Does This Blog Get More Traffic From Google or Twitter?

I made the assertion last evening in Seattle that this blog gets more traffic from Twitter than Google. Mike Arrington called bullshit on that statement. The interchange is about 1:50 mins into this video:

Fred Wilson at Naked Truth Seattle 2009 from Jason Preston on Vimeo.

Thanks to Jason Preston for capturing and uploading it and sending me a link to the video. The transcription (courtesy of Simulscribe video transcriptions) is here.

I offered to share my refer logs with Mike and I might as well share them with all of you as well.

Avc refers june & july

These are AVC's refer logs for the past 30 days. It says clearly that Google drives more traffic than Twitter (14.8% to 8.2%). However, some percentage of the direct (direct + avc.blogs.com) visits are coming from third party twitter clients and URL shorteners like bit.ly.

To demonstrate this, here are the refer logs from the same 30 day period one year ago.

Avc traffic june 2008

Last year, this blog ran at avc.blogs.com and there was no avc.com, so the direct traffic was 22%. Now it is 39%. Some of that increase is simply more people typing in avc.com into their browser. But certainly some of it is traffic coming from third party twitter clients and URL shorteners.

If just 7% of the 17% increase in direct traffic is a result of twitter links that are not being counted as twitter, then it is true that this blog gets more traffic from twitter than google.

I wish I could be more scientific about this. Maybe the bit.ly guys can help. My assertion last night was based on data plus gut instinct. That's not going to be enough to satisfy Arrington I'm afraid.

Reblog this post [with Zemanta]
#Weblogs

Comments (Archived):

  1. Jan Schultink

    It would be interesting to see the difference for your “latest blog post” and older “archive” posts.

  2. Anonymous

    Does bit.ly not feature as a referer?

    1. fredwilson

      not in google analytics

      1. Adrian Palacios

        fred, since you use google analytics you can shorten a tagged url in bit.ly. seriously, it’s too many steps, but if there’s any community to ask for some programming shortcut, or if someone knows where a shortcut was already made, this is it…1. copy the link you want to shorten… let’s use this post for example (so: http://www.avc.com/a_vc/200…2. go to the analytics URL builder: http://www.google.com/suppo…3. enter the URL you copied into the “website url” field.4. change “campaign source” to ‘twitter.com’ and “campaign medium” to ‘tweet’ and “campaign name” to, uh, whatever. really, you can put whatever you want in these fields. in the way i’ve outlined, in analytis under traffic sources it will appear as “twitter.com / tweet”5. click “generate url”6. copy the url that was generated (it would be http://www.avc.com/a_vc/200… for this post)7. assuming you’re using tweetdeck, just paste the url you just copied, shorten, and BAM! you’re done and you will know *exactly* what kind of traffic twitter is driving. do this for a 30 day period… heck, maybe just a *week*… and we’ll all know your google vs. twitter dilemmaokay, i just realized… if you add this to the end of every url that links to your blog:?utm_source=twitter.com&utm_medium=tweet&utm_campaign=avcyou won’t even have to mess with the url builder, etc.who knows javascript? can anyone make a button that would automatically append this to the end of a link fred wants to shorten? we can settle this debate pronto 😉

        1. ShanaC

          As great as this sounds- the data needs to be totally raw. Little to no SEo, very little linking in. You want to start empty or nearly empty. This is not controlled.

