Every lipstick tells a story of beauty. But for Marie, Quality Manager at one of the most prestigious cosmetics brands in Europe, every lipstick also told a story of increasing pressure. "With hundreds of shades, constant launches, and luxury consumers demanding perfection, our quality control teams were facing a tough battle," she recalls.

At the beginning of 2023, in her factory, Marie watched her quality control operators meticulously inspect thousands of lipsticks. Despite their expertise, she knew they were facing an impossible task. Operator fatigue led to inconsistent results: some defective products reached customers, while perfectly fine lipsticks were thrown away. With rapid innovation cycles in the beauty industry requiring increasingly frequent product launches, the traditional manual inspection system had reached its breaking point.

"We were losing talented operators to burnout, and our waste rates were impacting both our sustainability goals and our financial results," explains Marie. "When you handle premium products at our scale - across multiple production lines in Europe and the U.S. - even a 1% error rate leads to significant losses."

The challenge was complex:

• Modern production lines coexisted with equipment dating back several decades

• Each product launch brought new quality control challenges

• The premium market position demanded zero tolerance for defects

• Labor shortages made recruiting quality control operators increasingly difficult

• Visual fatigue led to inconsistent inspection results

Previous attempts to automate quality control had failed. "We tested many inspection systems, Marie confides, but none could match the expert eye of our operators - until now."

The beauty industry's struggle with automated quality control is deeper than it seems. "Vision systems have long been the benchmark in many manufacturing sectors, Marie explains, but luxury cosmetics pose unique challenges that have thwarted even the most advanced systems."

Traditional vision systems face multiple obstacles in inspecting lipsticks:

• Highly reflective surfaces creating unpredictable light patterns

• Subtle color variations that can mask real defects

• Complex geometric shapes requiring inspection from multiple angles

• Hundreds of different shades and finishes to master

• New formulations constantly pushing the boundaries of traditional inspection methods

"What makes their algorithm better, when this is a notoriously difficult use case in the vision industry?" Marie's question goes to the heart of a challenge that has frustrated quality leaders in the cosmetics industry for years. Previous attempts at automation often failed in three ways:

• False positives leading to excessive waste

• Missed defects due to complex surface properties

• Inability to adapt to new product variants

Industry statistics were telling:

• 75% of vision systems installed in cosmetic quality control did not meet expectations

• Traditional systems required up to 3 months of recalibration for each new product launch. Even AI-based systems were too slow to adapt due to their reliance on internal or external AI engineers for program creation.

• Most factories resigned to manual inspection as the "necessary evil" of luxury cosmetics production

This context makes the transformation even more remarkable. As a Quality Control Supervisor noted: "In the first 30 minutes of inspection, performance was similar between humans and Scortex. After 30 minutes, human attention was already declining, while Scortex remained steady. The rejection rate is the same, but if it had been only humans, more missed defects would have occurred."

This advance challenges long-held industry assumptions about the capabilities of automated inspection. While other sectors, such as automotive or electronics, had automated their quality control years ago, luxury cosmetics remained an unexplored frontier - until now.

The Path of Transformation

After thorough research, Marie and her team collaborated with us to implement an AI-based quality control system capable of matching - and ultimately surpassing - human inspection capabilities. What distinguished this solution was particularly striking for the Quality Team Leader: "What makes their algorithm better, when this is a notoriously difficult use case in the vision industry?"

The implementation revealed a compelling answer through direct performance comparisons. As noted by a Quality Control Supervisor: "In the first 30 minutes of inspection, performance was similar between humans and Scortex. After 30 minutes, human attention was already declining, while Scortex remained steady. The rejection rate is the same, but if it had been only humans, more missed defects would have occurred."

Key steps of the transformation:

• Maintaining a constant defect detection rate throughout shifts

• Eliminating the variable of operator fatigue in quality control

• Uniform inspection quality, regardless of shift duration or time of day

The results: beyond quality control

Within six months, the transformation exceeded the most optimistic expectations:

Impact on quality:

• Improvement in defect detection rate, ensuring no critical defect is delivered

• Reduction of false rejections, maintaining a stable rejection level despite increased ejection of critical defects, preventing unnecessary waste of high-end products

• Maintaining quality control consistency even during 8-hour shifts

Human impact:

"The role of our quality control team has evolved from repetitive inspection to high-value quality management," Marie enthusiastically states. "Instead of suffering from visual fatigue, our operators now focus on continuous improvement and innovation."

Business impact:

• Reduction in operational costs through the automation of manual tasks, decreased waste, and improved efficiency. Up to 2 operators per station were removed from the inspection task.

• 25% increase in production capacity due to faster and more reliable inspection

• Reduction in new product launch time due to quick creation of inspection programs: only 15 minutes to create a new inspection program

"The strength of these results lies in their consistency. While initial rejection rates remained similar to those of human inspection, the unwavering attention of the system ensured that fewer defective products reached customers - a crucial factor for a luxury brand where each product represents not only a sale but also the brand's reputation."

Vision for the Future

Building on this success, Marie and her team are now exploring additional applications of the technology. "This journey has shown us that when you combine human expertise with the right technology, you don't just solve problems - you create new possibilities," she concludes.

The Scortex system will be deployed on more than 10 lines in the client's factories.

To illustrate how AI enables the overcoming of traditional approaches, we explain in detail how to go beyond quality sorting machines with AI in a dedicated article.

Here are other articles that may interest you:

• Quality inspection of shiny parts: an impossible challenge?

• Customer interview: Automating quality control at Toly using Scortex's Spark technology

• Spark for the cosmetics industry

• Spark Multi View Video: up to 4 cameras simultaneously to detect multiple defects in a single pass