Threading the needle for fashion resale success

Overview

ThredUp unified its data on the Databricks Data + AI Platform, using ML models for personalization, pricing, and inventory flow, and adopted Unity Catalog and the AI Playground for LLM experimentation, cutting ML model training from weeks to days and reducing new analyst onboarding from two weeks to four days.

The challenge

Onboarding new data analysts and scientists took up to two weeks, and even then they wouldn't start delivering outputs until after two months; the fragmented, siloed data platform made it hard to generate actionable insights quickly, and training machine learning models could take days due to resource constraints.

The solution

ThredUp adopted the Databricks Data + AI Platform, starting with Databricks Notebooks, then integrating Delta Lake for ACID transactions and schema enforcement, Unity Catalog for governance and democratized access, and a serverless architecture to dynamically scale ML model training, plus the AI Playground for experimenting with LLMs and generative AI.

Reported business value

Onboarding new analysts and data scientists dropped from two weeks to four days, a 71% decrease; new engineers now produce MVPs in as little as two weeks compared with two months previously; and ThredUp saves approximately half a million dollars annually by avoiding additional hires.