Year: 2026
Type: preprint
Abstract: Whether automated bargaining benefits electronic commerce depends on the transaction route it changes, not only on the quality of its negotiated offers. A checkout-relative value model is developed to evaluate an optional seller-side bargaining agent against existing posted-price checkout. The model separates new-sale margin, discounts on otherwise feasible checkout purchases, unrecovered sales and participant interaction costs. It derives conditional cost and fallback limits for seller and buyer outcomes. Public product-price and negotiation inputs support a matched computational comparison of two rules and two frozen fitted Q iteration policies across 193 evaluation products. In the reference configuration, the full-state learner adds 0.0144 seller and 0.0039 buyer listing-price units per original opportunity, but a simpler learner provides higher combined value. Stubborn offers produce seller losses, while low valuations produce buyer losses after interaction costs. These results show how a seller-favoured bargaining policy can differ from a configuration that benefits both participants on average. The contribution is a reproducible model linking executable policy outcomes to the economic consequences of an optional transaction channel. Participation, customer values, fallback and effort costs remain stipulated; the findings are conditional computational evidence, not estimates of actual adoption, customer experience or realised commercial effects. Positive reference averages conceal product heterogeneity: only 82 of 193 product contexts yield positive means for both parties. Structural sensitivities further qualify deployment claims.
Source: Research Square
Authors: Bankatesh Choudhary, Meenakshi Sharma
Institution: Birla Institute of Technology, Mesra
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225109548