Twitter drives 4 times as much traffic as you think it does

Jonathan

Over the last few weeks, TechCrunch has run a couple posts using their own referrer logs to measure how sharing on various social services drives traffic. In these and other analyses based solely on referrer information, Twitter performs surprisingly poorly relative to expectations many of us have based on our own observations of the volume of link sharing on Twitter.

Does that mean the people you follow on Twitter who share links all the time are that atypical? Do most normal people just not click on links in Tweets? Is LinkedIn far more popular with the rest of the world than it seems to be with the people you know?

No, no, and no. There is a much simpler answer behind this disparity: referrers are a poor way to attribute traffic from social sharing.

Referrer analysis is based on the outdated metaphor of the web as a network of links between static pages that could only be navigated by browsers. Today’s web is built around social streams and other APIs that are consumed via dynamic web applications, desktop clients, mobile apps, and even other web services, all of which render referrers obsolete as an attribution mechanism.

awe.sm was built for the modern web — a network of people, not pages — to track the results of Tweets, Likes, emails, and other sharing activities no matter what path they follow. So our system knows with certainty where each link was originally shared in addition to all the places where it was ultimately clicked (i.e. referrers). This approach gives us a unique set of data that demonstrates just how misleading referrer information can be.

And in the case of links shared on Twitter, it’s very misleading: the referral traffic one sees from Twitter.com is less than 25% of the traffic actually driven by Twitter.

Twitter is the perfect storm for referral traffic

We looked at awe.sm data from the first 6 months of 2011 spanning links to over 33,000 sites, and the numbers were astounding:

  • only 24.4% of clicks on links shared on Twitter had twitter.com in the referrer;
  • 62.6% of clicks on links shared on Twitter had no referrer information at all (i.e. they would show up as ‘Direct Traffic’ in Google Analytics);
  • and 13.0% of clicks on links shared on Twitter had another site as the referrer (e.g. facebook.com, linkedin.com).


Twitter is the quintessential modern web service — all the ways to consume Twitter, even Twitter.com, are just clients for the Twitter API — so the failure to effectively track it using such an outmoded methodology as referrer analysis should come as little surprise. Twitter’s openness and the many resulting ways users interact with it are what have made it so successful, but they are also the things that have made its value largely invisible to publishers.

‘Direct Traffic’ explained

When a user clicks a link in any kind of non-browser client, from Outlook to a desktop AIR app to the countless mobile and tablet apps, no referrer information is passed for that visit and your analytics software basically throws up its hands and puts the visit in the ‘Direct Traffic’ bucket. The assumptions behind this fallback behavior show just how arcane referrer analysis is — if a visit didn’t come from another webpage (i.e. no referrer data), someone must have typed the URL directly into their browser address bar.

If you’ve spent the last few years wondering why the proportion of ‘Direct Traffic’ to your site has been on the rise, the answer is the growing usage of non-browser clients, especially on mobile. And since 2/3 of Twitter consumption is happening in desktop and mobile clients*, it’s safe to say that a lot of your ‘Direct Traffic’ is actually coming from Twitter.

How Twitter sends traffic through other sites

While the incredible growth of mobile apps and desktop clients and their importance in the Twitter ecosystem is news to no one, the value Twitter drives through content syndication is a bit more surprising: more than 1 in 8 visits driven by Twitter sharing are actually referred from other sites. Many other sites use Twitter’s API to pull in Tweets that they display on their own sites, where links in those Tweets are then clicked. For example, look at this screenshot of my LinkedIn activity stream. Notice that every update says ‘via Twitter.’ Yet when someone clicks on one of those links, the referrer will be linkedin.com, even though it only got to LinkedIn because someone shared it on Twitter first.

The same is true of Tweets syndicated to Facebook, About.Me, and myriad other websites that allow users to connect your Twitter feed directly. And because Twitter’s API is open and most Tweets are public by default, there are also many applications and sites that display Tweets based on hashtags, search terms, and other criteria without a user ever needing to connect their own feed.

In addition to the programmatic syndication of Tweets through Twitter’s API, sharing is fundamentally social and the human element is responsible for much of the serendipity that makes social media so powerful. A great example of that is this Tweet by @zeyneparsel, who only had 144 followers at the time. However, she happens to be a self-proclaimed “veteran hipsterologist” and this Tweet was on the subject of hipsterism (?!). As a result, the link contained in her Tweet ended up being included in a Psychology Today blog post on hipsterism (see UPDATE 3), which drove a significant amount of traffic.

In these cases, which showcase the amplification effect that makes Twitter so uniquely valuable to publishers and marketers, analyzing referrer data alone would attribute traffic to a variety of other sites, even though it all originated with sharing on Twitter.

