Investment firms today face a growing challenge with alternative data: how to turn an ever-expanding universe of unstructured content into something that can be analysed systematically.
Social media, news articles, disclosure documents and, increasingly, podcasts can all contain information relevant to investment decisions. But before that material can be useful to quantitative or discretionary investors, it has to be filtered, classified, attributed and converted into signals that can fit into existing research and trading workflows.That is the problem Bridgewise is looking to address following its acquisition of Context Analytics earlier this year, combining the latter’s alternative-data and sentiment engine capabilities with Bridgewise’s broader investment AI solutions.
“Our mission with Context Analytics is to convert unstructured data sources into structured data,” Bridgewise CEO Gaby Diamant tells Market & Alt Data Insight. “For example, if we have corporate filing documents that are several hundred pages long, we can convert that into structured, machine-readable data, enabling institutional investors to extract core signals quickly, through our predictive sentiment analysis metrics. We also have separate feeds for other alternative sources like news, social media, podcasts, etc .”
From conversation to signal
The social media sentiment offering draws on sources including a direct partnership with X, as well as StockTwits and news content, converting unstructured discussions online into quantitative signals that can be consumed directly by investment systems or through user-facing applications.
The company’s social sentiment feed monitors more than 7,850 global equities and has tracked more than 14 years of historical data for backtesting and research. The output is designed around several layers: quantitative sentiment factors, analysis of activity and source attribution, and AI-generated summaries explaining what may be driving the numbers.
Diamant points out that different users want different things from the same underlying data. A quantitative fund may simply want to add a signal layer into an existing model, while a portfolio manager may need to understand what caused a stock to move and where the underlying discussion originated.
“We’re able to take all of that unstructured data from social media and news, and extract a quantitative sentiment score based on a number of factors,” he says. “Within our model, we’re also able to filter out the noise, the bots and the material that we don’t want to be basing signals on, so that we can produce a much cleaner signal. Alongside the score, we can also provide textual analysis and context explaining what might be behind the sentiment shift.”
That filtering process highlights a broader shift in alternative data. Data scarcity is no longer the issue. Now it’s all about signal engineering: separating useful information from noise at sufficient scale, speed and consistency for institutional use.
Listening to longer-form conviction
Bridgewise is also pushing sentiment analysis into an area that has historically been much harder to process systematically: podcasts.
Where social media can provide rapid reactions to events, longer-form audio potentially offers something different. Interviews and discussions lasting 30 or 40 minutes can reveal developing narratives, changes in conviction and emerging themes that may not be evident from shorter-form content.
The company’s Podcast Sentiment Analysis Feed converts that material into structured data through a multi-step funnel. The feed transcribes episodes, identifies the companies mentioned, classifies the risks and opportunities via sentiment analysis, then identifies the securities’ relevancy to the episode to produce the final output of structured JSON and an AI summary, per company mention. Bridgewise says the underlying universe covers more than 4.5 million podcasts globally, with more than 51,000 episodes tracked across financial and business media, more than 6,800 securities identified and over 1.4 million summary records generated to date.
That opens up a potentially significant new area for alternative data teams. Content that might previously have required an analyst to spend considerable time listening, taking notes and manually identifying relevant references can increasingly be transformed into datasets that can be screened, queried, scored and incorporated into research workflows.
The use cases are also different from those associated with fast-moving social feeds. Bridgewise positions podcast data in areas such as trend analysis, longer-horizon investment strategies and reputational risk monitoring, where sustained discussion may matter more than immediate reaction.
Connecting the signal to the workflow
The acquisition of Context Analytics also points to a broader convergence between alternative-data processing and the applications through which investors consume intelligence.
Diamant cites the market reaction to Moderna’s announcement of positive Phase III results for its personalised mRNA cancer treatment, developed with Merck, as an example. The treatment significantly delayed the recurrence or spread of melanoma compared with Keytruda alone, sending Moderna’s shares sharply higher. Context Analytics’ systems could identify changes in news and social sentiment, while the wider Bridgewise platform could then surface the signal, provide explanatory context and allow users to interrogate the information through its finance-specific AI assistant Bridget™.
“The interesting part for us is bringing those capabilities together,” he says. “You can get an alert that something is happening, see the social sentiment score and what is being discussed online, and then ask further questions to understand why it is moving and what the context is. So rather than having the signal sitting on its own, you can move from the alert to the underlying sentiment and then into deeper analysis.”
That progression – from raw content, to structured data, to quantitative signal, to contextual explanation – may ultimately prove more significant than any single new dataset.
For alternative-data users, the universe of potentially investable information is continuing to expand. As audio, social discussion and other forms of unstructured content become increasingly machine-readable, the competitive question is shifting from what information firms can acquire to how effectively they can convert it into decisions.
Bridgewise is a gold sponsor of the A-Team Group/Eagle Alpha Alternative Data Conference in New York in September.
Subscribe to our newsletter
