Senstream
Neuromarketing Solution
Topics
Connected Devices, AI, Neuromarketing
My Role
Interaction Design, Product Manager, Data Collection
Senstream's groundbreaking Researcher Ring harnesses biometric data to revolutionize media analysis, achieving 93%+ accuracy in detecting audience reactions and unlocking new frontiers in neuromarketing.
Background
Senstream is a pioneering startup that has developed the first mobile-connected ring designed to produce lab-grade human biometric data. The flagship product, the Researcher Ring, features a patented capability for a ring to accurately read electrodermal activity (EDA), which is considered the gold standard for assessing the activation of a subject's nervous system.
The potential applications of the Researcher Ring are vast, with the ability to analyze an individual's unconscious emotional and mental states. This case study explores one such application: media analysis.
Objective
By leveraging EDA to measure moment-by-moment reactions to media, the Senstream system can identify key moments that elicit involuntary physiological responses from viewers. This field of study, often referred to as Neuromarketing, offers valuable insights into how audiences emotionally and cognitively engage with content. Utilizing the Researcher Ring in conjunction with a specially trained machine learning model, Senstream aims to pinpoint moments in media that generate significant audience reactivity.
Key Learnings
Training Data Creation
Developing an effective training dataset required manually analyzing hundreds of data recordings presented as line charts. Specific patterns and "shapes" in the charts were manually identified to form a "positive" dataset that indicates significant reactions.
​
Data Interpretation Challenges
Real-time detection of activations proved challenging, as models required up to 40 seconds of data to establish a baseline before detecting spikes that signify notable reactions. A proprietary combination of 40-second data windows, supplemented by smaller 5-second segments, enhanced the precision of event identification.
​
Accounting for Response Delays
Individual physiological response times to media stimuli vary, introducing complexities in aligning biometric data with media timelines. Training datasets were specifically designed to account for these delays, ensuring more accurate mapping of reactions to stimuli.
Outcomes
Model Performance
The machine learning models developed for the Senstream system achieved accuracy levels in the lower-mid 90% range, using a standard 80/20 training-test split.
Field Testing
The Researcher Ring was successfully field-tested in a media reaction study conducted by Dr. Paul Bolls at the University of Washington, Media Studies. In this study, 30 participants, comprised of both Democrats and Republicans, wore the rings while watching the Harris/Trump debate, providing valuable data on political media engagement.
Beta Version of Media Analysis
The beta version of the mobile app allows the user to wear the research ring and watch content uploaded to the app.​ After the content is complete, the subject's data is analyzed and key moments of activation are overlayed on top of a playback of the tested media.
​
This application is currently being tested with various partners to collect feedback for upcoming iterations.
