Fintime The economy in context
Sections
Technology

AI Samaya Outperformed Wall Street Consensus in 40 Out of 50 Forecasts

In an experiment by Fortune and Samaya, the startup's model more accurately predicted Q3 metrics for 40 out of 50 companies than Bloomberg's consensus. The sample covered 10 industries.

19 Nov, 14:10Read in Russian
fortune.com

Out of 50 companies, Samaya's model more frequently provided more accurate assessments than analysts for the most critical metric—earnings per share or revenue. Forecasts were made at least one day before the release of financial reports and later compared with actual results.

The model performed better, in particular, for Palantir and Permian Resources. Analysts were more accurate in forecasts for Yum Brands, as well as in assessing Amazon's and Google's revenue shortfalls relative to expectations.

Samaya CEO Maitra Raghu stated that the consensus more often prevailed for the largest companies, for which many private research reports are available. The startup used publicly available data—financial reports, news, and interview transcripts. In the future, it may incorporate analyst notes, though Raghu fears this could increase the AI's tendency toward consensus.

The authors note that the experiment covered only one reporting season and a small sample. According to market researchers, analyst forecasts are influenced by reputational risks and pressure from investor relations departments. AI could transform this system but may also adopt its biases.