Twitter’s “Who to Follow” – Bad English, Great Functionality

Around a month ago I heard people talking about Twitter’s bucket test (definition) of their new personalized suggested user list and this past week I actually got to experience it myself.  The results for me have been excellent and in this post I’ll explain why this is critical to Twitter’s continued adoption.

Twitter has implemented its recommended users on your Twitter.com home page as the “who to follow” (see graphic above) and you have a list of two recommended people to follow.  Not to be a stickler for English – especially on this blog where I often race through my posts with very little editing and therefore make many mistakes – but the list should actually say “whom to follow” as you’ll see in this guide to proper grammar.

“Who” is always used for the subject of a sentence and “whom” is meant to be used as an object.  You can always tell by inserting “he” or “him” in place of the word “who” or “whom” and the result will be obvious.  In “who to follow” it is clear that Twitter is directing “you” (the subject) to follow a user “him” (the object). So I *think* it should actually be “whom to follow.”  Who knows? ;-)

But who cares? I am very pleased with the results.  I’ve written about this topic before.  Twitter has has a huge advantage over Facebook in certain types of communication and relationships due to the asymmetric following model.  I can follow dozens of people who don’t know me and get the advantage of all of the links that they provide me as well as feeling like I get to  know them a bit better as humans rather than historically only hearing from them them behind press releases.

I follow people like Steve Case, Bill Gates and Marc Cuban – none of whom I know.  In a “symmetric” following model they’d have to choose between following me back or not accepting my request.  In Twitter we can have a one-way relationship.

Interestingly, on Twitter even though somebody may follow me whom I don’t follow back, that person can still communicate with me by using the @ sign so Twitter has become this wonderfully open communication platform and when not abused can create a quick conversation with somebody normally not in your personal network.  I don’t always have the time to respond to every @ message but I read every one and I try to respond to many.

But the biggest problem of Twitter to date has been getting the right people to follow you (which is why I wrote about the topic in the link just provided).  Here’s the problem in a nutshell: in Twitter you may be following somebody whom you know well and he’s not following you back and doesn’t even know it.  When you have 50-100 followers it’s pretty easy to audit whom you’re following but when you have 10,000 it’s impossible.

Every week or so I monitor all of the new people who are following me and I follow back people who are: friends, interesting or people with whom I occasionally have a dialog and provide great links (e.g. Atul Arora whom I’ve never met but is one of the best linkers out there).

But I’m sure there are some friends whom I’m not following and should be.  And more obviously to me is that there are a whole host of people whom I know well and aren’t following me.  (some I’m sure because they choose not to! ;-))

One strategy I talked about in the post linked above on how to get people like this to follow you is to try and unfollow them and then re-follow them in hopes that they will notice you when they are alerted of their new followers.  That works sometimes but is kind of lame so I rarely do it.  I also find it even more lame to have to say to somebody “I noticed you’re not following me on Twitter” – it makes you feel so narcissistic.

But now there’s the perfect way to get the right people following you!  The Twitter algorithm as of today seems to base your recommended followers based on who is followed the most by other people whom you follow.  (see graphic at right, which you find when you click on “view all” from home page on Twitter.com) Therefore your friends are much more likely to all be following similar people and therefore appear in each other’s recommended list and that’s exactly the point!

So how do I know it’s working?  In the past week I’ve noticed 7 or 8 new followers with whom I’m friends and who have large follower lists.  These are people whom I would expect to normally be following me since we speak frequently but I’ve always guessed that they didn’t realize they weren’t following me because they had large follower bases.

It can’t be coincidence that all of a sudden so many with large follower bases and who are close enough friends all suddenly started following me in the same week.

And similarly I’ve started following a bunch of new people.  Many of these are people whom I know and some are just recommended and followed by my friends – a chance for me to get to know some new people.  And if we’re all following more of the “right” people and discovering new people then Twitter engagement will continue to grow.

I love Twitter.  Once people get over the “I don’t care what you ate for lunch” hang up – it really is changing the nature of communications.

The algorithm is alive and well.  And working.  Brilliantly.  Shame about the English.  Whom should you be following?

** Note: any spelling or grammar errors in this post were purely intentional and designed to be sure you were actually paying attention.

