Case Study Consumer Goods

Achieving Six Sigma quality on a production line with persistent defect variance

A consumer goods manufacturer achieved Six Sigma quality on a production line that had never broken below 3.2 sigma in seven years — by identifying a cross-sensor correlation pattern that was invisible at sampled data volumes.

3.2σ → 6σ

Quality improvement on target production line

67%

Reduction in defect rate

$14M

Annual waste reduction

7 years

Duration of unexplained quality variance — resolved in 3 weeks

100%

Sensor channel coverage required to surface the pattern

3 weeks

From full coverage deployment to root cause identification

The Organisation

Global Consumer Goods Manufacturer · Consumer Goods

The Challenge

One production line had produced quality variance that resisted every improvement initiative for seven years. Extensive sampling-based analysis had failed to identify a consistent root cause. The process engineers had concluded the variance was inherent to the product design. The data to identify the actual cause existed in the sensor stream — a subtle interaction between temperature, humidity, and line speed — but it only became statistically visible when all three sensor channels were analyzed together at full temporal resolution.

The Approach

Full sensor coverage with LLM reasoning configured to identify multi-variable correlations across sensor channels at full temporal resolution. The model was tasked with finding patterns that explain defect occurrence — not just monitoring for known anomalies.

"Seven years of improvement initiatives, Six Sigma projects, and process redesigns. The answer was in the data the whole time — we just couldn't see it because we were working with a fraction of what the sensors were telling us."

Global Head of Manufacturing Excellence

Key Finding

The root cause was a three-way interaction: when ambient humidity exceeded 68%, line speed was above 94% of rated capacity, and the primary heat exchanger was in the second half of its maintenance cycle, defect probability increased by 340%. Each variable individually was within specification. The combination was the cause. This pattern required all three sensor channels at full temporal resolution to become statistically significant — it was completely invisible in sampled data.

Results at a Glance
Quality improvement on target production line 3.2σ → 6σ
Reduction in defect rate 67%
Annual waste reduction $14M
Duration of unexplained quality variance — resolved in 3 weeks 7 years
Sensor channel coverage required to surface the pattern 100%
From full coverage deployment to root cause identification 3 weeks
Get in Touch

Talk to us about your data.

Tell us about your event stream and we'll show you what full LLM reasoning coverage looks like for your environment.

Or book a call directly →