Category Archives: Tools

Twitter Secrets of the Obama Campaign: #1-Tracking Clicks

Posted on by

[For the background on this series, please see the Introduction first]

Did you know they’re tracking your every click?

You probably did.

Do you know how and why they do it?

The secret is the link.

Follow Obama on Twitter or Facebook, and you will see a steady stream of postings with quotes, information about events, and statements of policy.  And because of the 140-character brevity of Twitter, most of these tweets are just teasers with hyperlinks that lead to further discussion, photos, or videos.  Like this one:

You can tell it’s a video, obviously, but the URL looks strange: OFA.BO/FF1NZ1.  It’s clearly not YouTube or Vimeo or any other well known video web site.  But when you click the link, you find yourself nonetheless at YouTube:

How’d you get there?  That’s mystery #1…

But the plot gets thicker when later on you notice a pair of similar Obama tweets:

Both tweets are on the same subject and if you click either link you will be taken to the same anti-Romney page:

But look closely at the two tweets again: each tweet has a different OFA.BO URL. Why? That’s mystery #2…

Fortunately, the answers to these mysteries are fairly straightforward.

With mystery #1, as you no doubt suspect, the Obama campaign is using a custom URL shortener (for more details on URL shorteners, see this article).  You’re probably familiar with more popular URL shorteners like bitly and TinyURL.  The Obama campaign is using one of these services (see Asides at the end), but working with their own private domain, OFA.BO.

When you click on the link http://OFA.BO/FF1NZ1, the OFA.BO URL shortener (which is just a specialized web site) “unshortens” FF1NZ1 and tells your browser the actual web site you need to visit is at http://www.youtube.com/watch?v=uoEY1W2Fqxs. Of course, your browser hides this two-step process from you.  Instead, it quickly takes you to the YouTube page – so quickly that you don’t notice the stop at OFA.BO. It gives the illusion that if you click on the link in Twitter you’ll go right to YouTube 

But while you were going through OFA.BO, your click is captured by the Obama campaign:  they know which specific tweet caught your eye and got you to click on the video.  Very tricky!

The motivation for doing this is pretty simple.  Like any marketer, the Obama campaign wants to know what works and what doesn’t.  A tweet that gets very few clicks is a failure and won’t be repeated or imitated.  But a tweet that gets an enormous amount of clicks is a success and will be used as a model for future tweets.

This is especially useful for them when the campaign is trying to figure out how to phrase or explain something.  That’s the solution to mystery #2, where the two similar tweets have different shortened URLs: the Obama campaign wants to know which of the two messages  gets the the biggest reaction, and so it tracks each tweet by giving it a different URL.  Is it the tweet focusing on Bain’s profit at the expense of jobs? Or is it the other focusing on what sounds like mismanagement?  By the time you’re reading this, the Obama campaign has undoubtedly recorded enough clicks to know if either or both gets people fired up.

The value of this analysis extends well beyond Twitter: as messages are tested on Twitter and the best phrasing discovered, the campaign knows better how to present Obama’s messages in person and in traditional media.   If you hear Obama speak on the subject, and he focuses on how much money Bain made from Ampad, you know the first one got all the clicks.  You can almost imagine them using Twitter not for Twitter’s sake, but to research how to put an important speech together!

Of course Mitt Romney’s team has its own URL shortener: MI.TT.  They too can track clicks.  Interestingly, they haven’t figured how to test multiple tweets to hone a message.  That’s probably because they don’t tweet anywhere near as often as Obama does right now, and multiple tweets on the same subject are rare for them.  But in the times they do, the campaign doesn’t use unique codes:

 

Granted, there isn’t much difference between these two tweets, but Romney’s team has no way of knowing which of the two is more popular. You snooze, you lose… Politics, like football, can be a game of inches.

There are some key take-aways from this first Twitter Secret of the Obama Campaign:

  • Always use a URL shortener for links you post on Twitter (or Facebook).  Make sure the link shortener has a convenient way of producing reports on people’s clicks. 
  • Try out different forms of your messages, and pay attention to which one(s) work the best. It’s more effort, but you’ll be rewarded with a better knowledge of what succeeds on Twitter (and perhaps beyond).
We’re lucky that the tools to do this kind of tracking and testing are readily available to all of us, not just to well financed campaigns.  You can use something as simple as Twitter’s free Tweetdeck to compose and schedule messages along with bitly to track responses (also for free), or move to an integrated application like Hootsuite which provides scheduling, link shortening, multiple social network management, and analytics all in one package. See the Resources section at the end for some suggested places to start.

For some of you, this secret seems like it is just common sense.  And if it does, I apologize.  But I’m always amazed by the number of people who don’t use anything but the most primitive tools for using Twitter. As an example, in this analysis I did of what Twitter tools US Senators use, over 50% of their usage is via the web interface at Twitter’s homepage.  As a result, they have no way of knowing whether people are responding to their tweets or not.  And these are politicians with jobs of national importance — and who only keep their jobs if the public approves of them!  They should know better.

