Peloton published the stats of their purchase cohorts in their S-1 (and probably in subsequent reports, but I haven't read them). At least among those who maintain a $39/month Peloton subscription ("Connected Fitness Subscribers"), the frequency of use increases over time. From https://www.sec.gov/Archives/edgar/data/1639825/000119312519...:
> We have consistently seen workouts increase over time. On average, our Connected Fitness Subscribers completed 7.5, 8.4, and 11.5 workouts per month in fiscal 2017, 2018, and 2019, respectively. Usage drives value and loyalty, which is evidenced by our exceptional weighted-average 12-month Connected Fitness Subscriber retention rate of 95% across all fiscal year cohorts since fiscal 2016.
…
> Average Monthly Workouts per Connected Fitness Subscriber are higher for our most recent cohorts and engagement for each cohort has been consistent or improved over time.
S-1 page 63 has more. Also, at least as of the S-1, churn (called "Average Net Monthly Connected Fitness Churn") from those cohorts was also very low - roughly 0.5-0.9% per quarter or ~3%/year. That means Connected Fitness Subscribers includes the vast majority of those who purchased a bike.
In 2019 when the S-1 was filed, I put a few more details here: https://twitter.com/troyd/status/1167582830425075712