Exploring the Science Behind Food and Flavor Analysis

The global food and beverage sector is a competitive place; and if you want to make it to the top, product consistency, quality, and sensory appeal matter more than ever. Consumers expect the food and drink they buy to deliver the same flavor experience across batches and across time. If a product tastes “off” or has an unexpected aroma, companies may be forced to issue costly recalls and suffer reputational damage.

Human taste testers have played a key role in the food and beverage industry for many years; the reactions and feedback they provide can suggest clues as to what is causing a particular nasty taste, for example. Still, flavor remains one of the most complex attributes to objectively measure.

Analytical chemistry approaches are now beginning to close the gap between what we can taste/smell and what we can scientifically measure in a food’s composition.

To learn more about the scientific challenges inherent in measuring flavor, the techniques currently being applied in this field, and how emerging technologies such as machine learning could shape the future, Technology Networks spoke with Manuel Hutterli, PhD, market manager for food flavor & fragrances, and Rohan Thakur, PhD, managing director at TOFWERK.

Flavor analysis is critical for quality control

Can you describe why there is a need for the scientific analysis of food and beverage aroma, flavors, and off-flavors?

Flavor is not just a desirable attribute in food and beverages—it is often the make-or-break factor of product success. Scientific analysis provides a way to prevent issues by identifying off-flavors early and giving producers the opportunity to optimize their formulations.

However, translating sensory perception into measurable chemical data is far from straightforward. The relationship between what we taste and the underlying chemical composition is highly complex.

“With food flavor, it is pretty tricky because you have a very big discrepancy between what you perceive and what the actual composition of the food is. Very tiny amounts of certain compounds can have a huge effect on your perception, whereas large amounts of other compounds might not,” Hutterli explained. “This kind of discrepancy also complicates, to some extent, the tasks in scientific analysis.”

Our overall perception of food and drink flavor is a result of many, many individual ingredients with their own unique flavor profiles coming together to form a larger whole. Trying to tie many ingredients to a product’s flavor is, as Thakur dubbed it, “the ultimate multivariate experiment.”

“Each individual ingredient has its own flavor profile, and when you mix them all together you are assuming that mix is perfect, but actually, it is far from it. The manufacturing process itself adds more complications into the mix,” Thakur added.

Drivers behind food testing demand:

• Flavor directly determines product acceptance, brand reputation, and repeat purchase behavior
• Unpleasant tastes—known as “off-flavors”—can cause costly recalls and reputational harm
• Analytical chemistry enables objective validation alongside sensory testing
• Regulatory and quality control issues further drive the desire for objective measurement

Challenges of analyzing volatile compounds in complex food matrices

Do volatile organic compounds (VOCs) present challenges for analytical measurement?

Volatile organic compounds (VOCs) underpin much of what we perceive as flavor and aroma, but they are notoriously difficult to measure. One of the primary challenges is sensitivity—key compounds relevant to sensory perception often occur at extremely low concentrations. Detecting these trace-level molecules requires highly sensitive instrumentation.

Beyond sensitivity, the complexity of food matrices poses an additional obstacle. Food and beverage systems contain hundreds to thousands of compounds that interact in dynamic and often unpredictable ways. Crucially, as Hutterli noted, flavor is not static; it evolves during consumption, with aroma release changing during chewing, swallowing, and aftertaste. Capturing these temporal changes adds another layer of analytical difficulty.

“Perception is not just one point in time. You can eat something and taste it, but then there’s also an aftertaste. This dynamic system is a challenge; you have to measure the right thing at the right time if you want to analyze your samples,” Hutterli said.

Because of this, proper sample preparation is critical. While VOCs can be measured via headspace analysis—a process that effectively “sniffs” the air above your product—this does not always correlate with what will be experienced during consumption. As you chew, and as your saliva breaks down a product, different VOCs can be released in different amounts.

“With the added chemical diversity of all these VOCs, it is obviously impossible to measure everything with one method; you have to combine different analytical procedures,” Hutterli said.

Key analytical hurdles in VOC measurement:

• Ultra-low concentration compounds require exceptional sensitivity
• Complex mixtures make comprehensive analysis difficult
• Temporal changes in aroma release complicate measurement
• Headspace measurements alone may not fully represent sensory perception

Combining analytical techniques for comprehensive flavor profiling

What analytical techniques are used, and how do they complement each other?

Historically, gas chromatography–mass spectrometry (GC-MS) has been considered the gold standard for food analysis. This is often combined with headspace sampling or other methods, such as solid-phase microextraction (SPME), to extract and concentrate volatile and semi-volatile components in the sample.

