Friday, April 17, 2009

Thoughts About Twitter

Two observations from today's class discussion in Competing With Social Networks seemed especially important for understanding Twitter's future:
  1. The service has an exceptionally high abandonment rate.
  2. Twitter's network is comprised of asymmetrical ties, i.e., those followed need not reciprocate.
The question is: To what extent does #2 explain #1?

Before tackling that question, it's worth noting that high abandonment rates are common with buzz-fueled, fast-growing social networking services. CSN Prof. Misiek Piskorski pointed out in class that 40% of Facebook profiles are barren. In Second Life, 90% of new users abandon their avatars after just a few hours. Second Life faces two barriers to retaining users: fiendishly complex and crash-prone software that takes weeks to master, plus a big "what's this for?" problem. By comparison, Twitter is very easy to use, yet its abandonment rate is still very high. In Managing Networked Businesses this week, a class guest suggested that Twitter's abandonment rate may be as high as 98%!

In CSN, the class advanced a hypothesis that Twitter's low engagement levels might be explained by:
  1. A "howling at the moon" problem. There's little incentive to produce content if no-one is listening. Twitter's asymmetric tie structure exacerbates this problem. Some users are happy to produce with no audience, but most fail to see the point, so they quickly abandon their account.
  2. Twitter's lean, text-only format. Expressing yourself in 140 characters, without photos, etc. was held to be intimidating or at least in some way unsatisfying for many users.
I don't buy the second point. The mass market is very comfortable with IM and SMS, which typically display only short text messages.

With respect to the first point, is a new user's difficulty finding followers: a) a temporary problem that will fade as the service grows and improves friend search tools? or b) a permanent problem endemic to a service with asymmetric ties? I favor the first hypothesis. Twitter will soon have a few million loyal users comprised of the digerati/social media elite/SXSW crowd; self-anointed prophets without followers who like to howl at the moon; and celebrity fans. After it reaches a threshold scale, mass market users should be able to sign up and find some real world friends/colleagues/acquaintances who'll follow them.

If my hypothesis is correct, there are still questions about how Twitter will manage the conflicts it is likely to encounter as the service evolves:
  1. Can two fundamentally different modes of use — one-way broadcast by celebrities/gurus vs. one- and two-way communications between real world friends — continue to coexist on the same site? Technically, yes: you can follow Ashton and Oprah and Leo Laporte, and never post; I can post frequently for friends and fill my feed with @replies; others may use Twitter in both ways. But will the service ever have a coherent brand identity if it continues to accommodate both usage modes? That's a very tricky marketing problem. If I had to bet, communication between friends will gradually overwhelm celebrity micro-blogging.
  2. How will Twitter handle the inevitable tension between early and late adopters? Newbies always cause friction in social networking services because they do not understand community norms. In addressing this conflict, the asymmetry of network ties should be a benefit for Twitter: Ms. SXWS need never follow Mr. Hoi Polloi, although she may need to contend with some annoying @replies that he generates. On the other hand, Twitter will be unusually vulnerable to spam problems due to asymmetric ties and the ease with which bots can auto-search and @reply to public conversations.
  3. How will Twitter handle the blurring of social context that occurs when users co-mingle personal and professional relationships? Misiek suggested that Twitter might offer a very effective way for co-workers to coordinate and communication. This rings true, but mixing social and business relationships with a single account is a high maintenance task, as we are learning from Facebook. Users could maintain two accounts, or Twitter could support filtering for groups, but this takes more work on the part of the user.
  4. How will monetization strategies impact value perceived by the community? Daniel Palestrant, founder/CEO of Sermo, an online professional network for physicians, contends that almost online communities eventually 'arc,' that is, grow rapidly then decline, often because they introduce a revenue model in conflict with the community's preferences. Example: Facebook cluttering the site with distracting, low-CPM ads or using profile/behavioral data to target ads in creepy ways. How will Twitter handle this challenge?


  1. Tom,

    Thanks for coming to class and capturing the essence of our discussion. I am also very grateful for your comments – it is always great to draw on your vast experience and learn from it. Also, a big shout-out to Bill H. for providing the data and analyses on how people use Twitter… I only wish someone from Twitter was there to witness the discussion as CSN’ers had many suggestions that would vastly improve the strategy and the implementation at Twitter.

    So as I think about our discussion, I think it helps to distinguish between questions regarding Twitter. First, is Twitter a great one-to-many platform, similar to YouTube or Wikipedia? I think the answer to that is a resounding “yes.” With the asymmetric relationships, Twitter can help us follow people or brands for fresh and valuable content. With people self-selecting into following others, we do not need to worry about the issues regarding targeted advertising, so the monetization model should work well too. The remaining questions I have here are – why is this better than say following a celebrity on MySpace or Facebook and receiving updates this way? Or put differently, couldn’t Facebook add directional ties and essentially replicate Twitter while keeping the rest of Facebook intact? Friendster has already done a similar thing in Asia with Fan Profiles.

    The second type of discussion is whether Twitter has the potential to become a many-to-many platform much like MySpace or Facebook. This is where I think we have more disagreement. On the one hand, we see that some people use it to communicate with others, and the @feature encourages conversations, so some people suspect that there is a potential for Twitter to go many-to-many. On the other hand, the data show that 75% of Twitter users have zero tweets, and a small number of people tweet a lot! This does not look like a many-to-many platform.

    This begs the question of why so few people tweet? The first hypothesis is that most people do not have followers. From other research, we know that adding a visible audience of friends increases the percentage of people contributing all types of content by 10 times (e.g. 1.5% people contributing 50% content to 15% people contributing 50% content). So, suppose we fix that and people get their followers – will it get them to tweet more? I am concerned that this may be not enough and here is where the medium becomes important. The reason for this is that the increase in content creation for friends for visual media, such as pictures, videos, is approximately 7 times higher than for text… The opposite is the case in private relationships when no-one else is watching – this is why text and IM is so big in SMS context – there the whole exchange is very private. But Twitter is fairly public, so that has me worried about Twitter as a many-to-many platform…

    As an aside, I think there is a simple explanation why there is so little text production… People do not produce content on-line when they expect some sort of face loss, either (i) because they get criticized publicly for the stuff they posted, or (ii) they do not get a response or a comment, when there is a clear expectation that they should. Pictures are awesome in this respect – pictures usually represent us in social relationships with others (or exotic travel locations), so they usually get some praise from other people in the pictures. Pictures also do not have a clear expectation of eliciting a response. Text is different – posting a comment is often subject to more criticism (recall how many times you’ve heard people complaining about lame status updates on FB vs. lame pictures on FB), and there is also an implicit expectation of a response. So, text is harder to produce when many people are watching.

    On your other points, I completely agree with you that mixing Twitter for work and non-work uses is not a good idea. What I had in mind is more of a separate platform for work (like Yammer), or introducing privacy controls to Twitter, as Facebook did. More generally, my point was the competitive advantage of Twitter over, say, Facebook and LinkedIn is that it only focuses on text and strips away all other social data, such as pictures, and this is most useful at work. There are also substantial benefits of the Twitter-like technology over discussion boards at work, because you do not need to stop and enter data in those boards – the data is generated automatically as you’re trying to get your job done.

    Finally, a quick comment on monetization of social networks... I think there is ample evidence to show that not all networks ‘arc’ when monetized. LinkedIn’s community grew quite dramatically partly because it started monetizing ($100m) and MySpace is monetizing well ($750m) with advertising without offending its user base. I think that the secret is in monetizing in the way that creates value to the users – in the LinkedIn’s case is facilitating relationships between passive candidates and recruiters, in MySpace’s case it is the ability for users to incorporate brands into expressing their identity on their pages. The ‘arc’ problem is more salient in the Facebook’s attempts to monetize through advertising – but I believe that even Facebook will make a lot of money out of its social graph – mainly through FB Connect.

