Spangle AI

Spangle AI — Two distinct layers. (1) First-party (the licensable asset): conversational shopping-session telemetry from Spangle's AI shopping agents deployed on retailer sites — natural-language product-intent queries, the recommendations the agent surfaced, the shopper's acceptance/rejection/ refinement of each suggestion, the resulting browse→add-to-cart→purchase path, and the AI-influenced vs direct-traffic attribution behind each order. (2) This page: a maintained secondary-data reference aggregating dated Adobe Digital Insights and Similarweb figures on AI-referred retail traffic (62% Y…

Spangle AI offers (Alternative, Sentiment, Reference) — Every shopping conversation yields a multi-turn transcript plus the full recommendation set and the downstream order outcome, so volume tracks conversation count — plausibly hundreds of thousands to a few million conversation transcripts, each carrying multiple product-preference judgements, if deployment is in the tens of retailers. Nothing published; this is the first figure to confirm with the vendor. The published research page itself is a small curated statistics set..

RLHF / preference-pair datasets for shopping and recommendation models from accept-reject-refine sequences; conversational-commerce training data (intent→product-grounded-response pairs with purchase ground truth); e-commerce personalisation and retrieval fine-tuning; agentic-commerce attribution and incrementality models separating AI-influenced from direct revenue; consumer demand and category-trend sensing from unprompted natural-language product intent.

The data is with 2 years of history.

Coverage spans US; Internet & Direct Marketing Retail, Application Software; alternative, sentiment, reference; equities.