Streaming Media and Measurement
In the “connected” world of Internet delivered video, we have a number of advantages over the broadcast world in terms of audience measurement. First, and most obviously, we are connected to the same network - the Internet. Note that we often talk about being connected “to” the audience, but in actuality this may not be the case. While the users are connected to an IP network (their ISP's last mile access network), the content service provider is (with the exception of IPTV) invariably connected to a different IP network. There are usually many other networks spanning the delivery chain between the users, their ISP, the core Internet backbones, and the content service provider (CSP)'s publishing origins. Some forms of data such as those transferred using HTTP and TCP may transit these networks reliably and consistently, and these have become very successful in the delivery of ad hoc on-demand data - as we will explore in Chapter 10 when looking at the notionally more scalable multicast routing and the User Datagram Protocol (UDP) delivery methods. However, we often do not tie reliable, accountable delivery inherently to the delivery model. The reliability and accountability is often layered in as an application layer function that examines the network activity and then makes an “out-ofband” return path (with the logging, and sometimes the user control data shares this return path). So everything from metadata, recommendation and selection to bitrate, program dwell time, etc., can be captured, limited only by the programming ingenuity of the end user application design team.
The truth is that it is very difficult to give a meaningful interpretation of such data sets. Perspective is everything when interpreting sampled types of information. Fortunately, the most interesting data are top-line macroeconomic indicators such as overall traffic and overall audience size. A graph showing variance in such key performance indicators over a period of time is universally more informative than detailed granular reference to particular volumes. Trend is more important than specifics.