Traditional broadcast television is still largely tracked using very small audience samples, and recent polling failures show you how dependable that is in the diffused media world we now inhabit. BARB, Nielsen and other TV pollsters are as guilty of pseudo science and influence by interested parties and their political polling colleagues, but no one is willing to call out the status quo which benefits all those involved in different ways.
To be fair, coming up with a more accurate measure would be challenging.
Online metrics are more accurate, but this can offer its own challenges, especially if you're dealing with anonymised viewers. A hundred views from a single IP address could be construed as a hundred visits by one user, for example, but all the viewers may be individuals in a single corporate subnet.
And then advertisers use their own metrics to ascertain playouts online, which may or may not correspond to those of the webcaster.
Netflix issues very few details of its viewing figures - only what they need when reporting to financial markets and regulators, but they do from time to time offer valuable insight into qualitative trends in the media delivery industry.
Their latest nugget is that their viewers are alternating between bing TV series viewing and watching movies.
We're living in a world where statistical or quantitative analysis is becoming more and more dubious, so it may be time to pay more attention to qualitative analysis.