I have to confess, I haven't ever used Bayesian linear regression, but I will point out one major problem with your analysis. Specifically, that you assume the value of a reading minute (or a read, or a view) is constant.

It turns out that the value of a member's reading time isn't constant. According to my analysis (see https://medium.com/financial-strategy/cracking-the-code-for-mediums-current-partner-payment-calculation-52809b12b1b0), the rising edge of the distribution behaves like a sum of a constant (~$0.023/minute or $1.38/hour) and an exponent function of the cumulative reading time.

For the article analyzed there, the value of a minute of reading time peaked above $0.30 ($18/hour). The part I haven't cracked yet is what happens after the bulk of reading time is done. While over time, Medium reduces the value of a reading minute from the exponentially (literally) high value it can reach, back down to ~$0.023, it doesn't happen quickly.

As an example of how the post-peak behavior is hard to figure out, I looked at another of my articles, https://themakingofamillionaire.com/protect-your-assets-best-get-out-now-or-keep-your-money-in-stocks-144626eabba0. This article accumulated over 110 hours of reading time, and it too showed an increasing value of a reading minute. Curiously, after the daily reading time peak was over and it dropped to a tail of about 5 minutes/day, it continued earning far higher daily amounts than seems plausible. Specifically, even on some days with zero reading time, it earned between $1 and $5!