SocioTwitting: developing metrics for Twitter volume vs. Twitter influence

by Urs E. Gattiker on 2008/04/28 · 4 comments 1 views

in a analytics taking action,b why benchmark analytics,c micro-blogging Twitter

One can get lost in the conversation metrics for Twitter, so I began mapping a few measurements of my own to see if they made sense.

It might be of interest to understand how one could measure the reach, influence, reputation management through through micro-blogging using some conversation metrics that tell the whole story. I take a stab at the problem here.

I have talked about the micro-blogging phenomenon Twitter here:

Twitter – means getting a better conference experience

I have also pointed out that Twitter does not really allow one to have a conversation in the classical sense of the word:

Twitter: wake up and smell the coffee – have a conversation instead

Twitter is one of these social media tools we can use to talk with others and it enables each one of us to show presence in the virtual space.

Ever more people are using Twitter for all kinds of purposes and I got intrigued about how to measure what is going on – conversational metrics.

When we visit Twitter and a particular person’s ‘micro-blog’ such as ComMetrics on Twitter, you can see the following data:

    – the number of people following us on Twitter (Twitter labels this indicator with Followers – you can see the number but cannot see the list of people with pictures – only individual person that owns the Twitter account can see this),

    – the number of people the person is following on Twitter (Twitter labels this group Following some like Twitterholic call them friends – a term whose use I believe is inappropriate in this context – this info can be viewed on the person’s Twitter profile page on the web), AND

    Updates – some people update the account 20 or more times each day, others 3-6 times a day and less on weekends (U.S. more frequent, Europe silent on weekends – not attempting to stereotype or generalize but my experience so far).You can get the above data points also by looking at this:

    Twitterholic takes some of these public data points from Twitter about ComMetrics

Getting data about Twitter

There are several ways to get information from Twitter. One is to use the person’s RSS feed that lists the last 30 posts or so:

ComMetrics – RSS feed via Twitter

Another approach is collecting data from the people one follows. Such data enables one to obtain certain metrics as illustrated here:

Twitter spewage amongst my contacts. 2008-04-28 Dave Winer

Alternatively, one may put it in a spreadsheet as was done here:

What’s Your Twitter Noise Ratio? 2008-04-25 – Louise Gray

What metrics can be used for benchmarking?

Regardless of which approach we use, we have to address some methodological issues, such as:

1 number of tweets posted during period – some people post a lot, others very little, so how is the impact in comparison to the number of tweets a person posts – who gets the biggest bang for the buck.

Looking at the amount of tweets (see also below) it might be interesting to compare people who tweet every few hours compared to those that post more than one tweet every hour of the day! Can you still find the gems amongst all the chaff?

2 number of tweets ‘addressed’ to a person / total number of tweets in time period – addressed tweets are those starting with the @ sign such as : @ComMetrics

    Dominic Jonesdomjones @ComMetrics Yes, I do sleep if people leave me alone :-) In reality there are clones of me dotted across the time zones, inlcuding on ships.

Nevertheless, this statistic by itself may, again, be meaningless because it may just indicate who wants to influence Pistachio. She seems to ‘tweet’ to tons of people as these usage statistics suggest.

3 – beef ratio, number of tweets with URL / total number of tweets posted. Tweets with URL are those giving a link to a blog post or white paper:

Jowyang, besides some chatter (e.g., telling us he is going offline to hanging out at the pool), tries to provide his followers with nice links adding much beef to his tweets and keeping the conversation going.

4 Followers / Following ratio – in practice, if you follow 5,000 people and you have 6,000 following your tweets (followers) – it takes time to keep track of it all. In turn, some people revert to drastic measures, such as:

Stowe Boyd wants you (trying to get him interested in your product or service) to pitch him on Twitter using this approach only

The above helps Stowe Boyd keep track of things. This way he knows which vendor wants to pitch what product to him. Of course, why would I want to tell the world that I need to influence Stowe Boyd. I prefer talking to him without everybody knowing if I wish to influence hime or anybody else for that matter.

The problem of not spending too much time reading these tweets while not loosing those one needs is real one for people who have a huge following. Just imagine, 10% if those following you (600 out of 6,000) send you a personalized message or reply using @ every other day or so. Maybe if one scans these messages only, one spends about 10 seconds at most on average. Nevertheless, it could add up to 1 hour each day. That is a lot of time if not too much.

Nonetheless, what does it mean if your ratio is something like 5 to 6, whereby 5 times as many people follow you as you follow yourself?

By itself, it may not mean that much. However, with some more indicators it may show a trend that people care and want to hear what you have to say.

5 Influence is hard to track. For instance, Tweeterboard tracks who replies publicly to a given user with the @ symbol and their user name. An example would be an answer, such as @Stephtara or @GoingSolo. Tweeterboard also keeps track about the persons one replies to using Twitter. This means those messages one sends out with @ sign and user name at the beginning of the message.

It then claims to apply a secret algorithm [boy, everybody in the U.S. seems to use algorithms … they all must be wizards at the numbers game ;-)] to assign a number of points to a user. Similar to Google, value is conferred to a user through links from people who have had high value conferred to them by links from others, and so on.

Therefore, what Tweeterboard does has little to do with influence but much with what we call crowd scoring instead. In short, the more @ComMetrics messages I get from others, the more influence I seem to have. Not a very logical or helpful measure, is it?

