How wemodel flavour

We begin by analysing thousands of recipes across diverse culinary traditions—Italian, Thai, Mexican, Japanese, French, Indian, and beyond. This recipe data reveals which ingredients cultures around the world have successfully paired through centuries of experimentation. When tomato and basil appear together in thousands of Italian dishes, or ginger and soy sauce dominate Asian cuisine, these patterns represent accumulated culinary wisdom.

We use statistical analysis to identify ingredient combinations that appear together more frequently than random chance would predict. This helps us distinguish meaningful pairings from coincidental co-occurrences. Popular combinations become our foundation—proven patterns that guide our understanding of flavour harmony.

Each ingredient in our database contains detailed information about its aroma compounds—the volatile organic molecules that create what we perceive as flavour. Using data from gas chromatography-mass spectrometry analysis and scientific literature, we map the chemical fingerprint of over 4,250 ingredients.

We then analyse successful recipe combinations to identify the underlying chemical patterns. Why do strawberry and basil work so well together? Both share linalool (floral notes) and certain esters (fruity sweetness). This chemical overlap creates harmony. Some pairings work through shared compounds, while others succeed through complementary profiles—contrasting elements that balance each other.

Beyond chemistry, we analyse how people actually describe ingredients. We process tasting notes, culinary texts, professional reviews, and sensory studies to extract flavour descriptors—words like "citrusy," "earthy," "floral," or "umami." Using natural language processing, we transform these descriptions into structured data that captures the subjective experience of taste.

This text-based approach complements our chemical analysis. Some flavour relationships are cultural or contextual in ways that molecules alone don't fully explain. The combination of chemical data and human sensory language gives us a more complete picture.

For each ingredient, we create a comprehensive flavour profile that includes:

• Chemical composition: Key aroma compounds and their concentrations
• Sensory descriptors: How the ingredient tastes and smells to human palates
• Culinary context: How it's traditionally used across different cuisines
• Pairing patterns: Which ingredients it commonly appears with in recipes

These profiles become the foundation for all our recommendations.

With thousands of flavour profiles mapped, we can identify harmony patterns—the principles that make ingredient combinations successful. We've learned that certain chemical families pair well together, that specific flavour note combinations appear repeatedly across cuisines, and that some pairings work through similarity while others succeed through contrast.

When you search for pairings on our platform, we analyse multiple factors:

We score pairings based on both scientific evidence and culinary precedent, giving you recommendations you can trust.

For famous combinations, we show you exactly why they work. Chocolate and coffee both contain roasted pyrazines that reinforce each other. Lamb and rosemary pair well because rosemary's camphoraceous notes mask gamey flavours while complementing richness. Understanding these principles helps you develop better intuition about flavour.

Our approach also enables discovery beyond traditional boundaries. By understanding which flavour patterns create harmony, we can suggest innovative pairings that haven't been widely explored but follow the same chemical and sensory logic as proven combinations.

If yuzu and ricotta share similar compound profiles with lemon and cream cheese—a classic pairing—we can confidently suggest this less common combination. These predictions are grounded in the same principles that explain why traditional pairings work.

Our system learns and evolves. As users interact with the platform—rating pairings, flagging inconsistencies, and contributing their experience—we refine our models. Professional chefs, food scientists, and home cooks all bring valuable perspectives that help us improve recommendations over time.

This methodology represents our commitment to combining the best of culinary tradition with rigorous food science, giving you pairing suggestions that are both reliable and inspiring.