Re: [videoblogging] Approximating subscribers

Sat, 12 Nov 2005 08:43:43 -0800

David, you are still missing some crucial background info regarding the reports that you are generating. While its true, everyone has a unique IP at a certain time and those IPs can rotate, its a very regular case to have the ISP "REPORT" an entire range of IPs as a single IP. For instance, out of over 100,000 page views in one day, 3% or over 3000 people of our audience, are reported to come from a single IP. There are several ISP that do this. Our biggest counts come in at 3%, 2% and many 1%'s.

Also, for example, one day of traffic that comes to mind involved an interview with Hoder from Iran. On that day, the greatest amount of traffic came from a single proxy IP that was likely being used by Iranians. Everyone was shown to come from a single IP.

If it were just a little bit off base, it wouldn't seem like that big of deal. But I can see how your numbers would easily become skewed by huge percentages - so much so that its not even a reasonable estimate.

Thus, without the help of warrants and the FBI, or a full time mathematician, I have not yet heard of any way to accurately describe how many unique individuals are visiting your website.

On Nov 12, 2005, at 3:10 AM, David Meade wrote:

I never said I was going soly by IP address. 

If one viewer pings my rss page 1 time or 137,256,748 times a day ... they still count as 1 subscriber.  I'm not concerned with how many times they are pinging or how many times they are downloading, only that this identifiable user is still polling for updates ... still subscribed.

As mentioned above I have a a whole database table full of rules to give some logic to the system.  IP is not the sole method of determination of "user" ... it's simply an important part of the formula.  Believe me, I'm not counting people more than once.  If anything I'm being overly restrictive.

IP address alone is indeed too obscure ... but 1000 different people from the samp IP is unlikely since most of these estimates work off of a 24 hour time period.  DHCP leases (the ones that acutally change and would allow for different people on the same IP) dont usually change too many times (if at all) within 24 hours - which is the time span of records evaluated to estimate subscribers in my system.

Worst case scenerio, is that you have a segment of viewers who are all behind the same NAT'd address in a corporate environment ... which would all apear as the same IP address.  Even then, it would take an UNBELIEVABLY popular feed to risk getting 1000 people in one day from behind the same NAT'd address.  RB may have this issue... I dont. :-P  But even if I did ... I dont rely soley on IP address.

EVEN if you do have 1000 or 2000 or 40000 people behind the same IP address ... how many times they each download the video isn't speaking to how many subscribers there are.  You could have 99% of them checking constantly for updates and not downloading them as often as made available.  Subscribers != downloaders.  Subscribers = people who are continueing to see if a download is avaialable.

- Dave



On 11/11/05, andrew michael baron <[EMAIL PROTECTED]> wrote:
But that doesnt work when go by IP addresses because the unique IP address count is too obscure.

You may be pinging my xml page 12 times a day and only getting one video. And 1000 different people may be coming from the same IP.


SPONSORED LINKS
Individual Fireant Use

YAHOO! GROUPS LINKS





YAHOO! GROUPS LINKS



Reply via email to