What Tweeterboard shows about ComMetrics

Bottom Line

If you follow 500 people and they each tweet about 10 x a day, you are spending very much time with Twitter if not wasting valuable work time – remember you need to make money. In turn, maybe think about this statement:

Number following means: “It serves to enhance a user’s appearance, and is not a declaration of either interest (level or kind) or attention (listeners are not always listening).” Reputation, Conversational Index, Twitter, and Tweeterboard – 2008-04-16 – Adrian Chan

Accordingly, what is needed is a set of statistical indicators that give us a better approximation of reality than what we might get by using one indicator.

It is obvious that the indicators 1-4 presented above can play in important part in figuring out reputation and influence-related issues in regards to Twitter participation.

Stay tuned, we will bring you these indicators and other metrics for measuring your ‘influence’ on Twitter soon right here.

also of interest:
4 Lessons we can Learn from Mercedes-Benz, AOL and Wells Fargo The ComMetrics Index
what indicators are used to rank methodology – how we measure social media efforts


Check out: trend spotting – TWITTER – checklist for building trust with your global social media audience

  • adrian chan

    thanks for the summary — obtaining metrics on twitter for the purpose of measuring influence, topicality, continuity, social graph/networks will take some higher level math than what’s out there now!

    Social practices particular to twitter have emerged that skew and bias the value of simple metrics, as you’ve shown in each of your above points. Twitter’s keep-it-simple-stupid approach to follow/follower, @reply, @cite (@name in body of post) make those stats — unreliable because they’re used by twitters to establish presence and perception. Which is not a complaint. In fact twitter’s success owes in large part to the appearance of influence and popularity it can provide a frequent user.

    And in terms of conversation, twitter looks more like a very slow chat or fast status feed, and so is really an aggregation of individual missives than threaded discussion. I guess that’s stating the obvious, but sheer volume and popularity of twitter have us all wondering how best to analyze it for topical and conversational value.

    It’s a discontinuous form of talk and a disaggregated audience, through a channel that disintermediates — anyone is immediately proximate to anyone else. By laws of social interaction you’re going to get high levels of redundancy and low levels of follow through and reciprocity. In other words, it’s impossible to measure attention on twitter.

    That might be the crux of it, in the end: can we measure attention and “listening?” What I started to call signal/noise ratio in response to recent posts on twitter noise and volume.

    At this point only going outside of twitter to use influence and topical metrics from member activity on social networks and blogs might paint a reliable picture of who’s who and what they talk about.

    Over time, perhaps, with group twitter, hashtags, and channels, we might get a richer conversation space and thus a better view of who’s where, talking to whom, paying attention to what, and who’s listened to.

    At which point we’ll be able to call it AOL chat?



  • Urs E. Gattiker

    Adrian, have to agree with most if not all you say. Just wanted to add a few things to clarify further.

    Donomic Jones put it best yesterday on Twitter:

    “0. Number of blog posts written since started this personal twitter account. Must go now.”

    And the tool was depriving him of sleep. I suggested he go to bed earlier that morning (GMT time) to which he responded:

    2008-04-28 – “Dominic Jones domjones @ComMetrics Yes, I do sleep if people leave me alone :-) In reality there are clones of me dotted across the time zones, including on ships.”

    As I stated in an earlier post:

    this might be chatting or similar to instant messaging – exchanging word bits. Not a classical type of conversation we can hold in a coffee house.

    So what you state above I surely agree with. Nevertheless, as you pointed out as well, corporates want some measures. Hence, we went and developed some that provide us with an overall picture regarding social media – for Twitter as well.

    With the help of these tools and charts we get an overall view. This is part of the work that must be done. The rest is left to a human analyst who has to make sense out of it all.

    And methodology plays a big part in making sure that we get some indication of how much listening or attention a brand or person really get, whilst using Twitter.

    Thanks for the very insightful comment.

  • adrian chan


    Sleep deprivation is a sure sign of twitterholism! The tool is a slow synch chat. I find it not unlike IRC in the old days, or visiting stock message boards during the dot com days. The pull of what’s not there, and of what you might be missing if not paying attention, is the power of absence — a power that attracts — and is in contrast to the power of presence that governs face to face interactions. That’s the medium, in a nutshell.

    What we can’t negotiate, through physical presence, we negotiate mentally, and even emotionally, but through the force of attraction (and seduction even), rather than through extension of self-presence.

    It’s for that that there’s such a high degree of projection, introjection, introspection, reflection, and transference through mediated interactions. (I have a ton of stuff on this on my site).

    Hence, also, a need for psychological frameworks of user types, and of psychologically informed models of user behavior. Film director Elia Kazan described his “actors studio” approach to method acting “the art of turning psychology into behavior.” Same thing applies, IMHO, to online interactions.

    There’s a great deal of social presence bias in the following/follower activity on twitter, and I suspect a significant difference in the ratio between talking and listening. Some users post but don’t read, some read but don’t tweet as much. Similarly, on Facebook, some users maintain their presence on their own profile, while others leave posts on others’ profiles. Presence extension is a means of being visible. On twitter, there’s no presence unless you post. No measure of page views, visits, to provide users with a sense of acknowledgment. In strange ways it’s modeled on chat more than it is on the blog, and while it’s called micro-blogging by many, I think it’s more slow synch semi-public chat.

    Either way, you’re right: metrics are needed, and are possible. They’re certainly desired!

    cheers and I much enjoy these conversations!


  • Pingback: ♥ Suzie ♥

Previous post:

Next post: