Overview
Being Good was a long-running attempt to build a different way of describing and navigating music. Instead of treating genre, artist similarity or metadata as the main structure, the system explored how rhythm, texture and pitch could describe perceptual relationships between music and mood.
I started working on it in 2013 and stayed involved through many versions of the project. It was never one app that stayed the same for ten years; the research methods, interfaces and technology changed repeatedly as we learned what worked.
Role & Scope
I was a co-inventor and worked where the music research and the product met: perceptual evaluation, interaction and UX, prototype direction, music analysis, and translating the underlying system into something a person could actually use.
A lot of the work was hands-on. I listened to and evaluated music, compared model output, worked through product flows, tested prototypes, and helped shape how Rhythm, Texture and Pitch became a navigable system rather than three abstract values.
What I Built
2013 — Initial concept and perceptual model
2016–17 — Acoustic features + classification research
2020–21 — Multimood + 1,123 human evaluations
2022 — Mood Web Service
2023 — Spotify-connected prototypes
2018 / 2024 — Issued U.S. patents
Music information retrieval · RTP classification · Acoustic feature analysis · Machine learning · Spotify API · .NET/WPF · Web services
One of the hardest parts was turning subjective listening into data that could actually be tested. In the Multimood phase I manually evaluated 1,123 musical segments, assigning Rhythm, Texture and Pitch values in half-point increments.
Roughly 30% became training data. The remaining 771 held-out segments were then used to evaluate how closely computed RTP values and mood classifications matched my perception.
Later versions brought the research into working software: music could be analyzed, classified by mood, filtered by mood and intensity, assembled into playlists, and connected back to Spotify.
The work became a continuing patent family. I am named as an inventor on issued U.S. patents including US 9,875,304 B2, Music Selection and Organization Using Audio Fingerprints (2018), and US 11,899,713 B2, Music Streaming, Playlist Creation and Streaming Architecture (2024).
Being Good did not end as one clean consumer product. What came out of it was a body of music-perception research, a manually labeled dataset, multiple working software systems, several product directions and patented technology.
It is also the project that most changed how I think about design: sometimes drawing the interface is the easy part. The harder problem is deciding what the system means in the first place.