    My oh my – what was supposed to be a quick response – turned into an epistle… I look forward to more comments… Hopefully our students will chime in with their views!

  2. Tom - Great post and comment. Is there any chance you can publish a copy of the syllabus for the class? John Hagel

  3. Tom - I agree with John. A syllabus would be great. I had no idea you were teaching a class like this now. Glad to see it at HBS.

    And I'd add that when i speak on social media and Twitter I talk about several concepts including the importance of "feeding the network" - i.e. it's important as a responsible user to respond to the tweets, updates, blogs of others - to not simply post. Too many users *with* followers and friends on Twitter and FB suffer the isolating feeling that no one is listening.


  4. Tom,

    First, where's the love? That middle 45 minutes of Twitter conversation didn't spend weeks assembling, analyzing, and presenting itself! Thanks for the mention, though, Misiek.

    Second, with respect to the hypothesis that new user difficulty is a function of lack of friend search, I thought that the second body of data that we reviewed in class (the exploration of drivers of twitter usage) provided evidence to the contrary.

    Specifically, there doesn't appear to be a strong relationship between the number of tweets and A) number of followers B) number of following or C) number of "friends" (defined either HP's way or through mutual friendship).

    Put another way (and to put a different spin on the low-usage point by Misiek's comment) -- we already have a population of Twitter users that have overcome the obstacles of finding and adding friends. Yet, even these users are not driven to tweet! As a result I don't think that network-building functionality such as friend search would contribute significantly to Twitter usage.

    There is something very different that is driving users to communicate via Twitter. It's my hypothesis that this has to do with a hesitance to conduct one-to-many communication. Clearly there are sites that thrive on encouraging this communication: Yelp (does this well) and Blogger (does this poorly) are two examples. Incidentally, Yelp does have friend search functionality, but as a review-writer I find that this does little to help me "take the plunge" to write. If anything, it ups the ante!

    I am curious A) what Twitter can do to encourage this communication and B) if Twitter needs to be worrying about this at all given the growing fan/celeb social dynamic.

  5. Great commentary...some quick points...

    Can two fundamentally different modes of use — one-way broadcast by celebrities/gurus vs. one- and two-way communications between real world friends — continue to coexist on the same site?How will Twitter handle the inevitable tension between early and late adopters?I think the short answer is that they don't necessarily need to.

    It might be useful useful to separate out and Twitter's network and API-accessible platform.

    Today most people use to access Twitter, but if use continues to increase there will be more of an incentive to create custom clients for different uses (how about a hand-holding NewbieTwitter, a CelebTwitter, a PoliticsTwitter, etc.?) and "Twitter as a feature" mashups into other websites. itself seems to be following more of a strategy of highlighting "recommended" people to follow in the hopes that the "RSS-reader-substitute" functionality will provide enough incentive to draw users, who can then connect among each other. It's not a bad strategy at all, and as they get more marquee publishers, they may be able to better tailor the initial set of recommended users.

    "Howling at the moon" problemTwitter's lean, text-only formatFor new users, Twitter currently has a huge ambiguity problem. It's not immediately apparent what will happen to your 140 characters and "What are you doing?" is actually an innacurate query (how many people actually limit themselves to what they're doing on Twitter?).

    It's not evident that you can message Oprah, Comcast, or Jetblue if you type an @ before their name, or who gets to see a message if you reply to a friend or send a direct message. This is a huge problem for new users and it is basically an interface problem. Even the most basic part of the posting mechanism is unintuitive for many users - Oprah herself hit "refresh" instead of update on her very first tweet, deleting her message. There could be tens of thousands of people out there with the same issue, and that's a problem.

    Assuming people can be provided with better interfaces, the question of will they share becomes important.