Emerging methods are looking to go beyond this. Chemical ionization time-of-flight mass spectrometry (CI-TOF-MS) is quickly gaining momentum in the field as a solution for food and beverage analysis. These systems can detect thousands of compounds simultaneously with extremely high sensitivity and good mass resolution, enabling continuous monitoring of flavor profiles during production. Proton transfer reaction (PTR) is another soft CI technique that can be used in combination with MS to detect trace amounts of VOCs.

“What is fundamental with this chemical ionization is that it is “soft”, so it does not fragment the ions or molecules you are interested in. That means that what you are measuring actually is the parent ion,” Hutterli said. “This is one of the big advantages with CI-MS; you can be more or less selective to different compound classes and really specifically look at what you are interested in without being overloaded with fragment ions.”

“Picture this,” Thakur added. “CI is like showing you a photograph. Electron ionization (EI) is like showing you a jigsaw puzzle of the same photograph and making you depend on software to hopefully put that jigsaw puzzle back together correctly.”

In terms of application, these techniques are inherently well-suited to analyzing the flavor profiles of foods and beverages with pungent aromas, including coffee and beer. Hutterli’s team has also been working with these techniques to identify 2,4,6-Trichloroanisole (TCA) in wines. This compound is a VOC found in cork materials and is responsible for the “corked” flavor taint in some wines.

“It was known for decades that there was this specific flavor to do with cork taint, but it took three decades to actually identify it [as TCA],” Hutterli said. “Once it was identified, we could develop this targeted analysis and replace sensory panels.”

“Rather than having 20 people sniffing corks, with all the dust and everything else which makes it a health issue, we have one machine that can do this 24/7, and at a much lower detection limit,” he added.

Technologies shaping modern flavor analysis:

• GC-MS remains the established standard for compound identification
• Headspace sampling and SPME enable VOC extraction
• CI-TOF-MS enables real-time, high-throughput measurement
• Soft ionization techniques reduce fragmentation, preserving fragile compounds for analysis

The complementary roles of human sensory panels and analytical chemistry

How do analytical techniques complement human taste and smell panels?

Sensory panels made up of human “super-tasters” have a long history in the food and drink industry. The purpose of these improved analytical approaches is not necessarily to replace these roles. But where safety issues exist, the implementation of testing apparatus could cut down the risks to human health.

“It is an environmental safety hazard to have people smelling all of these corks individually,” Thakur continued, referencing Hutterli’s cork analysis work. “Here, we have this environmental or occupational hazard situation that the worker is being relieved from. We can automate such a process, which makes it safer.”

Also key is that, unlike analytical apparatus, humans are uniquely positioned to provide subjective judgments on flavor. Analytical tools should act as a complement to human feedback, which remains necessary if we want to work out which foods and beverages taste particularly rich, tart, earthy, bitter, etc., and why.

“If you really want to understand perception and the correlation between chemical composition, flavors, and aromas, that’s where the human nose and tongue really come into play,” Hutterli said. “They can actually define what tastes nice and what is bad. And then we can move over from the sensory realm to the analytical realm, where we can identify what compounds or mixtures are responsible for this taste and enable more efficient and more quantitative analysis.”

Balancing sensory testing and analytical tools

• Sensory panels establish flavor perception benchmarks
• Analytical methods enable objective, quantitative measurement
• Automation improves safety and scalability
• Combined approaches deliver the most robust insights

AI and real-time analytics are shaping the future of flavor science

What are the key future trends and opportunities in flavor and aroma analysis?

The next frontier in flavor science lies in data integration and predictive analytics, Thakur and Hutterli predict. Real-time mass spectrometry now generates vast datasets that capture dynamic changes in flavor profiles. According to both experts, this data is ideally suited to machine learning approaches that can identify patterns and predict outcomes.

Thakur sees artificial intelligence (AI) and machine learning (ML) as a transformative force, particularly in the context of software that can enable pattern recognition and predictive analytics in food manufacturing. Training better models by producing high-quality training datasets using real-time mass spectrometry data could help to enhance these predictive powers.

“With this real-time analysis that we have, we can measure thousands of compounds at extremely low detection limits, and this naturally opens the door to combine it with AI and machine learning,” Hutterli added. “This is the first time that the data is available in the quality, consistency, and volumes that you need for ML approaches, and I think there is a very bright future in this direction to improve process monitoring, optimization, and quality control.”

Emerging trends in flavor and aroma analytics

• AI and machine learning enable pattern recognition in complex datasets
• Real-time monitoring supports predictive process control
• Improved data quality is unlocking new analytical capabilities
• Applications extend to compliance, optimization, and product development

This content includes text that has been created with the assistance of generative AI and has undergone editorial review before publishing. Technology Networks' AI policy can be found here.