The learning curve for getting reports out of link shorteners is painless.  Anyone whose tweets are in service of a larger goal has no excuse not to begin to use these tools in their social media activities.  

Give it a shot and see what you learn.  When you do, you can feel proud that you’re using the same avant-garde techniques the President’s campaign team is using!

Keep up to date with future updates to this series by following me on Twitter and/or subscribing to updates to this website. To see all posts in this series, visit the overview page.

To move on to  Secret #2, click here.

Notes

  • Why have a private URL shortener like OFA.BO? One reason is just pure branding. But another reason comes when people re-tweet something from the campaign.  No matter whose tweet(s) you look at, if the URL starts with OFA.BO, you know that the URL originated from the Obama campaign and is safe to click.  If you see a bit.ly or any other link shortener’s URL, then you really don’t know — is it a link to real campaign content, or is it a link to a spam or phishing site?  If you are a large company with a reputation to protect, you should probably invest in your own private URL shortener for the same reason.  This cost is negligible.  You can even use the same services Obama and Romney use!
  • Twitter adds an extra wrinkle to the process of getting to your final URL, but it doesn’t fundamentally change the process.  Twitter runs all URLs (even ones already shortened elsewhere) through their t.co link shortener, so your browser actually goes to t.co first.  In the case of Obama’s links, t.co will return an ofa.bo URL to the browser.  Then the browser goes to ofa.bo and retrieves the YouTube URL.  Finally, the browser gets the YouTube page and you see the video.  You still probably don’t notice the process of bouncing around, but it is going to be slower than going right to YouTube.

Resources:

Asides:

  • The OFA.BO web site is managed by a company called ShortSwitch.
  • MI.TT is managed by Bitly.
  • OFA stands for “Obama For America”. .BO is the country domain for Bolivia. 
  • .TT (of MI.TT fame) is for Trinidad and Tobago.
  • Either country, if they wanted to annoy a candidate, could cancel the candidate’s domain and leave his URL shortener out of action. That’s not likely to happen, but we can guess that bit.ly was nervous recently since “.ly” is Libya’s  top level domain…

How to craft the most re-tweetable tweet you can!

Posted on by

You will find endless advice on the Internet on the art of writing a tweet that will catch the eye of your audience, all focused on how to find the right way to say what you want.  But there’s another, more mechanical aspect of writing a tweet, one that relates to not what is said, or how you say it, but the structure of the tweet itself.

Mechanically speaking, when you are writing a new tweet that you hope to have re-tweeted, there are four important things to consider:

  1. How long is your text?
  2. How many hash tags should you use?
  3. How many other Twitter users should you reference?
  4. Should you include a URL?

I looked at roughly 100,000 recent tweets and 100,000 recent re-tweets on Twitter to see if I could discover any pattern to them and, hopefully, some guidance that can be used in crafting the perfect tweet.

First up — how long should your tweet be? For most of us, the challenge in Twitter is getting our tweets to fit in the 140-character limitation.  My first draft of a tweet tends to be way over the limit, and the editing process consists of removing words, rephrasing things, and using short-cuts (like “&” instead of “and”) to get the length below the limit.  And this results in tweets that are close to 140 characters in length.

And when you look at the distribution of typical tweet (but not re-tweet) lengths, what you see is pretty much that:

Click on Image to Enlarge

Once you hit about 60 characters in length, the percentage of tweets remains fairly flat until 135 characters, where it takes a sharp rise — that’s the zone where people just can’t fit another word in. About 20% of all tweets are in the 131 to 140 character range.

But does maxing out your text produce a tweet that will be re-tweeted? Here’s a chart that shows the likelihood of a re-tweet by length of tweet:

Click to enlarge

This is almost the inverse of the average tweet lengths!

We can see, pretty dramatically, the probability of a tweet being re-tweeted drops as the tweet gets longer, until around 36 characters or so when it starts to level out.  Short tweets get re-tweeted!  Even holding tweets to around 30 characters is a dramatic improvement over adding just another word!

So if the ideal tweet is pretty short, what does that mean about hash tags?  Here’s a distribution of re-tweets by the number of hash tags in them:

Click to enlarge

And as we would expect from the dominance of short re-tweets, the number of hash tags is usually none, and occasionally one.  I’ve seen advice that suggests people use no more than two hash tags in a tweet, and this definitely bears that out.

If you look at the number of @references that people make in their tweets that are re-tweeted, the chart looks virtually the same.  Zero is very popular, one is OK, and over that drops off the list.

Now, to the last question: Should you include URLs in your tweets?  My natural inclination was yes, always, because that’s my call to action.  But lets see how that works with tweets and re-tweets.  Here’s the percentage of original (non-re-tweet) tweets that have URLs:

Click to enlarge

And here’s the percentage of re-tweets that have URLs:

Click to enlarge

We can see that re-tweets tend to slightly favor tweets without URLs, probably driven by the inclination people show to re-tweet short tweets.

For those of us who like to cram every last bit of information possible into a Tweet (and that’s me!), these statistics present a challenge.  Long tweets with lots of hash tags, references, and URLs just don’t generate the kind of engagement we’d like.  People often describe Twitter’s limit of 140 characters as being constraining, but the practical limits are even worse.

At the end of the day, you can’t prune so many words out of your content that you shear it of meaning.  But there is still some advice I can give:

  • Have one simple idea per tweet.  Don’t try to cram more in.  Words like “and” or “or” are dead give-aways that you’re putting multiple ideas in a tweet; you should split it up into two or more tweets.
  • If you have problems coming up with enough tweets to meet your goal, this is your chance to take one idea and play it out over several tweets.
  • But don’t spread a paragraph across multiple tweets.  You’ve seen people who write an essay in twitter and just break it up into 140 character chunks.  It’s hard to follow, and nobody will re-tweet an essay.
  • Adding a URL is not terribly harmful by itself — but it takes up space in your tweet.
  • Try to use one hash tag at most.

Remember, though, nobody re-tweets you because you’ve kept your length to 27 characters.  They re-tweet you because you have written something compelling.  Start with great content, and then use these guidelines to produce the very best tweet you can.  You’ll maximize your chances of being re-tweeted, and maximize your engagement with your audience.

The Limits of Automated Sentiment Analysis of Twitter

Posted on by

As part of my weekly analysis of the campaign race between Obama and Romney, I took a look at the sentiment of the tweets mentioning both of them.  That report will be up a bit later on today, but I wanted to share something interesting I’ve found while working on the report.

I did a manual sentiment analysis of the tweets to determine whether they are pro-candidate (or at least neutral) or anti-candidate.  The manual process consists of selecting a statistically valid sample at random and then reading each tweet to score it.  It’s slow, painstaking, but it produces the best results because I, as a human and one that is versed in current events, understand the text, the context, and can usually guess the writer’s intent.

But there are also automated tools for doing sentiment analysis, and I use these tools too.  These tools look at the text, do a simplistic parsing of the tweet, and assign a score based upon the words that are used: “Hate” gets a negative score while “Love” gets a positive score.  You might think that these tools are very crude, and you’d be right.  But that doesn’t mean they don’t deliver some useful insight.

Take, for example, last week’s (4/29-5/5/2012) tweets mentioning Obama.  Here’s how I scored them manually:

The margin of error is +/- 5% at a 95% confidence level (which is pretty much the gold standard for surveys).  How did the automated tool do:

Very close!  You’ll note there’s no “off topic” or “neutral” category in this chart.  That’s because a huge majority of tweets end up with a neutral score, and that’s really a failure of the scoring system rather than a real indication of indifference.  Still, both the 76% and 24% are within the margin of error of the manual survey, so we can say they produced the same results.  Fantastic! (Obama -Romney means I just looked at tweets that mentioned Obama but did not mention Romney — I didn’t want the score to be confused by negative-about-Romney tweets that mentioned Obama).

But does this always work? Let’s look at Romney’s sentiment as scored manually:

Mitt didn’t have that great of a Twitter week, although a lot of the negatives were other republicans not happy about losing Santorum and Newt, or people still supporting Ron Paul.  So a lot of the negatives are probably not supporters of Obama.

Let’s compare this to the automated sentiment analysis:

Quite different! Why? Mostly because although I tried to filter out posts that were really about Obama, there’s a lot of posts that snuck through without mentioning him.  #Julia and other anti-Obama hashtags were a common source.

The lesson here is that for automated tools you need to very carefully scrub the tweets being examined to make sure they are really on the topic you are interested in.  Or, put another way, because so many Romney supporters like to talk about how bad the President is instead of how great Romney is, it skews the automated analysis.

I’ll be trying to work in a better algorithm for analyzing sentiment, and I know others have made great strides in that direction as well.  But the key thing to remember is that automated sentiment analysis works best when the tweets are talking about the subject you’re interested in, and poorly when people mention your subject in passing while talking about something else.  When you see people give a score to a collection of tweets, you should generally assume that they are using automated tools rather than scoring them by hand.  And so you should keep in mind the limitations of those tools.

Update to Tag-O-Matic

Posted on by

I’ve released the next version of Tag-O-Matic, you can see the documentation and download information here.

Tag-o-Matic is a research tool for helping you to compose tweets.  The first version showed you the most popular hash tags for a given set of terms; the new version also shows you the most frequently referenced users for the terms as well.

Take a look at it, give it a whirl, and let me know what you think!