        2. fredwilson

          I’m intriqued and may do this but will I have to do all these steps every time I tweet out a post?And what about when other people tweet out one of my postsThe only way I am going to do all of this work is if goog analytics will capture all my incoming twitter traffic correctly

          1. Adrian Palacios

            when i finished writing my comment above, i realized there is only one step involved: add ?utm_source=twitter.com&utm_medium=tweet&utm_campaign=avc to the end of a link before you shorten/tweet it. somehow keep this little line of code handy, so that, for instance, your tweet this morning to the new blog post is not just[etc etc]the-fine-line-between-informing-and-spamming-your-followers.htmlrather, it should be[etc etc]the-fine-line-between-informing-and-spamming-your-followers.html?utm_source=twitter.com&utm_medium=tweet&utm_campaign=avcthis will also capture all the people who retweet your shortened link. the only problem is when *someone else* tweets a post of yours; most likely they will not append the utm info to the link, and thus that data will be lost :(the only thing i can think of for shoring up that missing data is the tweet & track link beneath every post; when i click on it, it automatically grabs your post’s url and shortens it… if there is some way to append the “?utm_source=twitter.com&utm_medium=tweet&utm_campaign=avc” to the end of your link before bit.ly automatically shortens it, that would take down the percentage of “missing” data even further.so, it’s not perfect, but it should get you pretty close. now that i’ve thought about this a bit, i see how difficult of an issue it is for bit.ly to add analytics tracking to their service!

          2. Adrian Palacios

            dang it; i just looked at the “reactions” to this morning’s post (“fine line between informing and spamming your followers”); the count of people who use your bit.ly link is less than those that either create their own, or use something like tinyurl, tr.im, etc.sorry, thought the solution i provided above would be good, but this is a lot harder than i originally thought :-

          3. ShanaC

            Could you partner with Some store selling clothing- two different locations- and create some sort of blog where every time someone buys an outfit, the store sends out a tweet on the person’s behalf with a photo with the option to log a comment back in about the outfit?But only at one location- so that the other isn’t generating traffic and can be used as control variable for say merchandise and sales comparisons?A) you can see if it drives up sales at location a versus location bB) you can see how far the links are passed via comments left? or tweeted back? Or traced? Or all of the above?Is this doable? This blog in some ways isn’t ideal because it has too many followers, you want something clean for the sake of analytics and something where it can be compared to a similar if not the same, situation. It’s very hard thinking up a way to track a tweet without linking it to some other behavior that is also traceable, and also can allow itself to have a control variable.

          4. fredwilson

            That’s the problem. We need to solve this at the analytics side

          5. Adrian Palacios

            fred, what do the actual server logs say? i’m not clear behind how these 3rd party apps work, but maybe these apps rely on a particular server?we recommend javascript-based page tagging for all our clients; it just makes sense. we recently gained a client that was relying on free software that analyzes server logs, though. for our purposes, the information was terrible. however, i noticed that instead of saying (making this example up) “x visits from sf.curbed.com” it would say “page2.syndication.com” (or whatever the heck google ad manager uses), or, it would would say “doubleclick.net” instead of the site the ad was served on…maybe you could check the server logs and see if traffic is showing?whoa…i never thought in my life i would be curious to see what data from the server logs say….

          6. fredwilson

            Whose server logs?

          7. Adrian Palacios

            whatever provider hosts avc.com

          8. fredwilson

            Fuggedaboutit.Typepad’s analytics are awful

          9. Adrian Palacios

            Fred, still been thinking about the issue of tracking Twitter traffic (say that five times fast…). I noticed that on the browser-based service, they track outgoing links with Google Analytics pageTracker._trackPageview function, which is great. However, a lot of the user apps utilize Adobe AIR; I am not a programmer, but maybe there can be some requirement to use the getUrl command? A lot of the publishers I advertise with use that somehow to track clicks.Anywho, just a thought. Still not perfect since it is basically publisher-side only tracking though…

          10. fredwilson

            That is the next step. They wanted to get click tracking working at twitter.com first. Not sure how they plan to implement it on the clients

  3. Ivan Kirigin

    As I mentioned on twitter, you can see this in Real Time™ http://bit.ly/wBQVX on your chartbeat144 total73 direct25 from twitter8 from googleBut you just posted a tweet to this link…

    1. Ivan Kirigin

      This made me think of a test you can run to estimate the direct traffic original source. Just look at the graph in direct traffic before, during, and after the bump you receive from posting a tweet to the blog. The absolute value of the organic direct traffic should remain roughly the same, so any increase in direct traffic is actually twitter direct traffic.

