Why Feed Subscriber Numbers are Meaningless
I am going to go out on a limb here and and claim that you don’t know how many people subscribe to your feeds. If you think you know, and you think the number is higher than 10 then there is no way to measure it. Sure, FeedBurner gives you a number. Statzen does to. While I am not claiming to know exactly how FeedBurner calculates their number, I absolutely know the general idea. And it is seriously flawed.
In order to track a subscriber number you have to count both web-based feed reader software (i.e. Google Reader, Bloglines, NewsGator, etc) and desktop based feed readers (NetNewsWire, FeedDemon, Thunderbird, etc).
The web based news readers typically pass the number of subscribers they represent in the USER AGENT header of the http request. Those are easy to count.
The way to count desktop based news reader is on the combination of the complete USER AGENT header and the requestin IP.
These methods are riddled with enough holes that the number becomes a guestimate. Examples would be:
Multiple feed urls pointing to one feed stat service using htaccess redirects. This causes web-based news readers to report conflicting numbers with each feed request.
People with desktop news reader may have multiple IP addresses in a given day. I may request a feed from my hotel, the office, a coffee shop, and my cellular card in a single day (sometimes in a single hour).
People with desktop aggregators share IP addresses at companies and schools. I think all of Belmont University still shows up as a single IP address. Everyone using NetNewsWire3.0 at Belmont looks like a single subscriber.
People sign up for multiple services. I have used 4-5 web based news readers. I think they all still report me as a subscriber to the feeds I had subscribed to.
I think these problems are particularly exacerbated with early adopters and geeks. Let’s take me and my subscription to Tech Crunch as a example. I subscribe to TechCrunch in NetNewsWire (1). I used to have NewsGator sync turned on in NetNewsWire, so NewsGator still fetched TechCrunch on my behalf and reports me as a subscriber (2). I subscribe to TechCrunch in GoogleReader which I sometimes use from my phone (3). I also subscribed to TechCrunch when I used to use Blogines (4). There are probably more. Already I am accounting for 4 of TechCrunch’s 50,000 subscribers. I am a pretty typical TechCrunch subscriber. So how many subscribers do they really have? 15,000? Could be.
I am just using TechCrunch as an easy example. The same holds true for any blog. It is likely exaggerated for a site full of people who try lots of new Web2.0 services.
So, in my mind, the number of subscribers a feed has is a parlor trick. The number is essentially meaningless. Truth be told, all numbers attached to web stats are likely meaningless these days. There are better ways to track attention.
Jackson, I have no reason to (or knowledge to back it up, if I did) disagree with you, but as some of your reasons would inflate numbers and others would deflate numbers, then perhaps it averages out well. I think most statistics tracking services are better at noting trends or making comparisons among those being tracked with the same service. Bottomline, however: I agree with you that as for providing precise numbers, many services that suggest they do, don’t.
Rex, I lazily did gloss over the fact of inflation and deflation happening simultaneously.
Yes and no. For looking at trends on a single blog then yeah, it is useful for identifying trends. When talking about making comparisons accross blogs, the tendency to inflate/deflate could easily go opposite ways.
Still, in oder to sell advertising you have to come up with something, and this something is probably better than nothing.