It's also a well-known psychological phenomenon that people will curtail usage if they know their usage will cost them more -- even if the amount is trivial. The last thing Internet-based companies like Netflix or Amazon want is for people to use the Internet less. Current metering systems really only "catch" the most egregious offenders (less than 1% of total customers). Most ISPs have a "business class" plan you can purchase for more that truly is unlimited, if you find yourself exceeding the caps on a residential account on a regular basis.
Those are effectively a source of hidden profit margins--NFL, Major League Baseball, et. al. get them instead of the telecom operators, but when looking at how much prices could drop, you have to consider that.
The huge marketing budgets don't go away if you suddenly split the market between a dozen or so providers. In fact, overall marketing spend across the industry would probably go up substantially because there would be more competition for limited consumer mindshare.
All I'm saying is that the marketing costs wouldn't go away even if the big cable companies do: they would just be replaced by a dozen smaller companies doing the marketing, billing and connectivity. These smaller ISPs would just be taking each others' customers rather than customers jumping ship to DSL or some other last-mile tech.
Why? It seems to me that metered internet could work just fine, if the prices are at least somewhat connected with reality (meaning cents per gigabyte at most, with a base cost not more than about $20/month). The form of metering we currently have in the US is usually punitive overage charges for data in excess of some unrealistically low monthly quota, and is applied on top of $60+/month subscription fees. That's obviously bad for internet content providers (since their whole point is to deter further usage), but I think reasonable prices are possible.