Improving social attribution

Last week, MG Siegler noted that Google+ started rewriting all outbound clicks to come from plus.google.com. Facebook has rewritten outbound links for quite a while due to phishing/malware and privacy concerns. And both LinkedIn and StumbleUpon frame all external pageviews, which means you can see all the views they drive. As t.co rolls out to 100% of the links shared on Twitter (a topic we’ve previously covered in some depth), they may very well start rewriting all clicks on t.co links to show Twitter as the referrer. This would ensure Twitter gets the credit they deserve for traffic they send to publishers, but it would have the downside of obfuscating the diverse paths that a tweeted link can take.

Until then, it’s possible to correctly attribute visits driven from Twitter sharing by tagging your outgoing links using a solution like Google Analytics campaign tracking parameters. For example, the Tweet Buttons on Business Insider use links like this:

http://www.businessinsider.com/closing-bell-july-12-2011-7?utm_source=twbutton&utm_medium=social&utm_term=&utm_content=&utm_campaign=moneygame

Google Analytics can then properly attribute traffic to those buttons. Google Analytics offers a handy URL Builder tool, and other analytics solutions, like Omniture, support similar campaign tracking parameters of their own.

Why awe.sm is, well, awesome :-)

And if all you want is an accurate count of the aggregate traffic Twitter drives to your site, that should be enough. But our customers have found there’s a lot more value to be had in understanding the mechanics that drive successful sharing — who is tweeting, what they’re tweeting, where it’s being tweeted from, when it’s being tweeted, etc. So in addition to automatically building the outbound links to integrate our social attribution with Google Analytics, Omniture, and other web analytics solutions, awe.sm tracks the performance of each Tweet (and Like, etc) individually. By connecting the rich information we have about the context of each share with the visits, pageviews, conversions, and revenues it drives, we enables our customers to go beyond just looking at social data and to start acting on it (and to build cool stuff like this).

If you’re interested in learning more about how awe.sm can help your business harness the value of social, please drop us a line to questions@awe.sm or just click here to chat with someone from our team right here.

* The full list of sources of clicks with no referrer information (i.e. ‘Direct Traffic’) not only includes mobile and desktop clients, but also web-users who have https security enabled for their Twitter accounts (which strips out referrer information).

This entry was posted in Business, Featured and tagged , , , , , , . Bookmark the permalink.

52 Responses to Twitter drives 4 times as much traffic as you think it does

  1. Zack says:

    This is interesting, and I’m glad someone finally wrote about it at length. A few months ago I took over the Tweeting for our publication and noticed that the numbers in Omniture simply didn’t make sense when compared to the volume of our Tweets, Followers, ReTweets, etc. In early May I began appending tracking codes at the ends of all of our tweets and saw an immediate discrepancy.

    I’ve been keeping track, and knew the number was somewhere around 25%. After reading this, I just re-crunched: Twitter only shows as a referral domain on 27.4% of the tweets I send out (I figured this out by taking the total number of incoming traffic with the tracking code – which is only used on Twitter – and dividing the total number of Twitter referrals that contain that tracking code. If my math is wrong here, please chime in … this isn’t necessarily my area of expertise. Shear curiosity led me to try to bridge the discrepancy). This is remarkably close to your numbers, which is interesting, because I wouldn’t think Twitter would under report with any consistency. But looking at my weekly, monthly, cumulative numbers, and taking into account your figures, it seems to be a pretty consistent 24-28%.

    Another note is that, unfortunately, tracking codes are only a reliable way to track the tweets that I distribute. There is no way to get our readers to append tracking codes to tweets when they visit our site. So I’m afraid that calculating total incoming traffic from Twitter takes a bit of guesswork (quadrupling it, I guess). If you have any insight on this aspect, I’d love to hear it.

    • Jonathan says:

      Hi Zack, thanks for reading and for the great comment!

      Your methodology for analyzing your own data seems pretty sound. And the consistency between your numbers and ours makes sense because it is indicative of how Twitter is used in general (i.e. lots of Twitter users are consuming Tweets through clients, etc).

      As for tracking the Tweets by your readers, awe.sm integrates with standard Tweet and Like buttons (https://github.com/awesm/awesm-dev-tools/wiki/Share-Buttons) as well as AddThis, AddToAny, and custom sharing calls-to-action you might add to your site. While this doesn’t capture *all* sharing by your visitors, we’ve found that an increasingly high percentage of sharing happens through these buttons, especially if they are the standard ones from Twitter and Facebook that users are used to. You can also manually configure these buttons to use your own tracking codes in most cases.

