AI

Learning to bid and bundle: a posterior-sampling agent for online retail

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 03/11/2026, 09:03:24 UTC.

Briefings

An online retailer launched a new agent designed to learn bidding strategies for perishable products by observing its own decisions. The agent draws a market from a posterior distribution, posts a bid and discount based on that market's exact solution, and avoids learning that bidding pays in weak markets. In a simulated florist, the agent achieved the lowest regret over 500 seasons compared to drawing once per season, though it escaped after only three seasons on average.

Key points

• The agent draws a market from an exact posterior distribution and posts a bid and discount based on that market's exactly solved season plan, ensuring every action is one a fully informed seller could take.
• In a simulated online florist, the agent achieved the lowest regret averaged over markets compared to drawing once per season, though it escaped after only three seasons on average.

Context

This research addresses the challenge of bidding for sponsored-search clicks on perishable products where the seller learns neither how clicks respond to its bid nor how strong and dispersed its market is. The agent learns both over repeated seasons from data its own decisions censor, aiming to minimize regret in simulated online florist settings.

Limitations

The evidence is limited to a simulated online florist setting and does not include real-world data. The agent escaped after about three seasons on average, and the study does not account for the cost of frequent waiting or the impact of discount frequency on market underrating.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 04/11/2026, 02:23:15 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeieskfj3yhiygptapcgq45rlxxlsw6jmlvjc3ajckftpbza6mlcb6u