GermanyConsumer goods2025 – ongoing

adidas

RAG chatbot for customer product reviews

adidas improved review-analysis efficiency by up to 30–40%, according to Databricks

WorkingVendor case studyVendor ↗

Problem

Review analysis remained largely manual. An existing chatbot returned generic answers and imposed high computing costs.

Approach

adidas embedded more than 2 million product reviews and indexed them in Databricks AI Search. Its retrieval-augmented generation pipeline combined relevant review excerpts with prompts and generated answers using models including Claude Haiku. The chatbot was designed to serve more than 50 decision-makers.

Results

up to 30–40%improvement in review-analysis efficiency reported by DatabricksVendor case study[1]

60%latency reduction reported by Databricks, from 15.5 seconds to 6 secondsVendor case study[1]

over 90%reduction in compute costs reported by DatabricksVendor case study[1]

Takeaway

The reported computing savings followed smaller retrieval contexts and changes to prompt design.

Sources

Driving product innovation through customer feedback
Databricks2025Vendor publication

Source summaries

[1]Databricks2025

Databricks described the review-retrieval architecture and operational chatbot. It reported efficiency gains of up to 30–40%, 60% lower latency and compute costs reduced by over 90%.

Full details are in the original sources.

Added 1 Nov 2026 · Updated 1 Nov 2026