The Data Scientists Loved

We started with streams of biometric signals — EDA, heart rate, HRV, ECG — flowing across clean, precise line charts. To researchers, they were rich with meaning. To marketers, they were unreadable.

The data was valid. But it didn’t explain itself. The signals existed without a story.

Emotion Isn’t Obvious

We expected emotion to look dramatic — a clear spike, a sharp rise.

Instead, it appeared in many forms. Sometimes gradual. Sometimes clustered. Sometimes subtle. The same human reaction could produce completely different patterns.

Without context, arousal was just movement on a chart.

Creating Structure

So we went manual. Thousands of moments were reviewed and labeled, grouping different-looking patterns that represented the same underlying reaction.

This created a dataset that could be used to train a machine learning model to evaluate live data. A 80/20 split of the training data resulted in 93% accuracy.

From Signals to Insight

Now with an ML model, we could apply that intelligence to any media and visualize the audience's reaction while they wore the Senstream ring.

This prototype has allowed us to measure emotional reactions across a range of media

Film Trailers

We could map emotional lift scene by scene, revealing which moments truly landed.

Sporting Events

We could track sustained intensity, pinpointing the seconds that drove peak engagement.

I started to explore different concepts around what might analysis look like with the aid of GenAI and running larger studies where we could view the data around individuals and the group as a whole.

Below is an interactive prototype built using Lovable.