Posts about socialgraph

Content farms v. curating farmers

Tweet: Content farms v curating farmers: Deeper insights in Demand Media’s model & finding opportunity in finding quality.

I spent an hour on the phone the other day with Steven Kydd, exec VP of Demand Studios, to understand their model—using algorithms to assign content creation based on search and advertising demand and to minimize cost and maximize revenue—because I wanted to learn a deeper layer of lessons than I think we’re hearing in the discussion of Demand’s allegedly evil genius.

The talk thus far misses their key insight and the opportunities they create. Much of what I see online is fear that Demand Media—with the slightly rechristened “Aol.” following fast behind—will cheapen content and flood the internet—that is, search results—with crap that’s just good enough to fool algorithms. Some also fear that while putting content creators to work they will put better content creators out of work: the dreaded deprofessionalization and deflation of media.

Michael Arrington marks the end of “hand-crafted content” (somewhere I hear Nick Carr and Andrew Keen cackling maniacally). And Read Write Web’s Richard MacManus worries that the web’s quality will suffer.

They may be right. But then again, the internet has always been filled with crap. So the challenge has always been how you find the cream. That’s where opportunities lie. That’s what Google saw. The new question is whether Google can keep ahead of the content farms and continually find new and better ways to find better stuff. I’ll bet on Google over crap-creators. But they better get cracking.

This is why, when I proposed an X prize to solve media’s key problems at Yale symposium, Clay Shirky responded with a call for work on what he called “algorithmic authority.” A few of my students’ proposals in my entrepreneurial journalism class tackled just this problem with discovering and prioritizing content for us: one using humans aided by algorithms, one using algorithms aided by humans; neither operated like a one-size-fits-all search engine (but then, soon, Google won’t either).

I think we may see search fall as the sole or even key means of discovery and filtering of quality content. I see three rings of discovery today: search (Google); algorithms (see: Google News, Daylife); and humans (see: Twitter). Note again that Bit.ly alone causes as many clicks a month—one billion—as Google News. Human power rises again. That’s what Fred Wilson says today when he argues that social beats search, because “it’s a lot harder to spam yourself into a social graph.” As search becomes more personal and no longer universal, SEO as a dark art and as the fertilizer for content farms will diminish and the social graph — our own circles of authority — will become more important in search as well. So I have faith that there are solutions to stem any rising tide of crap.

This is how I put it in a tweet: “algorithm-aided human writing will meet human-aided algorithmic curation; quality will rise.”

In all of this, I caution us not to miss Demand’s key insight: that the public should assign the creators, including journalists. The public often knows what it wants to know. I learned this lesson when I consulted at About.com and saw how they monitor search queries to see where there are questions for which the don’t have answers. When that happens, they go write answers; Demand automates the process. Makes sense.

This is not how we have operated in media: We decided which questions to answer because we asked them. What hubris! Today, I teach my students to find conversations on the internet and add journalism to them in the form of answers, corrections, reporting, explanations. In 2007, my students in a seminar at Burda in Munich and in my class at CUNY asked why the public doesn’t assign us and my entrepreneurial students in two classes have worked on that problem. Jay Rosen just started playing with this notion at ExplainThis.org, creating a platform for the public to ask reporters to report their questions. Demand and About are doing the same thing, only through search queries. Jeff Sonderman compiles some more examples. Where appropriate, reversing the assignment pipe is a good idea.

Demand is also creating a system they say will find the best writer for each assignment. We are free to disagree with their methods and results, but there’s insight here, too. Two students in my entrepreneurial journalism class won a grant to create a platform to do just this with local and hyperlocal news assignments (note that Kydd told me Demand isn’t touching news); I’ll report more on their project as it gets closer to launch. Can’t news organizations learn and steal some lessons from Demand? What if you wanted to create a content asset — say, a complete travel guide — and you opened up the process and its discreet tasks to a marketplace of paid contributors, enabling you to do larger projects at lower cost than before?

I always tell my students: Wherever you see a problem, look for the opportunity. That’s Arrington’s point: The next generation of content creation is here; deal with it. If you don’t like what Demand et al are doing, see the opportunity in it to surface quality content and to create competitive quality stars whose creations rise not just through algorithmic search cynicism but through human recommendation. Dig to the next layer.

* * *

I got lots of details from Kydd about the Demand method. In their view, they have combined content-creation and social-media platforms to enable content creators with “spare cycles”—his nerdy words—to earn money.

Kydd says 11 community members contribute to each article by fulfilling the discreet functions Demand identified: writing, copy-editing, copy-chiefing, reviewing titles, managing topic pages, checking facts. That is done by freelancers. The staff directs, edits, curates, and manages them. The algorithm makes all this more effective as it tracks content and ad demand and writes headlines for pieces it says will get traffic and earn money. Editors are 1.5 times more effective in creating assignments that will generate traffic, Kydd said, but the algorithm is 4.9 times better than creators.

Kydd said Demand pays from $0 (with revenue sharing) to $100 per piece; it averages at $20. Copy editors make $2.50-$3 per piece, which works out to $15-20 an hour. He said these people like to wake up and know there’s work they can do—there are 100k assignments waiting for takers right now—while they wait for old, human editors to respond to pitches. He said they also like being paid twice a week. Kydd said Demand employed 4,500 creators (text and video) and 400 copy editors in the last 30 days.

What amazed me most is that Demand uses its method not only for service content but for jokes at Cracked.com. Could an algorithm and social network replace Jay Leno? Easy.

: LATER: See also Doc Searls on junk food and chefs.

: Paul Marcum tweeted today: “Prediction: increasing clutter from algo content farms + mobile app convenience will have even @jeffjarvis paying for news by 2011.” I responded seeing the irony here: that value will come from aggregation and curation of quality content. But imagine then if the aggregators become more valuable than the creators and start charging; the creators (i.e., Murdoch) will go batshit. I’ve argued that in the link economy, there are two creations of value around content: from those who make the content and from those who bring together the public around it. Where is there greater value? We’ll see….

: LATER STILL: See Upendra Shardanand (founder of Daylife, where I’m a partner) on the need for new tools to create new handcrafted content. Problem is, he says, we’re using old text tools. See my related posts on storytelling and post-page media.