One problem he doesn't account for is that the community of users is often diverse and therefore ui complexity gets driven by corner cases. Often teams will start out with a visionary who holds the line for a while but gradually the pressure to add another bell or whistle increases until it is impossible to resist. For example coversion rates; if a feature is shown to drive up conversions by a small amount it will tend to get added, it doesn't take many of these discoveries to ruin an interface.
In the early stages of a project everyone is a beginner. Success is success because people become experts and experts are experts because they know how to do and want to do sophisticated things.
The example of the toilet is perfect. Good interfaces are designed to meet the needs of the user community. An expat in Japan is not a member of the community for some sense of "member" and "community". The solution for the outlying case is that the outsider asks for help. This transfers the expertise.
This touches on a problem our current product has. We're on the second iteration where the first iteration had almost no traction yet the design and features coming from the product side assume so much implicit knowledge of the business that there's no way a complete newbie can look at it and feel confident that they know what is going on.
Because we have no traction and thus no metrics we're in a vicious "I think it should work like this so let's do it that way because you can't prove people won't like it" cycle.
Sounds like you should check out "Don't make me think" by Steve Krug! It's a quick read and contains very simple effective advice on how to make your UI usable. He outlines ways to do very cheap user testing to add objectivity to the decisions you're describing.
but I think the author's point holds up under your criticism.
he's right that 99% of the time when you are looking at your account, you want the current trip. Take the 90% of the dashboard junk dedicated to "you might need this, so we're showing you just in case", and wrap it in a single, exposable "junk drawer" interface.
Or at the very least, call out the main content with styling, and make the junk more muted.
He calls this "dashboard ambush", but it's just as accurately characterized as a signal:noise problem. All the extraneous functionality is noise to 99% of use cases.
See the Product Hunt example. It's not that dashboards are bad, they should immediately be obvious what the user should do. The eBay example is good too - just wtf is going on there?