Thinklytics

Flex Ltd. (Automotive Division) · Manufacturing · Detroit, MI · 18 weeks

Quality data foundation built across 4 plants

A Tier 1 auto parts supplier faced high defect rates, 21 per 1,000 units, leading to $4.6M in annual warranty claims. Quality data was scattered across four plants, preventing clear visibility. We consolidated the data into a single system, enabling targeted quality improvements. This cut defects to 4 per 1,000 units and slashed most warranty claims.

Challenge

The supplier’s four plants tracked quality data in separate systems, each using its own defect classifications. Without a unified view, they couldn’t spot patterns or root causes across plants. With 21 defects per 1,000 units, warranty claims hit $4.6 million annually, putting their preferred supplier status with their biggest OEM customer at risk.

Approach

We created a consistent quality data system on Azure Synapse that aligned defect categories from all four plants. Then, we developed a real-time dashboard to catch process deviations immediately. We also built an analysis tool to identify recurring defect patterns affecting multiple plants.

Outcome

We cut defect escapes from 21 to 4 per 1,000 units in six months by refining inspection processes and targeting failure modes. This drove warranty claims down from $4.6 million to $880,000 annually. As a result, our biggest OEM customer renewed their preferred supplier status, directly attributing the decision to the improved product quality.

We made sure all four plants spoke the same language by standardizing defect categories. No more confusion, just clear, shared info.

Each plant was using its own old defect tracking system, so the data was all over the map and didn’t really line up. We spent about a month gathering everything and building one standard way to classify defects. That step was crucial, it let us spot trends across all plants and dig into the bigger problems hiding underneath.

How We Tracked Process Quality in Real Time

So, we built this SPC dashboard to track 240 process parameters across four plants, all happening in real time. The best part? It flags any out-of-control issues within 15 minutes. In the first month alone, it caught 8 problems the old system totally missed. That meant zero defective batches got through.

We noticed quality problems showing up at several plants. So, we jumped right in to fix them.

So, we took a deep dive into defect data from all the plants and noticed the same three issues kept showing up at different sites. The teams hadn’t caught these because they were only looking at their own local data. Once we zeroed in on those problems and tackled them directly, we cut warranty claims by $2.2 million. That was most of the $3.7 million we saved overall.

Results

  • $4.6M to $880K Annual warranty claims
  • 21 to 4 per 1,000 units Defect escapes per 1,000 units
  • 4 Plants on unified quality data
  • 18 wks Foundation delivered

We were losing $4.6 million a year on warranty claims, and our biggest customer was close to dropping us as a preferred supplier. Thinklytics helped us find issues we hadn’t caught because they were happening across all four plants. After that, we brought the defect rate down from 21 to 4 per 1,000 units.

VP of Quality, Tier 1 Auto Parts Supplier

In the client's terms

The question

Where does this number come from, and can the path be trusted?

The situation

Four plants tracked quality in separate systems with their own defect classifications. At 21 defect escapes per 1,000 units, warranty claims ran to $4.6M a year and preferred supplier status with the largest OEM customer was at risk.

What we found

Each plant was looking at its own defects and seeing noise. The same failure modes were recurring across plants, and because the classifications did not match, nothing that would have made the pattern visible could be run.

What it produced

Defect escapes fell from 21 to 4 per 1,000 units in six months as inspection was retargeted at the failure modes that analysis exposed. Warranty claims fell from $4.6M a year to $880K. The OEM renewed preferred supplier status and cited the quality improvement.

The roadmap

A consistent quality data layer on Azure Synapse aligned defect categories across all four plants, which is what makes cross-plant comparison mean anything. A live view flags process deviation as it happens rather than in a monthly review, and a pattern analysis tool surfaces failure modes recurring in more than one plant.

The work behind this result

Related case studies

Frequently asked questions

Why do defect patterns go unnoticed across plants?

Because each plant only sees its own. Flex's automotive division ran four plants with separate defect classifications, so cross-plant comparison produced nothing. Unified classification exposed three failure modes recurring at multiple sites.

What do defect escapes cost?

At 21 per 1,000 units this supplier was paying $4.6M a year in warranty claims and had preferred supplier status at risk with its largest OEM customer. Cutting escapes to 4 per 1,000 took claims to $880K and the OEM renewed, citing the quality improvement.

How fast should process deviations be detected?

Within minutes on a production line. The SPC layer here tracks 240 parameters across four plants and flags out-of-control conditions inside 15 minutes. In the first month it caught 8 deviations the previous system missed and no defective batches reached the customer.