Traffic Sources and Attention

There’s been good debate around how the source of traffic to sites is changing, shifting from the search engines to social sites such as Facebook and Twitter. I confirmed that I too am seeing a greater percentage of traffic come in via links shared on social sites and shared a colleague’s theory about what this would mean for Google’s advertising revenues. Fred Wilson also posted about this topic here and here.

What about attention? How does the average visitor from a social site compare to someone from a search engine? Niall Kennedy tweeted the following stats from his referral logs:

  • Twitter: 7 seconds
  • Digg: 20 seconds
  • StumbleUpon: 40 seconds
  • Facebook: 52 seconds
  • Delicious: 82 seconds


Here are my stats for the past year:

  • StumbleUpon: 40 seconds
  • Digg: 42 seconds
  • FriendFeed: 53 seconds
  • Facebook: 60 seconds
  • Twitter: 86 seconds*
  • Delicious: 110 seconds
  • Techmeme: 114
  • MyBlogLog: 176 seconds

Compared to the attention span of those coming from the major search engines we get:

  • (MSFT): 21 seconds
  • Yahoo: 35 seconds
  • Google : 40 seconds
  • Ask: 46 seconds

It would be interesting to see figures from other sites, especially online shopping sites which are the ones most interested in getting (and therefore likely to pay for) traffic. While it’s clear that visitors from social sites are more engaged with my blog because they tend to hang around a bit longer, that may not be the case with a shopping site where there is less intent to purchase than if they come from a search engine but Mark Essel thinks otherwise.

Increasingly, the flow of web links is being made between individuals via social media sites.  Your good fishing buddy who knows the Bay area, shares a link to his favorite supply store.  As focused communities become populated across geographic barriers, local quality referrals become more likely.  But what if you want to know what store fishermen prefer in San Francisco?  You could simply use twitter search for fishing san francisco.  In real time you could send a message to several individuals who are interested in fishing in that region.

Shared links are compelling but they need to be matched with impulse buying or discoverable when you’re looking for it. I’m thinking of O’Reilly’s flash discount shared via twitter (44% off to celebrate the 44th president) which was effective in bumping registration at the recent Web 2.0 Expo. The other way to generate business via shared links is to make them searchable so you can find what you need when you want it – but then we’re right back at sending traffic via search again.  Yes, you can search twitter but you can find this stuff on Google too.

The jury’s still out as I think this will be a slow shift of behavior that will take a long time to impact existing business models. The real prize is back to social search which would combine the best of the recommendation trusts of social networks with the ability to find what you need when you’re looking for it.

Facebook recommendations married to Google ‘s structure and ranking? That’s the subject of another post.


* I dug into the Twitter figures because they’re so out of whack with what Niall is seeing and it looks like there are a few visitors that hung out for a long time that are pushing that average higher than it should be.

Reblog this post [with Zemanta]
Current Events

AdSense, Self-Optimized

google behavioral targeting options

It’s obvious when you think about it. Instead of spending your energy throwing up hundreds of ads that dance around the edges in the hopes that one will magically trigger a random click of interest, why not ask your readers, “What do you want to see?”

Google announced a new program which changes the way they pick which Adsense ads appear on the websites you visit. Instead of looking at the content on the page and simple IP-based geo-targeting, the new program looks at your browser history and targets advertising based on your browsing interests. It’s known in the industry as behavioral targeting but Google has re-labeled it “Interest-based advertising.”

The system works using tracking cookies which is anonymous and tied to your browser so if you switch browsers or jump to another PC, your browsing history will not follow you (although technically it’s possible if they tied the cookies to your Google Account). Likewise, your browsing history is going to get muddled on shared PCs such as one you might find in a family room.

The most interesting thing about this new program is that they are allowing you to edit your profile using a new Ad Preferences Manager. As you can see in the screenshot above, you can pick and chose from 20 major categories and 600 sub-categories that are interesting to you and begin to shape the type of ads that are served to you. There is an option to opt out completely in which case you’ll get the regular content matched advertisements we’ve been seeing all along but having the ability to customize what you see is a bold step forward in transparency.

Arguments against such a system when I had brought it up in the past were that if you let everyone pick and chose your ads you run the risk of not serving a well-balanced mix of advertising, running dry in popular categories and having a glut of units from less popular categories. By putting direct user feedback into the equation, you could no longer tune your ad servers to optimize for maximum profit.

My counter to that argument is that by putting your readers in charge of what they see, you stand a much better chance of not only having people look at your ads (to see what type of ads get served) but also getting a much better sense of what interests your readers.

So here’s my crazy idea. Why not go a step further? Why not let people to search for ads directly? Every magazine has an Advertiser’s Index in the back, why isn’t there a web-based equivalent? It’s been done.

For more detail and commentary, Barry Schwartz has an extensive write-up on this development over on Search Engine Land.