  • http://twitter.com/PhilipHotchkiss Philip Hotchkiss

    I've noticed this feature, but haven't taken it seriously. Perhaps that's because I am so TweetDeck and Echofon centric, that I rarely use the Twitter home page.

    But after reading this post Mark, I'll explore if further and check out the algorithm's recommendations.

  • http://bothsidesofthetable.com msuster

    I've enjoyed it. I caught a few people I should have been following and chose to follow a few new people like @dens whom I don't know because many of my friends are following him. We'll see. The beauty of Twitter is that you can always unfollow people if they become boring or spammy! Good to see you this week.

  • http://twitter.com/rubenorozco Ruben Orozco

    The “Who to Follow” feature has been great!

    I've added tons of new people I didnt know about that have my same interests and/or friends in common.

  • http://twitter.com/GypsyRaven Gypsy Raven

    Something is off about the algorithm for me, because it's been suggesting people whom I've BLOCKED in the past and people that have been inactive for over 6 months.

    I've also noticed that Twitter seems to have randomly unfollowed some people for me, just to suggest them back to me as “Who to follow”. Both issues have casted a huge shadow on this feature. So I have to say, so far I do not like this feature as much because I find the same great people to follow through RT of other twitter friends.

  • Agilandam

    “whom to follow” sounds like an order and you take it as “hi hi Sir”.
    “who to follow” sounds like a suggestion… “you are the boss…decide”.
    That is just me.

  • http://qtp.blogspot.com/ sachxn

    I too like this fact (“I get to know them a bit better as humans rather than historically only hearing from them them behind press releases.”) very much and probably this is the only reason i am on twitter…….

  • http://twitter.com/kannannari kannannari

    ** Note: any spelling or grammar errors in this post were purely intentional and designed to be sure you were actually paying attention.

    Sure, Mark :-)

  • jamespatterson2

    Mark – if you could include a comment or two about how/ if you use the “listed” function, I would appreciate it. I added a “Follow Me” site this a.m. It's easy to see how they could monetize this.

  • http://throughput.us/ consultski

    It is a great start, but has some bugs. Like suggesting people that are blocked. Or in my case, suggesting people when I am out of follows.

  • http://twitter.com/vsagarv Vijaya Sagar

    On grammar: Until recently, twitter.com (when I am not logged in and go to someone's profile) would say something like this:
    “Get updates via SMS by texting follow msuster to 53000 in the India”

    It is easy to see where & how the bug crept in. While writing that string concatenation code, some dude there thought that every country in the world needs that article prefix :-) But the incredible thing is that they let that kind of a grammatical blunder to stare at millions of non-“the US of A” visitors for so many months.

  • http://www.ribbonfarm.com Venkat

    Interesting. My experience has not been as good, perhaps because I consciously attempt a 'listen to the periphery' strategy.

    From what I can tell, the algorithm uses a basic friend-of-friend+high clout filter, it recommends people who are being followed by people you follow or who follow you, who also have lots of followers.

    This has the well-known social graph effect of preferential attachment. In other words it makes the rich richer, poor poorer. It concentrates echo-chamber effects (or what the social graph theory people call small worlds).

    This does not work for me because I consciously attempt to curate my 'follow' list towards 'peripheral listening' (or 'weak link listening') … deliberately seeking out people who are in different necks of the social graph woods from me, and from whom (!) I might potentially hear stuff I wouldn't hear from the Web 2.0 echo chamber I normally inhabit.

    The usual suspect stuff gets to me anyway, via dozens of RTs, emails, RSS etc. It is the peripheral stuff where Twitter rocks.

    That's admittedly hard to easily turn into an algorithm. I remember some librarything/shelfari type site which would find people whose reading tastes were precisely the opposite of yours. It was cute for a while, but useless. What we need is an algorithm that gets you recommendations a tunable social “distance” away…maybe they're working on it.

    Venkat

  • http://twitter.com/wfjackson3 Willis F Jackson III

    I agree with you Mark, but the problem I keep running into is that Twitter keeps recommending really chatty people. They definitely have some needed enhancements to implement to make it really useful to everyone.

