WITHDRAWN — AI ROI Analysis: Evidence from 200 B2B Deployments (Dataset v1.3)

Engineering notes

Engineering notes will be added by the aipentium editorial team.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为大语言模型、生成式AI、ChatGPT相关技术、计算机视觉、深度学习等高价值论文补充中文说明。

Original abstract

WITHDRAWN — This record is withdrawn and superseded. The reference dataset is: AI ROI Dataset — 200 B2B AI deployments in France(2022-2025), DOI 10.5281/zenodo.17795133. Reason for withdrawal. The metadata of this record contained methodologicalstatements that could not be substantiated against the underlying file: - "audited methodology" and "financial audits": no independent audit was carried out on this dataset.- "survivorship bias correction": the published median was in fact computed on successful deployments only, which introduces rather than corrects a survivorship bias.- "outlier removal (top 5% Retail outliers driving the mean to +347%)": the Retail segment of the reference dataset has a mean of 147.7% and a maximum of 294.0%. No +347% mean is reproducible.- "Technologies used (ChatGPT 67%, Claude 18%, custom ML 8%, other 7%)": the reference dataset contains no technology variable. These figures are not supported by any published file.- "Failure rate: 27%": the reference dataset reports 17.5% (35 of 200).- "88.5% adoption in successful projects": 88.5% is the rate across the full cohort, not across successful deployments (89.7%). The reference dataset publishes both calculation scopes, declares the natureof its central metric (a gain-to-investment ratio, not a net ROI), and shipsa dependency-free reproduction script. Users who downloaded this record are advised to use 10.5281/zenodo.17795133.

Links and sources

Comments