Facebook, Twitter send more traffic than Google

Liz Gannes posted that Perez Hilton is now seeing more traffic coming in via Facebook than Google.

My colleague Udo Szabo at Nokia HQ in Finland has a theory that I call the Unified Theory of Interweb Economics. The theory goes something like this:

  1. Advertising is a function of your traffic volume, the more traffic that comes to your site, the higher rates you can charge.
  2. Social sites such as Facebook and Twitter have a lot of link sharing going on as friends post links to share them with each other.
  3. When Facebook and Twitter send more traffic than Google search referrals, advertising dollars will follow the source of that traffic.
  4. Google’s dominance in online advertising will be threatened.

One would think we’re seeing the first hints of that with and it will take a long time before we see this trend across the board. But in hindsight this is obvious and looking at my stats for the past month I find that 44% of my visitors are coming from search engines and 45% from referring sites with a vast majority coming from StumbleUpon, a social site for sharing links.

The more you think about it, the shift in balance of power has already taken place.  SEO is still a big industry but now there are companies that will help you with SMO (“social media optimization”).  On twitter we’re starting to see scam artists try and insert themselves into the conversation just as they used to do with Splogs.

And so the wheel turns round once again and there appears to be a new king of the hill – AOL Keyword > Yahoo Category Link  > Google Keyword Ranking > Social Site referral.

The jury’s still out as to which social site will send you more links but it’s looking more and more like Facebook. We used to have the Digg effect but no one talks about that anymore. Twitter certainly has the ability to send you a bunch of traffic in short order but the audience is still mostly the early-adopter set and any traffic will most likely be short-lived as the re-tweets scroll into the past.

Facebook, with 175 million users, certainly has the right broad-based distribution to broadcast a link and send back traffic. They have added new features such as FriendFeed’s “like” feature to make quick sharing easy. The Facebook Connect and Comments widgets also will help insert new links into Facebook for redistribution.

But while it’s great to get an influx of new visitors via a social site, it’s no good if it’s not a lasting reference. Has the pendulum swung to far the other way? There are some searches that work great on the real-time web, the latest viral video comes to mind, but others such as the listing for your local plumber just don’t work on something like twitter. Not that it’s worth anything but I still get a regular influx of traffic because I’m the #2 listing for insulting british slang even though the post is over three years old.

I think it’s fair to say that the social sites such as Facebook and Twitter will erode Google’s monopoly on online advertising. Google’s never been really good at building social sites and have shared in the growth of social networking in the past by providing the advertising engine for the most popular social sites. What happens with these sites start to go direct to the advertisers? Will Facebook advertising revenues come from traditional cost-per-click/auction model or is some other type of model required to succeed in a social networking site?

Current Events

Do Social Gestures a Business Model Make?

Is twitter a directory or a utility? This is the question that Charles Hudson raises in his post The Database of Intentions is More Valuable than the Database of Musings. While investigating prospective business models, he raises good questions about the ability of a collection of “accumulated musings” to determine intent which is what is most valuable to advertisers.

But maybe advertising  is not the great revenue driver of the next generation of startups after all, at least not advertising as we know it. Maybe it’s just me but I feel a need to make sense of all the stuff we share with each other. There seems to be value in tapping into the pulse of the “now web” but the methods of pulling meaning out of the noise seem crude. Keyword searches? Is that the best we can do?

Something went wrong with the Intense Debate comments on last night’s post on Keywords and Meaning. It’s unfortunate because there were some really thoughtful responses to the post which I’ll repeat in this post because they are worth reading.

Todd writes:

Keyword extraction from Twitter could be cool, but may kill of serendipitous discovery, my favorite aspect of Twitter. If keywords or meta-categories are predetermined truly unique hawtness, unprecedented new things ( a Twitter specialty ) will just get deleted? That would be FAIL.

I wonder if more of a “people with attributes” are really what’s needed. Example, I do want to know what’s going on with the latest developments for Symbian operating system, particularly activity streams and address book stuff. Rather than rely on keyword extraction, I could just assign an attribute to your tweets…


…I can be fairly assured news filtered by real humans, THEN assigned an attribute of my choosing will bring me some good results. A tag cloud of all tweets containing “symbian, activity stream, address book” would be noisy ( pollute with people asking each other for tech support? ), difficult to pull meaning from while drinking beer at my favorite bar.

Jonathan Strauss writes:

The TechCrunch post you cite was inspired by John Borthwick’s very interesting essay on how Google’s approach to content filtering breaks in the realm of what he calls the ‘Now Web.’ Like you say above: “Google’s PageRank, while valueable in sorting out the reputation and tossing the hucksters, is no good when applied to real-time news which is too fresh to build up a linkmap.”