  • bethtemple4u

    This feature is also good for their business because it keeps you engaged with new people to follow (sometimes the same avatars can have a 'pass by' affect) and, therefore, means you may come back to Twitter more often to find out whom they suggest next as well as returning to see what your new follows are saying. The list in 'view all' is also nicely crafted so you can eliminate from the list any followers you aren't interested in which I'm sure also fortifies the algorithm for future suggests. I agree its a great new feature … for us and them :)

  • http://rbeale.com RBeale

    While I enjoy the “who to follow” functionality, I think we need the ability to “hide” the feature from our main view. I recently wrote a blog post on the subject if you care to learn more: http://rbeale.com/social-media-marketing/twitte

  • http://bothsidesofthetable.com msuster

    Strange. I haven't experienced that.

  • http://bothsidesofthetable.com msuster

    Either way, “who to follow” is bad grammar

  • http://bothsidesofthetable.com msuster

    For me not the only reason but certainly a great reason. I also like: link sharing, keeping up with friends movements in a light weight way, asking the community questions, following my RSS stuff (like NY Times), IM people, etc.

  • http://bothsidesofthetable.com msuster

    I don't use the “listed” function. I see how people COULD use it but I don't tend to use it at all.

  • http://bothsidesofthetable.com msuster

    Suggesting people you've blocked sounds like a pretty big bug and a pretty easy fix for Twitter's developers. Hope one will read these comments and fix it.

  • http://bothsidesofthetable.com msuster

    Wow, that would sound funny to somebody who is actually in the India ;-)

  • http://bothsidesofthetable.com msuster

    Granted the feature won't perfectly for everybody's use case – but I think for mainstream users it wil drive up engagement.

  • http://bothsidesofthetable.com msuster

    If I follow OVERLY chatty people I just unfollow them after a couple of weeks.

  • http://berislavlopac.tumblr.com BerislavLopac

    The Twitter is great, indeed. Shame that it's so centralized, but as a Web service it can't really be anything else. Luckily there is solution for that, but it's not Twitter. :)

  • http://bothsidesofthetable.com msuster

    Yes, I agree with you that you should be able to hide the feature. I'll bet that eventually they build that in.

  • http://graysky.org graysky

    Good post & I agree that think is hugely good for Twitter. But I'm biased, my friend & I built “whoshouldifollow.com” in early 2008 because we saw a similar need. And Mr Tweet has been the leader to date among a few other tools. A couple thoughts after having played with it for a few days:

    - their algorithm has a cold-start issue. It is great for those who are already somewhat engaged but not super-useful to a brand new user IMHO. I would love to see an improvement to their category based suggestions to either let you search for topics you care about & find the top users (and their similar users) in those niches.

    - we supported filtering by geography since some groupings make more sense who share a location.

    - I think the “similar users” functionality will be very cool and useful for a bunch of purposes. Should be
    interesting to see what is built with it when the API is available.

  • http://www.victusspiritus.com/ Mark Essel

    I enjoyed Kevin Marshall's HiveMind application (@falicon) for similar reason. It uncovered who several friends were following but you weren't. Glad to see this happening, I'm sure it will ease in new users who have at least a handful of folks to tune in to. And it's preferred to the SUL

  • http://twitter.com/vsagarv Vijaya Sagar

    Yes, I wish:

    - Lynn Truss were Indian
    - and she were on twitter (@lynntruss isn't)
    - and she saw that message
    - and she wrote a whole new book on zero-tolerance :^)

  • http://reecepacheco.com reecepacheco

    ** Note: any spelling or grammar errors in this post were purely intentional and designed to be sure you were actually paying attention.

    ha… you sure about that? i haven't proofed it yet…

  • http://www.postlinearity.com gregorylent

    i despise it … a great illustration of the ignorance of not-very-granular algorithms … it gives me banal celebrities, people i already follow, or have blocked, or am completely uninterested in … and attempts to deprive me of following my own intuition

    visually, i really don't need a couple of more avatars hanging out on the right side

    solution? i hide all .. a couple of days later i have to do it again (there are workarounds online now for blocking it) … and doing this a few times, i am reminded again and again at how dumb the list is …

  • http://erica.biz ericabiz

    Hey Mark, I noticed you're not following me on Twitter. ;) I'm ericabiz there.

    I agree. It seems to work better than its Facebook variant, for what it's worth. I'm still waiting to see how Twitter will monetize itself, though!