  2. Peter Tanham says:

    Fantastically educational article Jonathan. “Direct” is always something that irked me in my GA profiles – and besides email sharing & IM, I couldn’t think of what else could be driving it so high – but Twitter clients (e.g. Tweetdeck) not passing referral info makes so much sense.

    This has cleared up something that has been bugging me for months – you’ve made my day!

  3. Matt Wolfe says:

    Here’s the only flaw in using a trackable URL shortener – the assumption that people are only accessing you site from the links YOU tweet. There needs to be a worldwide standard so that if, let’s say, Ashton Kutcher uses something other that awe.sm to tweet about your company, you can see that the traffic still comes from Twitter.

    • Jonathan says:

      Matt, you’re correct that these tracking methodologies can’t necessarily be applied universally. However, they can extend beyond just the links you Tweet.

      A large portion of the data we used for this analysis came from awe.sm-powered Tweet Buttons (see https://github.com/awesm/awesm-dev-tools/wiki/Share-Buttons#tweet) and other sharing tools that can be added to your site. While this won’t cover 100% of all sharing by visitors to your site, we’ve seen that use of standard Tweet and Like buttons (like the ones on this post) has dramatically increased the percentage of sharing by others that’s happening through means you can track.

  4. Pingback: Awe.sm Guys Say: Don't Believe Your Referrer Logs - Liz Gannes - Social - AllThingsD

  5. Susan Silver says:

    This is a great informative post. I didn’t know that Awe.sm was so robust in it’s tracking. I can see why Timely uses it as their default link shortner. I’ve trying to learn how to use GA better and the example you give above is a great way to do it. Thanks, Will spread the info!

  6. Pingback: Twitter bringt dir vier Mal mehr Traffic, als du sehen kannst » t3n News

  7. Paul L says:

    excellent write up. As more and more non-browser clients have been rising, I’ve also seen an incredible amount of “direct traffic” that just did not make sense intuitively. Lot of people are underestimating the value of social traffic… but it’s quite exciting to see how all the analytics around it will develop in the coming months.

  8. This was actually much easier to see before the Google/Twitter agreement on realtime expired on July 3rd. It had been possible to view mentions & links on many of the sites which cross populate with Twitter in Google Realtime. The cross population (sites which either feed Twitter or are feed by Twitter — somewhat dependent on how one has one’s profile preferences set) includes sites such as Friendfeed, Identi, Plurk, Pixelpipe, etc.

    In my opinion, the greatest loss while we await a revamped Google Realtime are Google Alerts which had been set up to monitor mentions, reTweets or links.

    When Google Realtime is reintroduced, it will, of course, include Google+. There are already a number of browser extensions & at least one use of OpenID to post from Google+ to Twitter. I haven’t yet seen a means to post from Twitter to Google+.

  9. Mike Cane says:

    >>>For example, the Tweet Buttons on Business Insider use links like this:

    And if I go and share that link via copy/paste, I rip out everything from “?” onward. So even if I hit a link in Twitter, then want to post it on my blog, that link gets cobbled down to its root. And I don’t think I’m alone in doing this.

  10. Meyer says:

    Now I can appreciate why my GA looks a bit off. One of many things to take with a grain of salt.

  11. Pingback: Twitter bringt vier Mal mehr Traffic als man sieht | Steadynews

  12. Pingback: Linkdump vom Do, 14. Juli 2011 bis Fr, 15. Juli 2011 Links synapsenschnappsen

  13. Pingback: Lesenswert: Tengelmann, Castaclip, Monte, Trolle, Google+, Save.tv, Twitter :: deutsche-startups.de

  14. Pingback: Traffic From Twitter Grossly Underestimated By Referral Analysis | The well-prepared mind

  15. Do you track links from only twitter.com or any other website/social media site as well?

  16. Pingback: If Social Marketing is So Big, Where Are All the New Visitors?

  17. Pingback: Twitter Drives 4x As Much Traffic As Google Says It Does [STUDY] - AllTwitter

  18. Thank you for these great explanations! I was always wondering why there’s so much “Direct Traffic” to my blog and sometimes i already thought not to tweet anymore because there are so few referrers from twitter. Your insights are the solution for this mystery and it’s absolutely plausible.

    Great piece of Work!

  19. Pingback: Are You Sure You’re Not Getting More Twitter Traffic Than You Realize? | WebProNews

  20. Pingback: 5 Key Lessons About Using Social Media in Your Business (Small Business E-commerce Link Digest – July, 15, 2011)

  21. Pingback: This Week in Social Analytics #7 at TweetReach Blog

  22. Simon says:

    .. just a few months ago i was researching the metric between the actual cost (man hours) of researching tweets, engaging followers, and the actual increase in lead generation and all the numbers over a 12 month period (6 after twitter) boggled me! over 300% increase in hard leads and full 5% con. increase for onsite transactions..

    great blog btw its awsome :p

  23. Jeff Hoffer says:

    I’ve also found that the pop up windows that occur when clicking a link on Twitter’s web client do not provide referrer URLs which would lend itself further to the Direct Link inferred data.