In the (relatively) static web, the network nodes are pages and the endorsement actions are the links between them which are effectively permanent as well as public, and thus crawlable. In the Now Web, the network nodes are people and the endorsements are ephemeral share actions, the majority of which are not public or crawlable (i.e. email, IM, Facebook — what I call the ‘Deep Now Web’). And so, authority also takes on a different form from the aggregate view that PageRank provides to the personal measure of how much influence an individual has with her social network on a particular topic at a given moment.

I agree that we need to have a means of systematically capturing the newly important metadata of share actions and that it needs to be done at the point of sharing (see Jeff Jonas). But, I believe the more easily adopted (and thus ultimately more useful) taxonomy will be one of contextual metadata (i.e. who/what/when/where/why/how) rather than the more personal folksonomy/tagging approach you suggest.

There was also reactions via twitter from Kevin Marks:

The act of sharing links, photos, or other metadata on social networks is an action, to a certain extent, that gesture is more interesting than the actual data itself. The fact that my usually dormant cycle racing friends are now extremely active on twitter these past few days as the Tour of California is on is as much an indicator of interest as the actual substance of their conversation.

Keywords are part of the picture – the complete context around who/when/where/why/how are just as important as the tidbit of data itself. The meta-data contains more clues than the data.

The cellphone is a rich source of meta-data which can be captured at the source, the moment of sharing. Feeding contexts captured from the cell phone would be a great way to add context to any act of sharing.  There are privacy concerns and ownership questions. There needs to be a real value demonstrated to the potential user before they give up some of this privacy. But that’s a topic for another post.

Reblog this post [with Zemanta]
Current Events

Keywords and Meaning

TechCrunch asks if twitter search gets us closer to being able to mine the world’s collective thoughts. We may be getting there as millions text their latest thoughts into their cellphones. With a simple text message, the hive mind has the potential for 4 billion nodes out in the real world (for comparison, the human brain has 100 billion neurons)

News junkies of the world turn to twitter as the latest source of raw, unfiltered information. Peering over the shoulder of various members of the House and Senate who twitter is a unique view into our government. What you see is a more intimate, human view of the people that make the news. Yet, how do you harness that noise and turn it’s output into information?

Twitter follows a long line of services which break through editorial filters, get at the source of a story so you can make your own judgements. Blogs occupied this space just a few years ago and real-time indexes such as Technorati rose to prominence as a way to get a jump on the news.

Sidenote: Alacra, admitting important news about companies breaks on the web, is launching Pulse which applies their analytics engine to extract company names from their hand-picked collection of 2,000 RSS feeds.

The need for speed is nothing new. Former Wall Street Journal newsman Craig Forman draws an arc that extends from the pigeons Baron Reuter used to deliver news of  Napoleon’s defeat at Waterloo to the real-time newswires used in the financial world today . If there’s a way for someone to profit from the knowing something before anyone else, there’s always going to be people looking for a way to get at a scoop and others looking for a way to deliver.

We want to look to twitter for the scoops but we are doomed to learn the same lessons as we have in the past about authenticity. What we gain in speed and convenience, we lose in validation and measured fact-checking. Google’s PageRank, while valueable in sorting out the reputation and tossing the hucksters, is no good when applied to real-time news which is too fresh to build up a linkmap.

Working for Dow Jones in Tokyo, I would work with bankers and reporters who would use digital newswires to deliver them the latest news from around the world. As a systems engineer setting up their workstations, I would often be asked to set up their news filters to narrow the feeds down to something reasonable (the typical newswire delivers hundreds of stories an hour, most subsribe to several newswires). In the late-90’s the tools were crude and after getting frustrated by throwing in a few keywords, I would get called in to refine things using additional tools such as company ticker symbols, or a few undocumented codes from a taxonomy of subjects that varied from newswire to newswire.

Today the problem of information overload has spread to the greater population trying to derive value from the rushing torrent of updates coming out of twitter and facebook. How do I manage all this stuff and figure out what’s important? We use the tools we have but if you think about it, Google Trends and twitter search are just keyword searches with very crude resolution. We have a long way to go before such tools will let us tap into the collective mind.

Perhaps it’s time for a crude taxonomy for social networks to help sort out the types of messages flowing back and forth? Imagine if all your tweets, facebook messages, and friendfeed streams came pre-tagged with the following tags or categories?

  • look at me, I’m doing something cool
  • check this out, it’s funny
  • books, movies, music, food, or sports
  • this is touching and will change your life
  • gadgets and meta, technology post about using technology
  • weather and the natural world
  • babies and kittens
  • my obscure hobby
  • breaking news, OMG!
  • make money now!

What other categories would you add? Librarians of the world, what keywords would you put into your search filters to help grep out what goes where? Categorization is the first step towards ranking and with ranking you get useful filters.