  4. disqus_v4XrgpXgPt

    Fred…if one data point would help – I found this post via your Twitter post and not google 🙂

  5. stanleyyork

    man arrington is quite the little thorn lol

  6. John Sharp

    I don’t see quite the amount of traffic that you see Fred, but I’m seeing the same phenom you mentioned in Seattle – in the past six months Twitter has come from behind to now be the largest source of visitors to our site, and it now provides 50% more referrals than Google.Also kind of interesting (at least to me, anyway), the Twitter referrals stay longer and view more pages.

  7. timraleigh

    I think you may be right Fred. While I subscribe to this blog via email, I got to this blog entry through a link on a comment on twitter by someone I follow. Based on my reading here (I could be wrong) it looks like Google analytics wouldn’t catch that traffic.

  8. marshal sandler

    When I check Analytics on my blog which is not a work of art Twitter Wins- Mr Arrington obviously a smart man, but the real bullshit is he is always playing gotcha ! http://en.wikipedia.org/wik… I think stats show that Twitter really has Gotcha- All Of Us -Shout out for Twitter a winner that can cut bait and fish !

  9. Dan Lewis

    The trade-off between Google (referral) and Twitter should be pretty high. The former is probably Google Reader/iGoogle-driven visits, which have likely ceded in high proportions to Twitter (myself excluded).

  10. Jason Preston

    More important than the actual data split is the point: Twitter is a large & rapidly growing source of traffic for avc (and therefore probably other sites as well).Glad the video was useful 😉

  11. niallsmart

    The relevance of Twitter as a referrer is obviously of huge interest (as presumably, they will explore the monetization options for this traffic flow). The % split versus Google is only part of that picture though – I’d also be interested to see the conversion rates for referrals via Twitter vs. Google (admittedly, this isn’t a relevant metric for a blog). Intuitively one would expect that referral via a friend’s tweet represents a stronger introduction/endorsement and is therefore more likely to lead to a transaction (c.f. http://peashootapp.com)Another consideration in this conversation is that leveraging Twitter for customer acquisition can be a lot easier than SEO – particularly if you’re in a position to use a ‘demand harvesting’ model (http://startup-marketing.co…. We actually drove more traffic from Twitter than Google search in the early days due to a fairly active Twitter campaign – i.e., identifying potential customers based on Twitter keyword searches and @replying to them. Using Twitter search in this respect is actually the inverse of Google’s search – you go find your customers, rather than rely on them finding you.Right now the tool support for this approach is immature and it’s not yet clear if the approach will scale, or whether unfiltered @replies will make Twitter too spammy – but it worked for us.

    1. fredwilson

      Great comment. We need more data and twitter needs to build in some governors to ensure spam is not useful

  12. Mayank Sharma

    You have to be careful while including statistics from bit.ly or other such link tracking sites as they count url lookup and not exactly a site visit. http://zebugroup.com/blog/2

    1. fredwilson

      I’m not counting them at all (yet)But that is useful info. Thanks for sharing it

  13. Danny Sullivan

    Twitter is definitely being undercounted. I did a piece on this earlier:http://searchengineland.com…I have a follow-up coming, but it might be undercounted from 80 to 1500% by analytics programs. Seriously.

    1. fredwilson

      So what’s your gut based on the logs I posted danny? Google or twitter sends more?

    2. fredwilson

      danny – i added your post at the bottom of mine as a related link. this just points to the fact that twitter has got to start providing some analytics tools.

      1. Danny Sullivan

        Thanks, Fred! In terms of my gut, Twitter’s certainly getting up there. I’d say it probably is exceeding your Google Organic traffic due to undercounting that Google Analytics is doing.A side issue is that your Google Organic traffic shouldn’t have dropped that much. Normally, I’d expect it to stay steady month to month or rise. The drop of 10,000 suggests you may have had a particularly good month a year ago. Trend would be nice to see.But still, I’d expect that same trend would show Twitter also rising month after month — and double your figures would be conservative in the “real” traffic that Twitter is probably sending.Hope to have more on this next week.

