audience attractor
A floor and a ceiling, not a curve
The hypothesis the playground exists to interrogate: personality-driven media audiences do not grow like a smooth line. They sit in bands. Once a streamer, talk-radio host, or podcaster reaches a certain viewership, habit and parasocial attachment hold the floor; niche capacity, identity lock-in, and format limits hold the ceiling. The attractor metaphor names that joint structure.
Viewership does not grow like a line. It moves between basins.
Core plus casual
The model separates a slow-moving core audience from a fast-moving casual audience. Core decays slowly through habit retention and grows by parasocial conversion of casuals. Casuals respond to platform exposure, cumulative advantage, saturation, and noise. The floor lives in the core. The ceiling lives in the platform and niche terms acting on the casual pool. A trust shock hits casuals first, a schedule collapse hits the core first.
How to read the landscape
The signature visualisation shows the potential well U(x) over log-viewers, with the empirical dwell histogram of the trajectory overlaid as lime fills. Wells are where structural pull is strongest; histogram bars are where the trajectory actually spent time. When the ball sits in a well and a bar sits over it, the basin reading is doing real work. When the bars are scattered and the ball is on a slope, the trajectory is wandering rather than orbiting.
Falsification, not confirmation
The model is built to be falsifiable in two directions. Drop habit, identity lock-in, and saturation toward zero and the trajectory should become a noisy walk with no dominant band: that is the wandering regime, and the attractor story does little work for that configuration. Slow-decay is the other falsifier: if even a creator with high habit and low discoverability fails to settle in a band, the floor is thinner than the metaphor claims.
Beyond Twitch
The dynamics are not platform-specific. Talk radio runs on commute routine and host familiarity. TV talk shows run on time-slot habit and guest pipelines. Podcasts run on feed subscription and parasocial routine. YouTube runs on recommendation memory and topic identity. Each medium adds its own ceiling and its own floor mechanism, but the two-pool structure and the basin language travel.
Model changelog
- ported the potential-landscape signature visualisation to a black-and-lime SVG with dwell histogram overlaid on the wells.
- ported the trajectory chart, drift curve, and band histogram to the black-and-lime palette via recharts.
- expanded the prototype five scenarios to six by adding a slow-decay scenario as a falsification target for the floor.
- replaced the prototype dwell-band readout with a four-tier regime classification (collapsed, wandering, stable basin, transitioned).
- added the standard scientific panel suite: sweep across each of twelve parameters, sensitivity tornado on final viewers, calibration table with reader-assigned expected finals, ten assumption entries.
- snapshot comparison: save a configuration, change parameters, see the dashed-orange ghost on both the landscape and the trajectory.
- narrative and reading generators tuned to the basin language (dominant band, dwell share, basin transitions, falsification).