams down to standardize the 214-item defect-code set and shared inspection thresholds, and only then enabled alerting. Those six weeks are why the 12-month numbers can be trusted.
“The 1.6% isn't what I value most. What matters is that when defects tick up now, we know which step and which shift immediately — before the lot leaves the plant, not after a customer complains.”
Other stories
Consolidated close across 14 campuses: from 12 days to 5 — payback in 9 months.
Read case study
Work-order cycle down 42% — costing closed in 4 days instead of 9.
Read case study
6,600 outlets, 92% of orders in the app — dealer receivables from 47 days to 31.
Read case study