Driving Trust with Auditable Transcripts
SVP of Technology & Product Strategy at FactSet on driving trust through auditability
This product turned earnings calls into a trading signal, not just a research document. Instead of helping an analyst read faster, it tried to detect whether management language, tone, and changes in wording pointed to an unusually large stock move after the call. That put FactSet in the business of selling probabilistic signals on top of its transcript feed, which is a much higher value layer than raw content delivery alone.
-
The workflow was concrete. FactSet already collected and distributed earnings call transcripts through CallStreet and related transcript products. High Impact Transcripts added an ML layer that scored the text for likely outsized moves, so a portfolio manager could triage which calls needed immediate attention.
-
This sat inside a broader pre-LLM signals business. The same interview places it alongside models for activism vulnerability, secondary offerings, bond issuance, and S&P 500 index changes. That suggests the product was sold as one more event prediction tool inside a wider quant and research workflow.
-
The market has since shifted from prediction first to explanation first. FactSet now emphasizes Transcript Assistant, which lets users query transcripts, pull guidance changes, summarize Q&A, and compare themes across calls. The core asset stayed the same, but the user interface moved from a black box score to auditable transcript analysis.
The next step is not a return to standalone text based alpha products, but embedding transcript intelligence directly into the research workstation. The winning products will combine signal generation, chat, summaries, and traceable evidence, so analysts can move from alert to source text to investment decision in one workflow.