  14. RacerRick

    It’s difficult to figure out all the traffic that originates via a link in a twitter stream.Certainly Twitter, and Facebook, will eventually blow by Google for referrals.Plus it’s more valuable because that link came from someone trusted.

  15. Brian Kung

    Why not just stop RSS feed for your next post and see how the numbers compare, or vice versa?

    1. fredwilson

      I don’t want to mess with the readers who use rss like that

  16. ShanaC

    Well, how about this.Is there a way to make two or three posts that are about absolutely nothing, as a test.. Somewhere else, that is not this blog. Promote the test elsewhere, and give the links via the following methods. Encourage people to comment if they receive the posts.Make one only available originally via twitter. No Seo, nothing. Clean slate, just twitter. Include as a variable bit.ly. Make a third clean one if you want to test that..Third post is also empty in content, but otherwise its structure underneath looks as usual.Release all at the same time, and track the analytic.Something like this will cause a baseline, I think. Either that, or Mike Arrington will be proved right. But you will have some data.

  17. Nate

    Arrington is an ass. His latest thing is painting people with the “you’re compromised by self interest” brush, which is just his own self serving method of goosing drama so people talk about him.Anyway, even if Google sends more “hits”, the quality of traffic from Twitter is much higher. When I click an AVC.com search result in Google it’s because I want some nugget of information and then I’m gone.When a trusted Twitter friend says “Read @fredwilson’s latest post on X because it’s awesome” I’m far more likely to A) follow @fredwilson and B) read Fred’s post and spread it to others.

  18. paramendra

    Fred. No wonder Mashable beat TechCrunch. Michael Arrington “gets” social media, but maybe not all the way. http://technbiz.blogspot.co

  19. kidmercury

    no worries boss, this is just another piece of evidence to prove once and for all that you’re superior to mikey. you’ve dropped truth, he’s blowin’ smoke. damn.

  20. Joe Lazarus

    Bit.ly should be able to give you a rough estimate on what percentage of their traffic to AVC is currently captured in your referral logs. I started to dig into some of your data using the Bit.ly and BackTweets API’s, but it was messy and I gave up. I assume it’s fairly easy for someone at Bit.ly to query their database to find all the links that resolve to avc.com and then compare their referral stats for clicks from twitter.com (and facebook, etc) on those links to clicks from “Email Clients, IM, AIR Apps, and Direct”. You could then use that ratio to estimate the portion of Twitter traffic that’s currently logged under “direct” in Google Analytics. It’s not exactly what you’re looking for, but it’s probably pretty close. Anyone can get all that data using the BackTweets & Bit.ly API’s.

    1. fredwilson

      Thanks JoeI’m gonna ask kortina to do this for me

  21. Yule Heibel

    Off-topic to analytics… I haven’t been able to watch the whole stream of the Naked Truth event, but yesterday I did catch the beginning, before the event itself started. Brian Westbrook interviewed Mike Arrington, impromptu – interesting exchange. I’m not familiar with Arrington and have only heard peripheral stuff about him (mostly negative, which I guess is what makes it to the top, haha).Anyway, Arrington did something really funny. He pushed back at Westbrook, demanded that the camera scan around to film the bar set up nearby. Then he claimed that everyone in Seattle seemed to be into boozing it up, whereas in SV – he claimed – everyone is way too busy doing business to bother having drinks: people in SV are serious, and people in Seattle are …not.Ehhh???That’s when it hit me: cognitive dissonance. Like an expert tennis player, Arrington tosses the ball into …well, a court you didn’t quite expect.Here’s what I mean: the rest of us who aren’t native or naturalized SVers, WE have our noses pressed against the shiny window that is the fab techno culture of that place, and it includes a spectacularly sybaritic mythology. For example: gourmet cafeteria meals at Google, massages, lunches at Il Fornaio (or wherever), flashy cars, shiny gadgets, you name it.So what does Arrington do? As those already outside the charmed circle get ready to have a bit of dipsomaniac fun, cut loose and have a glass of locavore wine, he tells them that they are decadent hedonists!Whoa! Cognitive dissonance!Sure made me think… Until I uncorked the next bottle, that is…

