In this real-world case study, RavenPack shows how its AI platform and NLP engine dig into internal digital content of a discretionary hedge fund manager to generate differentiated investment insights and trading signals from emails and instant messages.
Each organization owns and accumulates massive amounts of digital content, which largely remains under analyzed and untapped. By systematically extracting, structuring and enriching the fund’s own content in real time, RavenPack demonstrates that there is incremental alpha to be captured internally, as opposed to using public content alone.
The case study was conducted on the internal content of a $1 billion European discretionary hedge fund, which focuses on the utilities, infrastructure and commodities sectors in broader Europe, and the analyzed content (broker research, desk commentary and internal research notes) consists of three years of internal data (2016-2019), comprising of hundreds of thousands of emails, attachments and Skype messages in over 1,000 different file formats.
They find that:
- Approximately 80% of identified stock-related events were detected in the firm’s internal content, while 20% originated from public news and social media.
- Positive sentiment signals derived from internal content provide strong long-only signals for up to several weeks, while value from public news decays faster.
- Factor risk analysis of sentiment portfolios shows stable P&L coming from idiosyncratic stock price moves, demonstrating persistent alpha generation from a traditional factor model perspective