Google Reader Power Readers – unlocked

Browsing my feeds this morning I saw an ad for Google’s Power Readers feature appended on the bottom of a TechCrunch post. The ad pointed to the Google Power Reader page, an editorially crafted bundle of feeds made up of linkblog posts, generated by celebrities hand-picked and using Google Reader. This is the first time I’ve seen Google step up and take such an extensive editorial role in a product to the point where they are actively promoting an editorial voice.

It’s a smart way to promote not only the sharing feature of Google Reader but also Google Reader as place to consume feeds. My only criticism is that the subscribe option for this bundle of feeds is limited to . . . Google Reader.


There is a way to eventually make it to the source url for this bundle but you need to go down the path as if you were going to add it to your Google Reader account (meaning you need to login to Google) before they tell you the URL for the Journalists Shared Items page (via a re-direct URL which contains “source=prhomejournalistsall” which gives us a hint that the PR department is behind this. From there you can get to the RSS feed and subscribe to it as you will.

For kicks I’ve added the feed into a My Yahoo page with a few extra bits from Yahoo editorial added in as a bonus. A bundle of bundles if you will. You can grab it here.

Reblog this post [with Zemanta]
Current Events

Taking your finger off the button

Shares of United Airlines dropped 75% yesterday because of a poorly designed template. The Google News blog has all the gory details including screenshots of the Florida Sun-Sentinal site which included links to a old story, UAL files for Bankrupcy, in its automated “Most Viewed” sidebar widget.

The Google News robot crawled that link and because the destination page had the default header and no date stamp of when the story was originally published, Google News incorrectly interpreted the story from 2002 as today’s news. The dominos began to fall as downstream news agencies that obviously were short on fact-checkers re-circulated this old news as something new eventually finding its way on to the Bloomberg wire service.

This has happened in the past but never with such devistating consequences (UAL stock eventually recovered but still ended the day down 11% and is still down $1.50 from before the incident as of today). Recall the false Engadget-iPhone rumor and, Bloomberg again, which had been duped in the past, when a crafty short-seller found a way to mainline a fake story into the news desk of a lower tier press release service back in 2000. One can only wonder if this news will have an impact on the SEC’s recent recommendation that websites can serve as the official channel for financial earnings.

The news industry is under seige and there’s more pressure than ever to balance speed and economy of automation with the wisdom and judgement of human editors. But like riding a bicycle without your hands – you need to keep your eye on the road or you might end up looking like a fool.

Current Events

Google’s Flash-Eating Spider

This announcement is definitely cool and will open up whole new areas of the web to search. But truthfully I just wanted to post this because it lends itself to a great headline.

From the FAQ posted on the Google Webmaster Blog:

Q: What content can Google better index from these Flash files?
All of the text that users can see as they interact with your Flash file. If your website contains Flash, the textual content in your Flash files can be used when Google generates a snippet for your website. Also, the words that appear in your Flash files can be used to match query terms in Google searches.

In addition to finding and indexing the textual content in Flash files, we’re also discovering URLs that appear in Flash files, and feeding them into our crawling pipeline—just like we do with URLs that appear in non-Flash webpages. For example, if your Flash application contains links to pages inside your website, Google may now be better able to discover and crawl more of your website.

Current Events

Feedburner Stats Way Down

I noticed a big drop in the number of Feedburner subscribers to my blog over the past few days with the number of subscribers dropping nearly 50% starting sometime Thursday last week (May 8th). I noticed one other person reporting a drop and they pointed to Google Reader numbers being the culprit and, sure enough, if you look at the two graphs below, my Google Reader numbers are down significantly (230 vs. 60) but other readers (Netvibes for example) are down as well.

Anyone else notice this?

Current Events

Tim O’Reilly is Skeptical about OpenSocial

Tim O’Reilly sums up quite nicely the key problem we’re all waiting to see solved with OpenSocial

If all OpenSocial does is allow developers to port their applications more easily from one social network to another, that’s a big win for the developer, as they get to shop their application to users of every participating social network. But it provides little incremental value to the user, the real target. We don’t want to have the same application on multiple social networks. We want applications that can use data from multiple social networks.

Tim’s a great writer and is able to sum up what took me over 600 words to try and describe. But both these posts are worth a read for the comments.

In my post, Paul Linder from Hi-5, one of the launch partners for OpenSocial says that OAuth will be the preferred authentication method that will potentially bind identeties together across social networks. This is great because it’ll be neutral.

In Tim’s post, Kevin Marks from Google reminds everyone of the technical challenges of preserving privacy of data across social networks that honor the complex rules and “social norms” of each social network. In his comment, Kevin hints that “publicly articulated performative” social networks (i.e. twitter, MySpace) would be easier to integrate because the data such as friend connections are public and connections are made with that in mind.

This is why I think the momentum and growth will favor social networks that are built on openness. Just as my parents are frustrated with the permissioning layers of flickr that I have put in place for personal photos, the vast majority of users are going to opt for simple to understand integrations, those that are open.