Kirk,
Shawn is wrong when he said "Unique downloads is the best way to measure overall audience size". The unique IP information is for auditing purposes. It helps us when discussing measurement with an ad buyer or podcaster, it is easy to explain the unique IP total first (easily reproducible). When we explain that the delta between downloads and unique factors in large networks such as corporate networks, wifi hot spots, phone networks, university dorms, etc... it quickly makes sense.
When we see multiple requests from the same IP using the same user agent it goes into a special process in our system to identify the uniqueness of the download/play. In most cases (there are exceptions, e.g. wifi hotspot) we determine that there is only one download. We use information such as the byte range request data to determine which portion of the file is requested. More specifically we look at user agents which we see a lot of duplicate IPs from such as Stitcher and the podcast iOS app (mobile apps). These applications will break a download into smaller chunks by means of "byte range requests". These byte range requests when viewed by applications like AWStats appear as downloads. Sometimes they get this right and report them as "hits". In truth they are partial downloads that are then stitched together by the application.
We had a podcaster who had 100 unique downloads for an episode who reported another 900 downloads coming from one IP address in his AWStats all in the same day. We know that in all practical terms there is no way 90% (900 of 1,000 if we used his math) of his audience came from one IP address.
It is hard to gauge "subscribers" but we have some basic tips at the bottom of this page how to estimate them:
http://create.blubrry.com/manual/statistics-analytics-measuring-performance/ Essentially the first 3 days of downloads are, for the most part, coming from users subscribed to your podcast.