What challenges did PopChill face before AI implementation?

Improve personalization and search for pre-owned luxury ecommerce. Automate safety and policy screening for buyer/seller chat and identity verification. Reduce manual effort and speed up recommendation updates.

What was the impact of the AI solution?

Clicks on recommended items increased by 2.5X. One engineer spent two weeks to build the image search prototype.

Source

GCPEvidence: Medium50/100

PopChill: Vertex AI product recommendations, vector search, and Gemini moderation for pre-owned luxury ecommerce

Use case typeFraud detectionUpdated Jun 13, 2026

PopChill is a pre-owned luxury item ecommerce platform in Taiwan and Hong Kong. The company uses BigQuery, Vertex AI Search for commerce, Vertex AI Vector Search, Gemini 1.5 Flash, Text embeddings API, Multimodal embeddings API, and Vertex AI Vision. It built a data warehouse from Google Analytics and Google Ads data, refreshed recommendations every two hours, enabled image search, automated policy-violation screening, product categorization, and seller ID verification.

Organization: PopChill
Industry: Retail
Location: Taiwan
Published: June 2026

Reported outcomes

Impact: 2.5×

Other quantified impact

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Quantified impact: 2.5 x increase

Google Cloud Customer StoryJun 3, 2026Customer storyInferred claimMedium evidence strength

Clicks on recommended items increased by 2.5X.

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: PopChill
Provider: GCP
Maturity: Unknown
Linked source: Google Cloud Customer Story

No explicit deployment-stage evidence found.

Customer identity supportedSource describes one deploymentMaturity evidence evaluated

Primary read

Use case focus

Showing 3 of 4

1Personalization / Recommendation Engine
2Search and Discovery
3Content Moderation

Built a BigQuery warehouse ingesting analytics and ad data.
Used Vertex AI Search for commerce to refresh product recommendations from catalog and event data.
Used Vector Search with text and multimodal embeddings for keyword and image search.
Trained Gemini 1.5 Flash to flag risky chat messages, categorize products, and validate descriptions.
Used Vertex AI Vision to verify seller identity from ID images.

Technologies

BigQuery Vertex AI Search for Commerce Vertex AI Vertex AI Vector Search Gemini 1.5 Flash Text embeddings API Multimodal embeddings API Vertex AI Vision