Digital Conference 2010

Yahoo Scientist Questions ROI of Kardashian's Sponsored Tweets

At Ad Age Digital Conference, Duncan Watts Explains His Model for Predicting Value of Influencers on Twitter

By Published on .

NEW YORK ( -- Stop paying Kim Kardashian $10,000 per tweet.

That's the recommendation based on the work of Yahoo's principal research scientist Duncan Watts, who presented his findings at Advertising Age's Digital Conference today.

Duncan Watts
Duncan Watts Credit: Gary He
"If I had a fixed budget, I could get more value from a small amount of very influential [influencers], or a lot of smaller influencers, on Twitter," Mr. Watts said. "If you recruit enough people who, on average, influence just one other person, you could get a much better return on investment if you aggregated them and altogether paid them a tenth of what Kardashian gets."

It's timely research given Twitter's deployment of its business model yesterday, though Mr. Watts was clear his numbers were largely hypothetical. "I'm assuming a lot of things in this model," he explained, "but it's a good way of seeing what influencers on Twitter might be worth."

The insights grew out of Mr. Watts' larger work on social media, including an updated version of Stanley Milgram's famous test of the "Small World Problem." In the original test, Mr. Milgram concluded that, on average, two random individuals are separated by only six degrees of connections. Mr. Watts used e-mail instead of the postal service to test the theory, which led to a slightly higher number: eight.

But perhaps more enticingly, Mr. Watts also looked at a problem every marketer (and publisher) is currently trying to solve: How do you influence people online?

Search engine optimization is critical if you want to get your website noticed, and your products in front of consumers searching for your brand online. But SEO is more than just getting your site to perform well in organic search, it can also improve your site's overall performance. SEO expert C.J. Newton lays out important aspects to consider before you start rebuilding your website in this Ad Age Insights white paper, available now at

Testing rankings
To answer the question, Mr. Watts set up a web page of unknown music bands and purposefully reversed the popularity of songs as rated by the users to the site. He found that influence can indeed be engineered, as his experiment showed the more popular a song was ranked (however falsely) led to more downloads than it would otherwise have gotten.

"But the problem with doing that, we found, is the more you falsely rate, the less valuable the marketplace becomes," he said. "There were fewer overall downloads when you faked the popularity of songs."

Back to Twitter. "If you could imagine a platform that was ideally designed to identify influencers, it would be Twitter," Mr. Watts said. "This is not like Facebook. Twitter is a listening network, and it's also a talking network. Both sides are expressly interested in being influential."

But in looking at influencers, Mr. Watts found that it's incredibly hard to predict who will be a major factor on Twitter, a conclusion that runs counter to the prevailing wisdom of social epidemics popularized by the book "The Tipping Point." While he acknowledges there are certain personalities such as Kim Kardashian who can potentially trigger a larger cascade of re-tweets given her large amount of "followers" ("Tipping Point" enthusiasts call her a connector), close studies of social platforms reveal that influence is spread more efficiently and more reliably when done through many-to-many connections, rather than through a few highly connected individuals.

"Most of them will send tweets, and no one else re-tweets," Mr. Watts said. "A lot of times, not that many people are listening on Twitter."

Hear that Kim?

Most Popular