  • http://kirstenwinkler.com KirstenWinkler

    I think that means services like Mr Tweet are irrelevant now, filling holes etc. So what do you think will be the next building lot to finish? Klout?

  • http://www.missi.com/ Peter Beddows

    I like this new feature and have already found several tweeters through it that I figure would be worth following but about whom I previously I did not know.

    Granted it would be nice if you could “hide” the feature form visibility on the Home page, especially since it is readily available under the “Find People” tab, but the fact that you can actually check out the tweeters suggested “before” you connect to any of them gives you complete command/control/choice over whether or not to “follow” them: No one is twisting anyone's arm and it may well be that, from time to time as new people take up tweeting, there could be an occasional real gem amongst those suggestions even if otherwise 90% of them are amongst those you would typically choose (there's that word implying you control your choice again) to ignore and, what a relief that we can always “unfollow” anyone. Undoubtedly over time, the algorithm underlying this feature will be tweaked so as to yield better selection. Just look at the amazing things that Hunch has now accomplished over time with tweaking their algorithm.

    This new feature is particularly interesting to me: Since I have 2 twitter accounts, one under my concatenated first+last name as shown here that I typically use for more general observations about life, the other under @MISSIcom, an acronym for our business and web site, that we use for stuff relevant to happenings in the business world, to our clients and our own business ~ rather as, for example, @dharmesh uses his name for his personal twitter account and @hubspot is the business face.

    We are following different people under each account but there are a couple of people, such as @msuster (where have I heard that name before?) that I follow under both accounts so it is interesting to compare the two different “Who to follow” suggestions ~ and they are quite different suggestions.

    Secondly, as an ex-pat Brit' it always brings a gentle smile to my face when Americans get into discussing the rules of English: Haven't ya'll murdered the English language enough already that attempting to correct for grammar between you becomes something akin to a comedy of errors? {grin} I brought the house down at a public planning meeting once in New Jersey when I laid out plans for a new plant I was building and said “And this is where we will park the lorries!” Since then, I have come to accept that most of my friends and colleagues here in the US speak American while I speak English with an occasional Welsh accent.

    Having that off my chest, congratulations Mark on your excellent effort to correct twitter's grammar: As an abbreviated sentence, it seems to me that “Whom to follow” would be most suitable and acceptable; it is not a command in context here, it is merely a suggestion!

    Lastly, it is not my intent to lose or upset any of my American friends, after all, I'm also an American now; however, fortunately, since hardly anyone here is following me under either account, I guess I do not have to worry too much about losing any followers if I have upset any one with my comments. What a relief! lol. Maybe I'll even pick up a few without relying upon being found in “Who to follow” popping up on their screens.

  • http://bothsidesofthetable.com msuster

    I think the most important thing in your comments is that they have to give users the ability to hide the recommendations. I'm sure that must be coming. As for not serving up great recommendations to you – I'm sure it will improve over time. For what I'm getting right now it's working brilliantly but I take it as read that in a community of 190 million people it won't be a bullseye for everybody. Thanks for the feedback.

  • http://bothsidesofthetable.com msuster

    ;-) done

  • http://bothsidesofthetable.com msuster

    I love Klout. I encourage them to work hard to diversify.

  • http://bothsidesofthetable.com msuster

    Actually, when I lived in England I found myself correcting Brits all the time! ;-)

    If you want a brilliant book to read check out “Made in America” by Bill Bryson. It will dispel your commonly held misconception that we “buggered up your language” ;-)

  • http://www.missi.com/ Peter Beddows

    Touché Mark. OK, I'll admit not all Brits are perfect! And I'll also admit I figured you have a great sense of humour and would appreciate a little teasing. You certainly give as well as you get.

    I've heard of that book but never gotten around to reading it yet; maybe I should. Aside from that, I remember you mentioned living in the UK in a previous post but do not remember the details: Where and when?

    I've not been back to the UK since 2000 and have been in the States since '79 ~ first briefly in NY and NJ then most of the time in CA.

    My wife, who is American (including some native American blood) swears she will dump me if I ever lose my accent. I was fortunate to have gotten into one of the old schools where use of correct (Queen's) English was demanded at all times, errors even being punishable. However, my early career took me all over the UK and I agree with you that there are places where not only is “Correct Grammar” virtually a foreign idea but even the accents/dialects make it all but impossible to understand the “natives”.