  22. Lloyd Dewolf

    Thanks for this discussion. That bit.ly did not set itself as the referrer when none is provided is an interesting insight in itself for me — makes sense, but the other way probably would have as well.Danny Sullivan comment and linked article are also particular interesting to me.It seems that our tooling needs to be updated to identify this type of traffic with it becoming more prominent, because this non-web referred traffic doesn’t really fit in the “direct traffic” bin either. It wouldn’t seem to be that hard to fix in the twitter client and browser integration to hack into the referrers.The easy hack would be to use a shortener that does provide itself as the referrer, if one isn’t present. I bet http://code.google.com/p/urly/ could be hacked to do this very quickly (if it doesn’t already)

    1. fredwilson

      I think its really incumbent on google and all the other analytics providers to do the work to identify twitter traffic (and fb traffic too btw) correctly

  23. Melcrum

    I understand that Twitter is growing as a traffic source, but why do you think you are losing so much Google traffic (down 12,000 referrers from the same period a year ago)? It seems that the drop in your Google traffic is as interesting a story here as the rise in Twitter referrals. In the logs for the sites I manage, I’m seeing a boom in Twitter referrals for a lot of sites but haven’t seen the drop off in Google referrals that you have.

    1. fredwilson

      It may be due to the move from avc.blogs.com to avc.com. We followed best practices in an attempt to keep all of the google juice. But we may have lost some along the wayIts ok with me as most of the google traffic to this blog isn’t very high quality

  24. kortina

    Fred, I just recorded a quick screencast showing how to dig into the Direct traffic reports in Google Analytics to understand them a bit better: http://bit.ly/23Y7e7 – you and the other commenters may find this useful.The campaign tagging in Google analytics mentioned by some others is a great way to further segment out this stuff, but it does require some up front work. The method I show in the demo is less accurate, but useful if you want to estimate and have not been using utm vars.

    1. fredwilson

      Kortina – thanks so much. I really appreciate you doing this work. I twittered this post this morning

      1. kortina

        No problem. This stuff is tricky. I’ve been digging pretty deep into thisanalytics stuff for awhile, though, so I feel a bit responsible for tryingto help demystify it. Glad the vid helped.

  25. ilan peer

    i liked hearing Etsy conversion from social referrals match all others (google)cut on adwords, hire a social editor..etc.

  26. Jamie Lin

    Fred, since you have 40+ pcnt readers who use RSS clients, including the most popular Google Reader, to access your content, you might want to discount Google Reader referred traffic — as these aren’t the result of SEO. If you take that out, I wonder if Twitter is already bigger than SEO.

  27. wdg162000

    Isn’t it 31% direct, not 39%? Meaning that you need 7% of the 9-10% increase to be twitter in order for twitter traffic to surpass google.

    1. fredwilson

      Its 39pcnt because avc.blogs.com is direct too

    2. fredwilson

      Its 39pcnt because avc.blogs.com is direct too

  28. wdg162000

    sorry posted twice

  29. Uday Sreekanth

    It would be nice to segment the referrals by post also, for the more recent posts, to see possible topic influence.

  30. Mayank Sharma

    Fred, I discovered a very simple solution to accurately measure traffic coming in from twitter using bit.ly but with a simple hack. Maybe bit.ly can provide this as a feature to get correct results. Let me know if you want to try it out. Details at http://bit.ly/L1ys4