  24. Pingback: links for 2011-07-15 : The ChipCast || by Chip Mahaney

  25. Pingback: Daily bookmarks & places archive | Chipcinnati

  26. Pingback: La mentira del tráfico que viene de Twitter

  27. Pingback: links for 2011-07-16 | Läsvärt | Attefall Allehanda - Nyheter och recensioner

  28. Pingback: The Saturday Summary – 7/16/2011 — The Crossing of Marketing and IT

  29. Nice to see this blog post being shared widely. I realized this was an issue back in 2009, in the same that visits from email programs such as Outlook are lumped into the DIRECT visits bucket. I helped a client set-up an automated process to tag all URLs posted from his blog to Twitter. We discovered that Twitter visits were being under-counted by a factor of 8!

  30. Pingback: Twitter dirige 4 veces más tráfico de lo que se piensa | Clases de Periodismo

  31. Pingback: Voce Monday Morning Five | Voce Communications

  32. Pingback: Twitter Drives 4x as Much Traffic as You Think. Here's Why … | TechCrunch

  33. Pingback: Are You Sure You’re Not Getting More Twitter Traffic Than You Realize? Analytics programs may be getting it wrong

  34. Pingback: Quanto traffico porta Twitter?

  35. Pingback: The secret to Twitter success – it’s all about sharing links

  36. Pingback: The secret to Twitter success – it’s all about sharing links

  37. Brilliant post! I’m new to GA and have always wondered how my website ended up with so much direct traffic but less than 10% referrals from Twitter every time I tweet about a blog post.
    Now I’m going to go back to the stats and see what I can glean from the figures based on your blog post. Thanks!

  38. Pingback: The secret to Twitter success: Link sharing. - Twitter

  39. Pingback: The secret to success on Twitter: Link sharing.

  40. Pingback: Getting Mugged by the Huffington Post

  41. Pingback: Tracking Social Media with Web Analytics

  42. Mark McLaren says:

    The Google URL Builder tool is okay for a quick URL tag solution on the fly – better than nothing at all. But there is a really nice free URL tag Campaign Manager tool over on the Analytics Pros website.

    It’s an Excel spreadsheet that does what the Google URL Builder does AND it makes it easy to keep all the URLs you create in one place for future reference.

    Consistency is the key to tagging URLs. If more than one person is working on an online campaign, you need a way to keep track of who used what URL and what tags they assigned to it. You can share this spread sheet with everyone on your team, and use a URL shortener to create sets of URLs for a campaign.

    Full disclosure: I work with Caleb Whitmore, owner of Analytics Pros. But the spread sheet is completely free and there are no strings attached to the download.

  43. Pingback: Twitter Referrals Easier to Track in Analytics Now | WebProNews

  44. Nathan Zaru says:

    Hey guys, I had an interesting finding recently and I thought back to this post. We had an event that lasted a couple hours and the referrals from non browsers applications was 73% of overall inbound traffic. Is this on par or higher with what you have seen?

  45. Vikki Fraser says:

    Excellent post. Thank you for some clues as to the inordinate amount of Direct Traffic I receive!

  46. David B says:

    You just answered something that’s been bugging me for months – thank you.

    I keep hearing about how Twitter drives so much traffic yet a browse through my Analytics reveals Twitter as being one of the more ineffective referrers. I was also never able to put my finger on the fact I get so many “new visits” showing up with a referral source of “direct”. How are these people hitting my site directly if they’ve never been there?

    I think you solved the mystery.

  47. I’m wondering if this is true, as you know Twitter has automatic the no follow attribute, so there is a less change that you will have some traffic in the term of backlinks from Twitter. The most of the tweets are timestamp what means that it is for a short time, if the tweets is not reading in one hour, the tweet is history. Another thing is that Twitter has since half 2011 a filter what means that Tweets with a link will not be visible in a timeline, specially in the search box unless you have more than 2000 followers and you follow a couple. I have test this out a couple of months ago, and it seem to be that blogging and write good content is still the winner specially in Google where the most traffic will come from

  48. Pingback: Tracking ‘Dark Social’ – Uncovering the true value of your social media traffic | Our Socially Connected World

  49. Pingback: IBM’s Black Friday study got it more than a little wrong | awe.sm | Blog

  50. Pingback: Is it Better to Share on Google+, Facebook or Twitter? | Seo Marketing Biz