    I think there are some lessons from Digg, Facebook, and from the e-mail world that can help us here.

    Bill is right about people being hesitant to share with broad audiences, but e-mail gives us some lessons as to what people do share. What goes around as chain e-mails? All sorts of political, consumer, and social articles, calls to action, and especially outrage, viral humor, etc. People may be unwilling to share their own production, but they certainly have a lower hurdle for sending on third-party production.

    Twitter's "retweet" function is still only halfway supported, and I'm sure they're worried about the signal/noise implications of making it easier to retweet, but that would definitely lower the hurdle to production.

    Digg and Facebook both have interactions (the "Digg", the "poke", etc.) that also reduce the production hurdle and allow production-encouraging feedback. Twitter hasn't institutionalized any such mechanisms, but they may well decide to...or a third party could integrate an informal one into their own client.

    How will monetization strategies impact value perceived by the community?This is a really tough thing, and Twitter's distributed nature makes it all the more difficult. Twitter will have to find a monetization strategy for its and Twitter search platforms, but also find some way to gain value from the third-party clients and API users. It may end up being much easier for the third parties to monetize Twitter, and Twitter Inc. has to guard itself against becoming a low-value network provider while at the same time encouraging these third parties to come in and add value to their network.

  6. Great post. A lot that I would like to mention has been covered by Bill and Humberto.

    I would like to add that one reason that adoption is slow (or tepid aka "protected" accounts) is that anyone who has not been an active user of Twitter can not understand the various different uses the site offers.

    I think this is primarily driven by Twitter's departure from our traditional social conventions (public "speaking", 1 to many, acceptable "eavesdropping", etc). The fact that the site is able to accomplish so much in spite of these social changes suggests perhaps as much about the weaknesses of our other tools as it does about the power of "Twitter". In other words, a community of tech-minded tweeps have hacked a very simple text-based service to make it serve a bunch of functions they want, and EV et. al were smart enough to build an API to let them push it along...

    This makes me excited for the next "social-networking" apps that are yet to come.

  7. I agree with Misiek--communities don't always decline when monetized at a conflict. As long as people are fans of an aspect of a service, they'll pay. Phish concert goers, Harley Davidson bikers, and even Starbucks coffee buyers are members of communities who continue to pay high fees at the expense of community goals (i.e., each of these communities has high costs to participation yet each would prefer 'free' participation to achieve inclusiveness in improvisation, rebellion, and community).

    At scale, more twitter users will have more followers and thus content creation will increase. I think a fascinating question remains about the way to monetize "real time" social content. while facebook's mobile app still crashes my blackberry, twitter's clients encourage distribution and search of fresh news and reviews. content produced by "anytime, anywhere" sharing and discovery could take eyeballs away from everything from American Idol to Citysearch. can't wait to see how it evolves

    [as an aside, i wonder if Sermo's narrow focus precluded it from becoming a $500+M business like Gerson Lehrman or LinkedIN. Why not cater to tech communities so investors can get more answers, or pre-med students so healthcare recruiters get more job seekers?]

  8. Interesting to see how you and your students have analyzed twitter, Tom.

    One important point, often missed, is the value proposition of "listening" on twitter. This is advice I give to companies before coaching them on how to start "engaging"and building a community; it also applies to newbies. As a matter of fact, 60% of time on twitter should be spent listening. (An aside-there are a lot of analytics start-ups aggregating data now based on content and sentiment--will cover that in a blog post.)

    The power of twitter as a research and "listening" tool is underestimated.

    After listening, I show people a method called
    "Pick Two" to begin engaging. (There are studies on the exponential power of networking when 3 people are connected.) See related post,

    Currently, my "Web of Influence" methodology has already begun to go viral.

    Regarding monetization, I covered some ideas in my March 29th post on sponsored hashtags. Also, "" sponsored by Microsoft in beta appears to be a successful monetization tool for Twitter.

    Just some ideas. Thank you.