Immersion Blog

Sweating the Data?

Written by Paul J. Zak | Aug 31, 2020 12:00:00 PM

I love to sweat.  When working out.

In my academic lab, we measure arousal responses using small changes in electrical activity from the fingers as people sweat.  So, why doesn't Immersion measure sweat responses since we are "agnostic and promiscuous" when capturing neural signals?

News flash: electrodermal activity was part of the early models we developed to predict human actions. These models had a lot of redundancy built into them to ensure nothing was missed.  We would spend months cleaning and analyzing millisecond frequency data in order to build the most exquisite statistical models. This is what academics do--show off how smart they are to other academics.  Even Immersion's first minimum viable product that pulled data from a long-gone wearable included electrodermal responses. It was there, so we wanted to measure it.  The increase in predictive power from this data stream: a very respectable zero.   The arousal response from sweat is already captured by other neural signals in the key Immersion algorithms.

In fact, the sweat story is worse than no prediction.  The quality of electrodermal data from off-the-shelf wearables was awful when compared to medical grade devices.  The "ground truth" at Immersion has always been validating the data we get from smart watches by comparing it to data from devices that are used in your doctor's office.    There is an old saying in computer science: Garbage in, garbage out.  Over time, we found that using low quality electrodermal data from commercial wearables reduced the predictive accuracy of our algorithms.  So, we slimmed down and dropped the sweat.

Indeed, the full complement of Immersion algorithms have evolved substantially over the years to increase accuracy when predicting market outcomes. Today, our predictions are consistently better than 80%; for example, immersion predicts hit songs six months in advance of their release with 83% accuracy.   A very interesting additional finding is that people also enjoy immersive experiences.  Immersion both predicts what people will do in markets and also shows our clients what makes people happy.  It's a win-win for customers and companies.

So, don't sweat the usefulness of sweat.   Just because you can measure it doesn't mean you should. If you need to predict what your content will do, your best choice is to measure immersion.