Your Cart Has a Story. Here's How We Learned to Read It

Back to Devoured - June 04, 2026

DEVOURED

Your Cart Has a Story. Here's How We Learned to Read It

Data airetail Zepto

Zepto developed a Cart Contextual Model using a Transformer-based masked language model to predict user purchases in real time from shopping cart "sentences."

What: Zepto's new Cart Contextual Model treats shopping carts as "sentences" and applies a Transformer-based masked language model (MLM). It's trained on historical cart data, incorporating temporal, geographical, and product signals, along with inverse-frequency masking for long-tail items, to predict what users will buy next.

Why it matters: This demonstrates how e-commerce companies are adapting advanced NLP techniques, typically used for human language, to understand and predict complex user behavior patterns in transactional data, moving towards more intelligent, real-time personalization.

Decoder

Masked language model (MLM): A type of language model that is trained to predict missing words in a sentence, often by masking out certain words and trying to reconstruct them based on context. This allows the model to learn bidirectional relationships between words/items.
Transformer-based: A neural network architecture that revolutionized natural language processing, known for its attention mechanism which allows it to weigh the importance of different parts of the input sequence.

Original article

Zepto built a Cart Contextual Model that treats shopping carts as “sentences” and uses a Transformer-based masked language model (MLM) to infer user intent in real time as items are added. By training on historical cart patterns with temporal, geographical, and product signals plus inverse-frequency masking to handle long-tail items, the model predicts what else the user will likely buy.

Explore all 55 articles in today's digest