Could the way people feel provide an early indication of changes in consumer behaviour and financial markets? That’s the proposition being explored by The Real Feel, a US-based data company collecting daily information on Americans’ emotional states to develop a new source of alternative data for investment research.
Founded by former Reuters executive Lisa Hu, the company originally targeted marketers but has since attracted interest from hedge funds and macroeconomic researchers.
“I always thought there was a missing qualitative data layer around how people are feeling in a given moment,” she tells Market and Alt Data Insight. “I wasn’t even targeting the financial world; I was targeting the marketing world. But the fact that hedge funds and macro researchers at banks were coming to me was quite fascinating. I realised it’s because we’re a whole new alternative dataset for them. We’re collecting human emotion, and there are hypotheses around the correlation between human emotion and mood, market movements and economic outlooks.”
Measuring the National Mood
The Real Feel works with survey provider Pure Spectrum to collect approximately 1,500 responses weekly. Respondents rate their feelings from one to ten, select an emoji describing their mood and explain their feelings in free-form text.
The dataset combines numerical scores with qualitative responses and demographic information covering age, gender, race, income and occupation. Unlike conventional consumer confidence surveys, which focus on economic expectations, The Real Feel captures broader emotional states, potentially identifying changes that influence subsequent economic decisions.Demographic segmentation allows researchers to examine differences between population groups, although Hu acknowledges some over-representation of retired and unemployed respondents, requiring additional weighting.
From Mood to Market Signals
The Real Feel has undertaken exploratory work with Wolfe Research, incorporating emotional data alongside other inputs into equity forecasting models.
“We partnered with Wolfe Research last year, and they took our data as one input among others, fed it into their models and found that emotion-based data could be a predictor in forecasting the S&P 500, down to sector and company level,” says Hu. “That was just one year’s worth of data, and of course there needs to be more analysis behind it, but that’s one interesting use case.”
Other potential applications include consumer credit risk, where financial anxiety among lower-income groups could signal emerging pressures, and discretionary spending, where mood changes among higher earners might offer clues about future consumption.
The initial findings suggest potential value in combining emotional data with existing forecasting models, with further research needed to establish its contribution to predictive accuracy.
Extracting Signals with AI
The Real Feel uses the Hex analytics platform alongside large language models, including Anthropic’s Claude, to analyse free-form responses and identify recurring themes.
One application involves extracting expressions of fear across demographic groups, distinguishing concerns about employment, household finances and other uncertainties. This could support a consumer-focused fear indicator, offering a different perspective from the financial-market volatility measured by the VIX.
“We come up with scenarios around how we want to use AI. We might say, let’s explore mood versus consumer spending, or let’s explore mood across different income groups, because we’ve seen differences between those making more and less,” explains Hu. “But there have been scenarios where AI has told me what might be a good application. The consumer VIX example came from Claude. I didn’t think of that. It sometimes toggles between being a great analyst who executes and an adviser.”
Although The Real Feel offers an API for querying mood scores and requesting AI-generated summaries, Hu says many financial clients prefer the raw historical dataset for proprietary analysis and backtesting.
The company offers testing arrangements, pilots and ongoing feeds, and plans to enhance its API to provide more direct access to raw observations and derived indicators.
Establishing Predictive Value
The growing dataset is providing new opportunities to explore relationships between emotional responses, economic behaviour and financial markets, with further validation expected as the historical record expands.
“Having two years of data, I’d say we’re seeing real potential,” says Hu. “Measuring how Americans react to events is relatively accurate. Even during major storms, you can see a mood dip in certain regions. With use cases like gauging consumer fear or predicting the S&P 500, the early results are promising. Time is on our side as we gather more data to validate these signals. For funds, that’s the opportunity: getting in early on a dataset that compounds over time.”
For investors, the potential lies in identifying patterns in human emotion that offer fresh insights into consumer behaviour, economic conditions and market movements, adding another dimension to established sources of financial and alternative data.
Subscribe to our newsletter
