You have hit upon a couple of the important limitations in podcasting metrics. It’s good to measure unique downloads by IP in order to avoid inflated stats, but there is the risk of under-measuring….for example in the case of a corporate firewall.

If the stats package is sophisticated enough, I suppose there are some things that could be done to mitigate this. You could watch for different OS and browser combos hitting the file and make some other guesses as to when it might be more than one person accessing the file.

When it comes to bittorrent, I think all bets are off. You save on bandwidth, but your stats are a mess.

That one point is interesting as there is likely some value that can be gleaned from being able to do stats on old entries and seeing what information is still useful to people over time.

Anyhow, I’ll admit that I’m new to all of this but find it very interesting. I was curious about the metrics that come from individual IP hits. Do web servers have a way to distinguish users behind a firewall that all share one external IP? I ovten wonder about this since you and 20 friends at work can each subscribe to a podcast individually and potentially be considered one hit.

On another note, I’m curious how technologies like torrents might affect the overall landscape of statistical research? If people begin subscribing to a podcast but your outbound bandwidth is minimized by having another subscriber feed the episode to the end recipient do all my metric tools go away?