#1. Accurately measure subscriptions. That seems obvious but FeedBurner is actually terrible at one of their core features. It's not uncommon for your subscriber count to bottom out to zero from one day to the next as their system fails to work.

#2. Find a way to find and eliminate the cheaters. FeedBurner's subscribe-via-email feature is a great way for people to scam the system and raise their subscriber counts. All you have to do is make a bunch of dummy accounts, which is trivial for anyone with any kind of decent hosting package, and then use those accounts to subscribe to your feed.

---

Subscriber counts are an important metric for both ego and site valuation. The fact that FeedBurner has such a struggle with accurate counts hurts their credibility too much.

The problem is that the "subscriber count" metric is bullshit.

First I will explain what that count is. Then I explain what we think people want from the subscriber count and how we are working to address that need.

The subscribe count is based on the feed reading tool and the IP address (if it is a desktop client like NetNewsWire or FeedDemon). FeedBurner has explained a fair amount about this number and this post provides some good background (http://blogs.feedburner.com/feedburner/archives/2006/09/a_pe...).

The problem is that "subscriber" is supposed to equate to "person", but it does not. In email marketing this works because we have a person identifier (the email address). There is no similar identifier for feed. What that means is that out of TechCrunch's 100,000 subscribers, I as an individual represent 5+ subscribers (once on Google Reader, once on bloglines, once on NewsGator, once with NetNewsWire at home, once at NetNewsWire at work). I am a pretty typical TechCrunch subscriber so it is possible the number of "people" is actually 20% of the number of "subscribers".

The converse is that many corporations and Universities have only 1 outgoing IP address. So all FeedDemon subscribers at Belmont University here in Nashville show up as 1 subscriber.

Why this is important is because of what people want the "subscriber count" to be. They want it to be an apples to apples comparison to compare popularity across blogs. It is just too inaccurate to be that measure.

We initially had this count and as we looked at the data we felt that it was inaccurate at best and sometimes completely misleading. This was the driving force in our decision to remove this feature. It is a risk, but I think it represents an intellectual honesty that our customers will appreciate.

The other reason people like the "subscriber count" metric is to measure growth and decline in their individual blog. This is why having the number fluctuate so wildly in FeedBurner is so frustrating.

We have not solved the apples to apples comparison yet, but we have some promising progress.

As for measuring growth, once you break that out as a goal it is easy to see that there are other ways to answer this question.

As you said "The fact that FeedBurner has such a struggle with accurate counts hurts their credibility". Hopefully our willingness to find a better way, even if it may be perceived as a shortcoming in the short term, will help our credibility.

What do you think about our strategy here? (I really appreciate the feedback we are getting from HN. This is such a valuable resource!)

Do typical TechCrunch subscribers subscribe using five different feed readers? I don't subscribe to it anymore (I do follow @techcrunch on Twitter though - TC posts too often for my feed reading habits) but when I did I used, well, one feed reader. I assume that is pretty typical. Besides, your e-mail example doesn't work either if someone has five e-mail addresses!

Going off of the stats on several of my feeds (ranging from 18,000 "subscribers" down to about 1000) I see that in most cases Google Reader accounts for 50%-70% of subscribers. The numbers Google provide should be pretty accurate.. and so while there might be a slightly inaccurate swing in the remaining 30%, if the Google Feedfetcher number is showing 10,000 subscribers, I am pretty much guaranteed that's 10,000 unique Google users.

As a publisher with a significant audience for my niche, I find the subscriber count very useful and it provides much of the reason why I've stayed with FeedBurner despite the issues. It makes it easy for me to prove I'm the biggest in my niche because the closest people have counts over 30% lower than mine.. and I can't see the swing being that much considering how popular Google Reader is now.

It is a pretty nuanced issue. For instance, do TC subscribers USE 5 readers? No. But web-based readers that people quit using still report subscribers. I don't use 5 readers either, but I have multiple readers reporting me as a "subscriber". This comes from trying different web-based clients as well as using multiple IPs in a single day with my desktop based client.

While you are absolutely correct that Google Reader is going to report accurately how many people have subscribed to your feed using their system, that doesn't give you any indication as to how many of them have read a single item in the past 6 months. This is why FB started to add "reach". In the article I referenced, TC had over 100k subscribers while "reach" was topping out at about 25k.

So, all of that is to say, you are right. We are going to have to provide a way for bloggers like yourself to easily convey the worth of ads on their blog. Personally I believe everyone wins with accuracy.

I never figured out how to do it. There were two problems involved for me:

1) I couldn't tell whether it worked or not before I posted something on the blog, thus sending a feed out in the ether.

2) I didn't try very hard.

I don't know that much about RSS feeds and how they work, and frankly I don't care either. I just want it to work.

A straight and simple approach that doesn't require me to have any prior knowledge would work wonders. I think there are many people like me that have a blog, but don't really have the time or expertise to get acquainted with how feeds work and what you can do with them.

Btw: I love this:

Can statzen handle a very high traffic blog?

We think so. Maybe. Hopefully. [write me]

Hope you didn't post it anywhere else than HN where people will understand that this is pre-launch :-)

I know what you are saying is 110% true. RSS works best when it just works and people don't need to know "how". We are trying to make the feed integration part as easy as we can for each specific scenario.

Some people want to know how it works, so how would this work for you...

Simple step-by-step instructions that have no clutter to sift through, but a "?" link at each step for people who may want more information.

Would that have reduced the barrier to installation?

And yes, this is only posted here. I am glad you got a kick out of my placeholder copy :)

I'm currently thinking about the exact same thing, links to small help-boxes scattered around the place where users might have trouble, for my own project.

Often times, I find I just check which of my posts people like reading, and it gives me a tingle bit of feel-good before I close it and go look at something else. However, I often have to do 4 clicks to get there.

FYI, we have post specific attention data for both web and feed right on the first page after logging in. screenshot: http://flickr.com/photos/jaxn/3121401122/ (blue is web, orange is feed)

It is our feeling that post-centric analytics is really what bloggers want. It also allows us to do topical analytics as well.

Also, I don't currently include an "all time" view, but I am going to rethink that decision.

Thanks!

Do you include an invisible 1x1 picture in each RSS item so you get that data? (That would have been my first idea)

Or rather, it was perfect until Google bought it and then it tanked.

Please do keep an account of your efforts on a blog or something...

The scale of information they collect. I am paranoid. Google and other web services collect a LOT of data and I'm just not comfortable with it.

While I do not intend to stop using _all_ services altogether, I do plan to minimize it as much as practically feasible.

This includes using my own email (yes, the web-host/ISP, as the case may be, have some degree of access to it unless I set up my own hosting server)

Sanitising my search queries along with using random rotation proxies if possible. perhaps use TOR. This will depend on acceptable speed achievements. If it is too slow, I will think of other ways.

Blocking all _known_ google (and of course other services too) tracking applications, viz. Analytics, Feedburner, Adwords, Adsense etc

And others I'm still thinking about.