The science behind Palate DNA: how we map your flavor fingerprint

When you open most food apps, you're looking at a ranked list of restaurants sorted by average star rating, distance, or whatever the algorithm decided was trending that week. It's useful. But it's not personal.

The question we kept asking ourselves was: what would it look like if a food app actually understood your palate? Not just your order history. Not just your neighbourhood. Your actual sensory preferences — the things that make a dish feel right to you on a deep, almost instinctive level.

The five dimensions of taste

Palate DNA maps your flavor fingerprint across five sensory dimensions, each scored from 0 to 100:

• Umami (0–100) — your affinity for savory, brothy, deeply complex flavors. High umami scorers tend to gravitate toward aged cheeses, mushrooms, soy-braised proteins, and fermented foods.
• Spice (0–100) — not just heat tolerance, but your enjoyment of it. A score of 80 means you actively seek the burn; a score of 20 means you want flavor complexity without the fire.
• Richness (0–100) — your preference for fat, creaminess, and indulgence. High richness scorers love butter-forward sauces, fatty cuts, and full-fat dairy. Low scorers prefer clean, bright, lean preparations.
• Acidity (0–100) — your appetite for brightness, tang, and citrus-forward flavors. High acidity scorers love ceviche, vinegar-dressed salads, and fermented pickles. Low scorers prefer mellow, round flavor profiles.
• Sweetness (0–100) — your tolerance for and enjoyment of sweet notes in savory contexts. High sweetness scorers enjoy teriyaki glazes, honey-roasted vegetables, and sweet-spicy combinations.

These five dimensions aren't arbitrary. They map directly to the primary taste receptors on the human palate, and they're the dimensions that professional flavor scientists and Michelin-trained chefs use to describe and categorize dishes.

How we build your profile

When you first open Taste Twins, you go through a two-minute onboarding flow — the Sensory Sliders. You're not answering abstract questions about your "food personality." You're making real choices: drag the Umami slider to where it feels right, then the Spice slider, and so on.

But that's just the seed. The algorithm learns continuously from every interaction you have in the app:

• When you swipe right on a dish, we extract its flavor profile and weight your scores accordingly.
• When you save a restaurant to a plan, we analyze its menu and update your profile.
• When you rate a meal, we use that signal to fine-tune the model.

Over time, your Palate DNA becomes a living, evolving fingerprint — one that gets more accurate the more you use the app.

The matching engine

Once we have your five-dimension profile, finding your Flavor Twins is a geometric problem. We represent each user as a point in five-dimensional flavor space. Your Twins are the users whose points are closest to yours — measured by a weighted cosine similarity that accounts for the relative importance of each dimension.

The result is a match percentage: the famous 98% you see in the app. That number isn't marketing. It's a real similarity score, and it predicts with remarkable accuracy whether two people will love the same restaurant.

In our private alpha, users with a match score above 90% agreed on restaurant choices 94% of the time. That's the number we're most proud of.

What's next

We're currently working on two major extensions to the Palate DNA model: texture preferences (crunchy vs. soft, dense vs. airy) and cuisine-level affinity mapping (not just "do you like spicy food" but "do you like the specific spice profile of Sichuan cuisine vs. Nashville hot chicken vs. Thai green curry").

The goal is a palate model that's granular enough to recommend not just a restaurant, but a specific dish — and tell you exactly why it's right for you.

If you want to be among the first to try it, join our private alpha waitlist below.