Research & methods
· Field Note 0001 · Jun 2026 · 5 min · KWL Research
The paper that started Cadence. And the five tests that closed the question.
We did not invent this idea. We found robust science, tested it ourselves, learned where it breaks, and built a product that stays inside the evidence.
From a finance result to a brand read
Peer-reviewed findingMusic sentiment predicts stock returns, across 40 countries
Our extensionA hypothesis: mood shapes consumer behaviour, not only markets
CadenceOne read from six signal familiesMusicSearchNewsStreamingWikipediaMacro
A behavioural market readHow a market feels, and how a brand should speak
The finding
Everyone senses that what a country plays says something about how it feels. The useful questions are whether it is measurable, and whether it connects to anything that matters. There is a peer-reviewed answer to both.
Published in the Journal of Financial Economics by Alex Edmans and colleagues, peer-reviewed research rather than a trend report: the valence of the music a country streams predicts that country’s stock returns, across forty countries.
The study
40countries
2022published
JFEJournal of Financial Economics
Peer reviewedindependent academic work
Why it is powerful
The signal is revealed, not stated.
Nobody is asked how they feel. The measure is drawn from what people actually played, in their millions, day after day. Revealed preference is the economist’s gold standard because it is hard to fake.
Behaviour records what people did, not what they said.
What Cadence does with it
We take that revealed-mood signal and extend it one step the paper does not: downstream, from financial markets to the consumer behaviour brands care about. Read sentiment from observed behaviour rather than stated intention.
Music is the anchor, the signal with the published validation behind it. Around it we triangulate search, streaming, attention, news tone and macro, each normalised so a number in Spain reads like a number in Sweden. The music tells you how a market feels. The rest tell you what it is doing about it.
The architecture
Six signal familiesMusicSearchStreamingNewsWikipediaMacro
Normalisationposition-weighted, year-on-year, de-meaned across markets
Latent behavioural dimensionsmood, attention, trade-down, sovereignty, tone
Country archetypeone of the Four Europes
Brand guidancehow to speak to the market this quarter
Why six signals
No single signal explains a market. Music captures mood. Search captures intent. News captures narrative. Streaming captures attention. Wikipedia captures curiosity. Macro captures economic context. Together they produce a more stable behavioural read than any one source alone.
Where we stop
We tested forecasting four times and could not support it.
Taking the science seriously cuts against us, too. The literature backs music as a contemporaneous mood proxy. It does not license a forecast. So we use the signal for exactly what the evidence backs, and no further.
The evidence supports
The evidence does not support
✓Music reflects contemporaneous mood
✕Music predicts product demand
✓Revealed behaviour
✕Stated intention from surveys
✓Mood as an input to a read
✕Mood as a sales forecast
Why this is the position
The strongest AI products are built by making only the claims the evidence can support.
That is the posture behind Cadence: grounded in published research, extended with discipline, and explicit about where the evidence ends.
References
Edmans, A., Fernández-Pérez, A., Garel, A., & Indriawan, I. (2022). Music sentiment and stock returns around the world. Journal of Financial Economics, 145(2).