AI infrastructure for predictive consumer product development

The first-principles model for human taste, smell, and texture preference.

The most advanced AI solution that creates digital twins of consumers and chemistry, built to design and deconstruct products and ingredients that consumers will love.

See what it does

A person smelling a flower beside a callout reading AI Model, Taste and Smell, Trained.

Origins in wine. Now scaling into ready-to-drink, carbonated soft drinks, energy drinks, spirits, and more.

Consumer companies have spent decades summarizing what consumers say they want. We build the model that knows…not from words, but from chemistry.

Read the science

Where chemistry meets preference

01. The problem

Product development
is still trial and error.

For every product on a shelf, there are five that failed quietly. Reformulations that lost consumers. Line extensions that missed. Innovation cycles that took 18 months and ended in a focus-group shrug. The cost is structural. Teams iterate physically, validate after the fact, and discover what consumers actually wanted long after the budget is spent. Meanwhile, fortunes are being poured into AI that was never built for this, models trained on words and clicks that predict the past instead of what people will actually love.

The cost of the broken cycle

$1T

Spent annually launching consumer products

85%

Of new product launches fail

18 to 24 mo

Industry-standard time to market

02. First principles

Preference lives in chemistry,
not in words.

For decades, the industry has tried to predict what people want by mapping language (fruity, smooth, clean) back to ingredients. That mapping does not hold up. A consumer who says they like cherry will reject a wine engineered to taste like cherry. The descriptor is the echo. The chemistry is the cause.

Tastry's model is built from the molecule up. We capture the full symphony of chemistry in a product in our proprietary in-house lab, then connect that chemistry directly to consumer preference. The output is a prediction of how a real audience will feel about a product that does not exist yet.

The result is something the industry has wanted for a century. A direct, causal link from a product's chemistry to a consumer's preference. Words do not predict it. Chemistry does.

One molecule, three descriptors: benzaldehyde reads as cherry, marzipan, or nutty. Language is a poor proxy for preference.

What the consumer says

I like cherry.

What the consumer does

But I do not like this.

Read the science

03. The model

Digital twins of
chemistry and consumers.

One twin captures the complete chemistry of a product. The other captures how a real population will respond to it. Put them together and a brand can build a new formulation, reverse-engineer an existing one, and know who will love the result, before a physical sample is ever made.

PILLAR 01

Proprietary chemistry dataset

Every product captured as the full symphony of chemistry, not a sampled summary. The dataset is built category by category, geography by geography, and owned by Tastry.

PILLAR 02

Consumer digital twins

Real human preference, modeled. Each twin learns from sensory and behavioral signal so we can predict what a population will feel about a product that does not exist yet.

PILLAR 03

Future-proofed formulations

Reformulate against cost, regulatory pressure, or supply changes without losing the consumer. The model knows which substitutions preserve preference and which destroy it.

See the platform

U.S. Patent No. 11,847,684. Issued in the United States, China, Japan, and Canada.

04. What it does

Consumer-centric insight,
from first concept
to shelf.

Tastry does not replace one tool. It replaces the guesswork inside every stage of the development cycle, from forecasting demand to engineering the formulation to targeting the audience most likely to love it.

A US map with a single location pinpointed, representing nationwide opportunity forecasting.

01. STRATEGY

Strategic intelligence

Forecast store, local, regional, and nationwide opportunity in real time, before a brief is written.

Two overlapping circles with a separate dashed circle, representing SKU overlap and an unmet market gap.

02. PLANNING

Assortment planning

Identify gaps in the market. Quantify overlap between SKUs. Brief the next launch with intent.

A radar sweep locking onto a single target, representing a winning formulation found among many candidates.

03. R&D

Product development

Recipes engineered using real-time chemistry and preference data. A winning formulation in weeks, not years.

A crowd of people with the highest-value individuals highlighted in accent color.

04. TARGETING

Consumer targeting

Match brands to their highest-value customers. Segment by predicted depth of preference, not by lifestyle.

05. Proof

The model has
already been
graded.

Across all products tested with consumers in double-blind panels, Tastry's predicted preference scores matched actual consumer scores at 93%.

Tastry AI predicts how consumers will score a wine before it is launched: predicted and actual scores track at 93% accuracy across all wines tested.

“This is simply amazing data and information. To the general public, my guess is that any of the Tastry blends would be very well received.”

Josh Baker. Winemaker, Phase 2 Cellars

The recipe nobody else could match

Tastry delivered a custom flavor
in two weeks that multi-billion-dollar companies failed to match in 18 months.

01. The brief

A global CPG customer's target flavor.

A single, specific flavor profile. The brief sent to the best flavor houses in the world.

02. 18 months. 9 attempts.

The global flavor houses missed every single time.

Three of the largest flavor houses on earth. Nine attempts over 18 months. None succeeded.

03. Tastry. 2 weeks.

Matched the target on the first try.

One pass through the model. From chemistry alone, with no language descriptors involved. Two weeks from brief to match.

06. Where we started

We started with the
hardest sensory category
in the world.

The origin

Why wine. And why every CPG company is next.

Wine is the most chemically complex consumer product on earth. Hundreds of thousands of chemistry signals, every bottle different, preference impossible to predict by language. If the model could crack wine, it could crack anything. It did. Now it is coming for the rest of CPG.

See the wine work