Is StumbleUpon trumping Facebook in the internet attention wars?

Wirehead - Source: AndYaDontStop
Wirehead – Source: AndYaDontStop

The latest salvo in the internet attention wars has come in the form of figures from StatCounter.

A relatively small content driving service called StumbleUpon drove more than 50% of all social media referral traffic in the US, news the company’s founder and CEO Garrett Camp was quick to tweet.
In other words, Camp claims his company now generates more more referrals than Facebook, Twitter, Reddit and Digg combined.

Facebook, second place in the US statistics, drove almost 39% of American social media referral traffic (but is still the number one social media site globally).

To put this in perspective, Facebook claims around 750 million accounts. StumbleUpon claims around 10 million. Goliath, meet David.

The StumbleUpon figures represent a new front in the debate over content freedom versus focus. But they do not offer an easy solution.

Systematic serendipity

StumbleUpon is a discovery engine. Instead of searching directly for content you already know exists, discovery engines are designed to provide you with content you didn’t know about.

In practice, the service combines your positive and negative ratings of content with those of other users and then provides you with personalised recommendations to view other content.

In this version of web-surfing, you click from site to site based on the ratings of the sites you are shown, rather than by searching for specific things or following links on sites.

There is a somewhat mythical model of pre-internet media that suggests diverse stories were supplied to the public on diverse topics, all gathered under one umbrella.

The searchability of the internet, the ability to self-publish, and the ability to link only to those you desire, have lead some commentators to argue that the internet leads to less diverse opinions.

The blogosphere, some contend, is a dystopian echobox, not a utopian flowering of public debate.

Arguably, StumbleUpon provides an alternative to the echobox. “Stumbling” is a form of systematic serendipity.

The service seems to be built on the apparent value of diversity of both opinion and content sources. Could its new-found importance indicate that users value diversity? I think this is somewhat illusory.

Illusory diversity

StumbleUpon traffic is unusual because, while sites may get a lot of StumbleUpon hits, users appear to spend less time on each site than with referrals through sites such as Facebook.

Part of this stems from the somewhat ephemeral nature of the web itself.

Sites, on the whole, are less permanent than the StumbleUpon database, and sites are highly variable in terms of presentation.

Users may hit a site that has out of date information now, compared to when it was highly rated, or users may find the site is broken in current browsers.

It’s not surprising that users quickly “stumble” from site to site, looking for something good, immediately relevant, and working well.

Users, then, are not necessarily valuing diversity itself. They might experience diversity by accident, but that’s not the same as seeking out a variety of opinions.

Further, those sites that are skipped may well contain valuable alternative content, but users skipping old or broken sites are not making decisions about the content on its own merits, especially if they visit only for a few moments.

Finally, compared to Facebook, the userbase of StumbleUpon is primarily the geek-inclined.

Even at 10 million, their up- and downvotes are hardly representative of seriously diverse international opinion.

Top seven social media sites in the US as of August 2011. | statcounter.com
Top seven social media sites in the US as of August 2011. | statcounter.com

The problem of finding diverse content On the internet how you get to content is far more important than why:

Why you want to see content is nebulous and subjective. How you get there is material and objective: you must click links to get to content and every click is registered.

The internet attention wars centre on tracking and manipulating how you get to content, both with and without your knowledge.

The diversity of content online is wonderful in theory, but there are two problems:

1) The needle in a haystack. You want to find content, hosts want you to see their content, and drivers want to send you to content – but just what content?

2) This is less obvious to users but crucially important: how free are you to actually find diverse content?

People basically find content on the internet in three ways. Let’s say that you want to watch a funny video.

1) You could go directly to a content site, such as Funny or Die. But that depends on you knowing the site exists and taking the time to go there – or perhaps signing up for site alerts.

The pro here is that you are likely to see something funny; the con is that there’s a lot of it out there.

Given all the sites you might visit, seeing all the possible funny videos will take a long time.

How might you know which is funniest? How will you be exposed to unexpected content?

2) You could use a search engine, such as Google. The pro here is that you may have more chance of seeing something unexpected than if you just go to a single content host, although that does depend on how focused your search is.

Search engines tend to work best when you know what you are looking for – they work against diversity because that’s their job.

But there are even more cons to search engine results.

You’ll get far more results than you can look through even on one site.

But what funny video will come up first? Which is the funniest? And what do you know about the results? Are you seeing everything relevant?

How many different search terms could you use? How do you know what results are being brought to your attention and which are being hidden?

Many search engines try to learn about your habits and personalise your results. The pro of this is that you might get fewer irrelevant results; the con is what Eli Pariser calls the Filter Bubble, whereby your exposure to alternative views is increasingly limited.

Combine automated personalisation with unknown corporate and government manipulation of search results and there’s a recipe for powerful direction of attention.

For example, you might not see funny satirical videos about online services or political figures.

3) You can rely on the recommendations of others. The pro here is that some recommendations are personal, such as seeing videos posted on a Facebook friend’s Wall, directly or as the result of a “Like”.

Other recommendations are crowd-sourced, brought to your attention through various forms of aggregated voting.

Facebook is still on top worldwide. | statcounter.com
Facebook is still on top worldwide. | statcounter.com

There are a couple of ways of getting to such crowd-sourced recommendations. You could go to a social news website, such as Reddit, where users find content, post it, and then “upvote” and “downvote” posts.

This leads to some diversity of content through the averaged range of opinions, although such sites also have cultures that prefer certain forms of content.

Another method is to use crowd-sourced aggregation services such as social bookmarking sites (delicious, zootool, pinboard to name a few) or discovery engines. Which brings us, once again, to StumbleUpon.

Given the way the service works, StumbleUpon users have deliberately opted in to a combination of personalisation and crowd-sourced opinion.

Further, since StumbleUpon is a browser extension, users have the benefit of serendipity and recommendations everywhere, as opposed to having to go to a particular site to hope for serendipity.

Google is attempting to do the same with its +1 button.

Knowing you have opted in to recommendations is certainly better than the opaque personalisation of search engines.

But no-one, perhaps not even StumbleUpon, knows how much StumbleUpon is being gamed by promoters using sockpuppet accounts – alternative online identities used to inflate up and down results.

Where do you want to go?

The holy grail of internet marketers is crowd-sourced, automated, demographically-identified content recommendation.

But that will only work, in turn, if users can trust the recommendations.

Even if they can trust the recommendations, users should be pushing discovery engines and search engines in general for more open standards for recommending diverse, as well as personalised, content.

Is there a market for this? The stats would seem to suggest there is. But the market must work for consumers as well as the marketers.

The Conversation

This article was originally published on The Conversation.

Read the original article.