Authentication no longer depends on inspecting a single item in isolation. It depends on recognizing how that item fits within a much larger body of data. As counterfeit materials improve, the difference between real and fake often emerges only when compared across many samples.
That requirement introduces a different kind of analytical challenge. Instead of working with controlled matrices, authentication workflows must account for variation across materials, finishes, and manufacturing processes.
“Consumer products are more complex, using different materials, coatings, and variations,” reveals Jina Kim, Head of the Scientific Authentication Division at Bunjang Global and CEO of InsightviewTech. “The biggest challenge was dealing with that variability and building a reliable data set.”
Scale Begins with Data, Not Instrumentation
The ability to distinguish subtle differences depends less on the instrument and more on the dataset behind it. Without sufficient reference points, even high-quality measurements offer limited context.
“We’re sometimes analyzing 300 to 500 items every day,” Kim explains. “Over three years, we’ve been building data sets across 31 brands with more than 1,000 models.”
Working in analytical science?
Register for a FREE Separation Science account to subscribe to the Separation Science Newsletter.
As the dataset grows, patterns become more stable. Differences that appear insignificant in small samples begin to separate clearly when viewed across thousands of entries.
“We have collected more than 25,000 sets, and that helps us to see the difference,” Kim asserts. At that scale, authentication shifts from individual judgment to structured comparison.
Reducing Dependence on Expert Interpretation
Traditional authentication relies on trained specialists who can interpret subtle signals and make final determinations. That model limits throughput. Training alone creates a barrier to scaling.
“You need two to three years of training to make a final decision,” Kim notes. “That requires companies to invest a lot of time and money.”
As demand increases, workflows must distribute that expertise more efficiently. AI-assisted interpretation provides a way to embed decision logic into the system rather than relying entirely on individual experience.
Even so, full automation remains out of reach. “We do achieve 99% accuracy, but that 1% can make a huge difference,” Kim explains. “I still believe that human experts are needed.”
The result is not a replacement model, but a redistribution of effort. Systems handle routine classification, while experts focus on edge cases where uncertainty remains.
Extending Analytical Workflows Beyond the Lab
Scalability also depends on where decisions are made. Laboratory workflows introduce delays that limit responsiveness, especially in enforcement or supply chain settings.
“We are trying to incorporate machine learning tools into the apps,” advises Kim. “When they scan the XRF, they will be able to get the results right away.”
This shift allows authentication to move closer to the point of need. Instead of transporting samples to a lab, users can generate and interpret data on-site.
“There is demand for tools that can be used in the market, for real-time determination of whether a product is real or fake,” explains Kim. This possibility for field deployment shifts the role of analytical science from verification to intervention.
A Model Built for Increasing Complexity
Authentication workflows now operate under conditions that continue to evolve. Materials vary more widely, counterfeit quality improves, and the volume of items requiring verification grows.
“There is potential for use in almost any area that requires authentication,” Kim advises, adding that it’s particularly useful in cases where non-destructive analysis is required.
Meeting that demand requires systems that combine scale, speed, and interpretability. Large datasets provide context, AI translates data into decisions, and portable tools extend those decisions beyond the lab. Together, they define a model built to keep pace with a problem that is not standing still.
