Yes, there’s no way that we’ll ever be able to get nearly the control over our third party application pages as we would over our own blogs or websites, which is why some of my wishes are frankly destined to be nothing more than pipe dreams. However, I think it’s important to note that Facebook, a website that does control their ecosystem very closely, does provide somewhat robust tracking tools for its Fan Pages. It’s one of the reasons that I’m disheartened by the amount of tracking that Twitter provides.

Twitter is sort of like the baseball of social networks - everyone wants to analyze (and overanalyze) the existing numbers to see what it all really means. As such, it would greatly benefit those statistically-minded (read: obsessed) folks to have some more data points to use. The recent boom in baseball analysis has come from creating new, meaningful metrics (BABIP, wOBA, OPS+) around the current and old measurements (AVG, RBIs) Whether it happens or not, we’ll just have to wait and see. My goal in this post was to ask for a couple things that we don’t currently get (UUs, demographics, origin) along with some that we could ascertain (ListTotal, TweetReach). Some of this would be fantastic for personal use and analysis, especially for corporate brands, and others would be good for defining influence.

Unique Users: I was defining a “visit” or “pageview” as when a user visits your homepage (i.e. twitter.com/jaremy). It doesn’t account for the majority of the content I put out, which is read through feeds, but would give me an idea about how many people are actually interested enough to visit my page after reading a tweet or seeing my profile somewhere else. From there, I could later figure out what percentage of people (ballpark) are actually adding me based on following me. It might help me ascertain whether my biography or tweetstream is doing a good job of converting followers. This is different from TweetReach, which has to do with individual tweets.

ListTotal: I count anyone following the list that I’m on as “my follower”. I think the reason to do that is to track the total number of people tracking you on lists. Yes, there might be some overlap, but I’m skeptical as to how much there would really be. I can’t imagine that there is much more than a 5-10% difference in the total number of people following me on those lists and the total number of “followers”. Don’t quote me on that though :). But if we’re talking about someone on 300 lists, followed by 3,000 people, I’d imagine that there’s even less overlap. It would probably be worthwhile to do a case-by-case study of those numbers, but my guess is that unless someone is on relatively few lists and has relatively few followers, the amount of overlap is minimal. Therefore, I believe it’d certainly be a useful metric to track, at least when comparing apples to apples (i.e. not comparing someone featured on 25 lists with someone featured on 250). Using the tags, you could even just compare similar people (actors vs. actors, tech people vs tech people, bloggers vs bloggers).

FollowerOrigin: Yes, I think this would be very useful. But I agree, there’s absolutely an attribution issue. I don’t think you’d be able to come up with totally accurate answers for where ALL of your followers came from, but the same is true about online advertising (we measure clickthroughs and conversions, but oftentimes a conversion comes much later than the initial clickthrough, or without a clickthrough occurring at all). I don’t think the attribution issue makes the measurement any less worthwhile.

Demographics: This was one I thought that was the least likely. Facebook has a step up, because of the information tracked on profiles. Less than ideally, this tracking could be done with a service like comScore or Hitwise, but I’m always skeptical about their demographic information anyway. This would really rely on Twitter either a) changing the information they pull from profiles, or b) integrating some other service. Would be undoubtedly useful, but is nevertheless unlikely.

TweetReach: Best way to fudge this would be to take the total number of followers that each link could have been served to. So my 448 followers, plus the 1112 followers if you retweeted my post, etc. It’s honestly something that the new Twitter Retweet function might actually be good for. This is likely something that Twitter would need to eventually integrate, but I can’t imagine why they would not be able to put it together (# of Followers Tweeted + Person1-RT(# of followers) + Person2-RT(# of followers)… - OverlapFollowers). For a third-party team (read: Twitalyzer) to do this, they might have to fudge it more, to create an aggregate using a user’s most recent RT mentions to determine overall reach. Something similar to how Twitalyzer calculates influence, based on the number of followers for each account retweeting one of your posts publicly. This method wouldn’t allow you to measure an individual Tweet (which is what I’d ideally want), but would at least give you an idea on a day-by-day or week-by-week basis as to whether your Tweets are reaching more or less people.

The ideal, however, would be for TweetReach to show every single tweet that a user “saw”, or was “served” through a feed/stream. Again, there would be issues with this, as someone might be “served” 1000 tweets but really only read 50 of them, but it would at least (hopefully) not count inactive accounts or those who haven’t read their stream all day. A great ancillary feature of this would be to see which days/times are the best for Tweeting to your own list of followers, based on when your highest TweetReach is.

Hope that helps to clarify. I try to write my posts in a little broader strokes, but I am always happy to delve into greater detail :). Again, thanks for the responses, Jeff.

Let me start by stating that if my wife got me a great DLSR camera for Hanukkah, I would be pretty happy.

First lets get the obvious out of the way. The amount of data and control you have over your site or blog will always be drastically bigger than any third party applications, be it Twitter, Facebook or the next big thing. Also, you can’t really compared data from Facebook to Twitter. Facebook controls their ecosystem much, much closer than Twitter - which is why there are not over 50,000 applications based on Facebook like there are using Twitter. Facebook also happens to ask users (or allow for) for more demographic information than Twitter.

OK, now my 2 cents on your metrics

Unique Users: This is interesting. What do you mean by “how many people are visiting our own accounts” When I see your tweets on Tweetie or Hootsuite, I may be explicitly viewing your tweet…or I may not and you just might happened to be in be in my stream at that moment but I did not see it or browsed by it. What is a pageview in Twitter?

ListTotal: As I mentioned in your post about this matter, while I think the Lists have some great use for users, they are still very much a wild card in terms of terms of engagement and importance. I think you are taking some assumptions that based on your 25 lists (26 as of 10 minutes ago) you are on, that you have 302 followers. How many of are actually following you? Of those 25 lists, how many people are on more than one list?

FollowerOrigin: Capturing Referral Channel would be a great one. The tough part would be associating the correct point of conversion. For example, I actually found @jaremy from this website which took my to your profile on twitter.com. However, I did not follow until a few hours later using Tweetie on my iPhone. Who gets the attribution?

Demographics: Yup, but Twitter does not require much in terms of user information. Even location from your bio is too free-form. Go here http://www.twitalyzer.com/twitalyzer/list.asp?uri=list.asp and start to type in Seattle and see how many variations come up. Geo-location is good for real-time targeting, but not demographics. Until Twitter makes drastic changes to it’s user profiles, this will be on your wish list for quite some time.

TweetReach: This would be cool. Links are easy to track, views are not. Again, depending how I am viewing my stream, I may have been served up an “impression” of your tweet, but there is a good chance that I did not read it (not yours of course, those I read

“I believe that each and every one of them would be relatively easy to build in” OK. If you can put together the computations needed for these metrics, I will see if I can help you.

We’re looking into new analytics/items to build into our vScore all the time, but there’s currently a long queue of features and only so many developers.

By the way, I just got a new D3000 and I’m loving it.