  • http://www.missi.com/ Peter Beddows

    Of course, it was in London according to your “About” so you would have heard many people from all around the country there at some point or another. Probably heard better English in India in fact.

  • http://twitter.com/davidcottrell davidcottrell

    The beauty is, Twitter can learn a lot about whether I accept their recommendations. They can mine that data and know what I'm interested in reading, and target me even better for ads.

    They can do interesting linkedin-like graph analysis to recommend products that I actually want to buy. For example, it could guess that 5% of my closest friends are recent parents, guess my age, and as a result, offer me EcoMom deals. Nobody likes ads that are intrusive, but if all ads were things you could want or use…

  • http://bothsidesofthetable.com msuster

    Or better yet they could just ask us to fill out a short profile like we do in Facebook and then not have to infer everything about us. Never understood why they didn't ???

  • http://twitter.com/davidcottrell davidcottrell

    Good question, I don't know. It does seem to fit with the rest of the twitter minimalism though.

    I think actions speak louder than words, and that you can infer much more about a person from what they've chosen to “like”, and view, than what they say about themselves statically. Especially since views and interests may change over time depending on your mood, what you're doing, who you're with, etc.

    Netflix's algorithm contest had an interesting spin on it. You had to predict how someone would rate a given movie. They provided data sets with the movie name, the year of release, and sets of previous ratings and dates for the movie. You could use more information if you had access and the rights such as genre, but it wasn't needed. Predictions could be made based solely on how and when you'd rated other movies, how and when others rated it, and the date. Pretty interesting.

    Soon I'd imagine more people will use the wealth of data being created to answer interesting questions. Starting with, “What are the ads most likely to get Mark to click”

  • http://www.missi.com/ Peter Beddows

    These all seem like not only really good suggestions but also relatively easy to implement and beneficial to both twitter user and twitter business. Worth suggesting directly to twitter.

    Do we not all tend to respond positively to suggestions that answer questions we have even if we are not actively conscious thinking of the related question when a relevant answer pops up – impulse buy? On the other hand, if the information presented is not relevant but is frequent then we tend to shut it out and get annoyed so it would really pay to develop the underlying algorithm to ensure a greater percentage of relevancy than otherwise.

    With the models already actively in use on the three platforms referenced – Facebook, LinkedIn and Netflix – it should be possible to develop a really helpful extension of usefulness from the twitter user's perspective that also would be monetizable as all three examples have already accomplished.

  • http://www.hypedsound.com jonathanjaeger

    I have a website account that I run and my own person Twitter handle. The problem I get is that whenever I'm on one account, I get recommendations of people to follow from my other account. I'm not getting a lot of new interesting people. Maybe that will change if I “ex” some of them out of the list.

  • http://kennethrcarter.com/CoolStuff Kenneth R. Carter

    Facebook has its own grammatical error. Its recommendation should be ” People You *Might* Know”.

  • http://twitter.com/DarkSideGeek Bill Houle

    My edumacation taught me that 'whom' always follows a preposition. Although not present, there is an implied “Of” or “In” as a prefix to the “Who To Follow”, making it more appropriate as “[Of] Whom To Follow”.

    msuster grammar disclaimers aside, wouldn't that rule require this sentence to have its who/whom flipped in position, as follows: “The Twitter algorithm as of today seems to base your recommended followers based on _whom_ is followed the most by other people _who_ you follow”? Or maybe a double whom?? Aw heck, whom knows.

  • http://twitter.com/DarkSideGeek Bill Houle

    I agree, Venkat. Left to its own infinite devices, it would seem that my follows would become my follower's follows, and similarly cascaded for everyone else. The system would trend us all towards homogeneity, as if we are all just nodes of a Self Organizing Feature Map (Kohonen neural net).

    In your “periphery” strategy, you are at the center of a Venn diagram of a multitude of interests – the very epitome of “social circles”. But as your follows get accepted as your follower's follows, that intersect area gets larger and larger until it becomes an encompassing circle itself.

    There is joy in exchanging nuggets from the periphery.

  • Priscilla

    Hello, I read your tweat and I think you are on to something. I represent S.E.E. (Speak Excellent English) and if people are following you, I think I should to.

    I look forward to your future tweets.