I would trust an unaffiliated speed test app or website server-side analysis much, much more than any count of signal-bars.
My point is that people are putting a lot of stock in signal-bars, assuming if there are 4-5 bars local service is good (and conversely that drops in signal-bar count mean bad service). That's not my experience in SF at all; the correlation used to exist, but in the last year, it broke. I can have usable calls at 0-2 bars; I can be unable to initiate or hold a call at 4-5 bars. It's some other capacity issue that's not measured with signal-bars that's AT&Ts giant problem.
My point is that people are putting a lot of stock in signal-bars, assuming if there are 4-5 bars local service is good (and conversely that drops in signal-bar count mean bad service). That's not my experience in SF at all; the correlation used to exist, but in the last year, it broke. I can have usable calls at 0-2 bars; I can be unable to initiate or hold a call at 4-5 bars. It's some other capacity issue that's not measured with signal-bars that's AT&Ts giant problem.