Understanding Taste Trends Using Machine Learning

Quick Summary

Challenge

A FoodTech company needed to analyse fragmented and unstructured menu data to understand evolving food trends across geographies.

Solution

Tatras Data developed an AI-powered platform using web scraping, embeddings, and machine learning to classify and analyse food data at scale.

Result

The solution enabled data-driven insights into regional trends, improving decision-making and product strategy.

Tech Stack

AI: Machine Learning Models Pretrained Embedding Models (NLP) Semantic Analysis | Dev: Web Scraping Frameworks Data Pipelines & ETL Systems Dashboarding & Visualisation Tools

The Challenge

A FoodTech innovator wanted to understand evolving food trends across different geographies to drive better business decisions.

However:

Data on menus and dishes was fragmented across platforms
Trends were difficult to identify due to unstructured and inconsistent data
There was no unified way to categorise cuisines, dishes, and flavours
Businesses lacked actionable insights into regional preferences and emerging trends

The client needed a solution to aggregate, structure, and analyse food data at scale to uncover meaningful trends.

A Day in the Life: Before Our Solution

A product or marketing team tries to understand emerging food trends. They rely on manual research, anecdotal insights, or limited datasets.

As the process unfolds:

Trend identification is slow and subjective
Insights lack scale and consistency
Regional variations are hard to quantify
Strategic decisions are made with incomplete data

For the business:

Opportunities to capitalise on trends are missed
Product innovation is less targeted
Competitive advantage is limited

👉 The result: fragmented insights, slower decision-making, and missed market opportunities.

Solution

1. Core Innovation

Tatras Data developed an AI-powered food trend analytics platform using machine learning and data pipelines. The solution:

Built web scrapers to extract menu data from food aggregators
Used pretrained embedding models to capture semantic relationships between dishes and cuisines
Applied machine learning algorithms to classify restaurants, dishes, and flavours
Standardised data into structured taxonomies for analysis
Created dashboards to visualise regional trends and evolving preferences
Enabled periodic updates through automated data pipelines

2. Key Features

Web scraping and large-scale data aggregation
Semantic analysis using pretrained embeddings
Machine learning-based classification of dishes and cuisines
Trend analysis across geographies
Dashboarding for business insights
Automated data pipelines for continuous updates

3. What Made This Hard

Processing large volumes of unstructured menu data
Capturing nuanced relationships between dishes, cuisines, and flavours
Ensuring consistent classification across different regions and formats
Designing taxonomies that are both comprehensive and actionable
Maintaining up-to-date insights with continuous data ingestion

Outcomes

✅ Enabled data-driven identification of food trends across regions

✅ Improved visibility into evolving consumer preferences

✅ Provided actionable insights for product and marketing strategies

✅ Automated data collection and analysis workflows

✅ Delivered scalable infrastructure for continuous trend monitoring