Whatnot's Data Engine Fuels Hyper-Growth

Live-shopping sensation Whatnot has detailed how it navigated its explosive growth, transforming what could have been a data infrastructure nightmare into a competitive edge. At Snowflake Summit 2026, the company shared its strategy for handling billions of daily events while maintaining seamless user experiences.

Whatnot's trajectory has outpaced e-commerce giants, generating $8 billion in gross merchandise volume in 2025 alone and adding over 20 million new accounts. This hyper-growth generates a relentless stream of data from auction bids, chats, and transactions, all crucial for the platform's real-time functionality.

From Bottleneck to Blueprint

Initially, a single data team managed everything using dbt, but this quickly became a bottleneck. Whatnot adopted a modular data stack, enabling individual business units to manage their own Snowflake warehouses and data pipelines via infrastructure as code.

This decentralization, however, introduced a new challenge: maintaining company-wide visibility and control over costs and performance.

AI Analysts for the Masses

To address the human bottleneck of data scientists swamped with ad hoc requests, Whatnot evolved its analytics approach. Early attempts involved an AI Slack bot, followed by integrated semantic views. By 2026, they deployed Hex Threads, a custom data companion powered by Snowflake Cortex Agents.

This agentic era allows employees to interact with data conversationally, eliminating the need for deep SQL knowledge or understanding complex data structures. Within 90 days of launch, over 80% of Whatnot's employees were actively using the solution, with 17 departments reaching 100% utilization.

Teams now use conversational AI for strategic tasks like tracking international trends and building seller churn models, compressing analysis that once took weeks into minutes.

Democratizing Data, Controlling Costs

Giving teams freedom to query and spin up resources necessitated robust monitoring. Whatnot focused on fast, affordable, and readable observability.

Snowflake's updated telemetry engine, via Snowflake Trail, now offers 10x faster event ingestion, solving the slow logging bottleneck. This allows for real-time pipeline error detection, a significant improvement over the previous three-to-four-hour delays.

Furthermore, AI-assisted observability workflows allow users to create complex infrastructure alerts through natural language requests, abstracting away the need for intricate SQL coding.

Proactive Operations and 'Epistemic Hygiene'

Looking ahead, Whatnot aims to shift from reactive problem-solving to proactive prevention. Scaling AI analytics revealed organizational friction, highlighting the need for clear data modeling and metric definitions.

To ensure data trustworthiness in their decentralized AI systems, Whatnot enforces strict guidelines. Agents must use probabilistic language, differentiate observation from interpretation, and avoid declaring causation from correlation.

This combination of modular data stacks, real-time event logs, and conversational AI transforms the data infrastructure black box into a strategic advantage, enhancing both internal operations and the end-user experience.