Live from SMX East: The Real Time and Social Search Landscape

By Janet Driscoll Miller | Oct 4, 2010
More Articles by Janet


After an early and rainy flight into LGA this morning, I started my visit to SMX East with the Social and YouTube track: The Real Time and Social Search Landscape, featuring Jeremy Hylton of Google, Brian Theodore of Yahoo!, Meredith Ringel Morris of Microsoft Research, and Othman Laraki, of Twitter.

Jeremy Hylton – Google

Jeremy Hylton from Google was up first, and he is one of the lead engineers on the real-time search product. He started by discussing Google’s investment in search technology, including Instant, Caffeine, Social, Realtime, and many other changes that users may not have even realized happened.

Realtime search gets results from Twitter, FriendFeed, Jaiku, Buzz, Facebook and Myspace pages and show up in your Universal Search results. Launched in December 2009, Realtime is now available in many languages, including Japanese, Russian and Spanish. You can also restrict the Realtime to a geographical area and subscribe to the results in RSS or receive them to your email via Google Alerts (select Updates).

Othman Loraki – Twitter

Othman was up next and works on search at Twitter.  He started with an overview of Twitter Search, which is an inverted index. It has the capability to process of over 1,000 tweets per second and over 12,000 queries per second (or more than 1 billion tweets per day). Twitter Search is used to power the Twitter Search API (for mobile, apps, widgets, etc.), searches on Twitter.com, and other numerous services on Twitter.com (such as related tweets).

Some example types of Twitter Searches include:

  • following an event or news in realtime
  • addressing a hyperlocal question (why did the fire truck show up across the street?)
  • communicate with a realtime community
  • vanity searches (@ mentions)
  • standing search around an interest (such as online reputation monitoring)
  • gadgets

Othman said that the relationship between the search engine (Twitter) and the information is different in realtime. A traditional search is about routing you to the right place, but realtime search is about filtering all of the information to pinpoint what you want to know. He also added the “pothole theory” — it’s all about context. You care about the pothole on YOUR street, but not likely on other streets.

So what does ranking mean in a realtime world? The impact is different because of time of consumption. While PageRank was a great way to rank websites, it is relatively static, versus the realtime world where the value of information changes based on microtrends and events.

So what does all of this mean for SEO? Traditional SEO involves “hacking” the search algorithm. However, realtime SEO is about “hacking” human interest in topics. So how can you think about objectives for SEO? For traditional websites, you care about website actions, such as pageviews. With realtime, action is more immediate and perhaps engagement is the better measurement. Potential measurements might be followers, list inclusions, retweets, trending topics, clicks/pageviews.

Meredith Ringel Morris – Microsoft Research

Meredith was up next and explained that she works for Microsoft Research, which examines what may happen in the search  landscape in the next few years.

She feels that searches will be more about searching versus asking. For instance, she asked her friends on Facebook about her trip to New York in addition to doing a search on Bing. They did a survey to learn more about how people use social media to ask questions of their networks:

  • 29% recommendations
  • 22% opinion
  • 17% factual
  • 14% rhetorical
  • 9% invitation
  • 7% favor
  • 3% social connection

They also looked at topics that people asked questions of their social networks:

  • 29% technology
  • 17% entertainment
  • 12% home and family
  • 11% professional
  • 8% places
  • 6% restaurants
  • 5% current events

People don’t tend to use social networks to ask questions about health, finance, and things like porn — likely due to privacy — but they do use search engines greatly for those purposes.

Next they did an in-person observation study using Facebook versus Bing to find information for a problem. Over half of the testers got responses back from friends via Facebook faster than completing their search engine search. Their confidence in correctness of what they found from friends in Facebook and on the search engine result was higher when the same/similar result was found in both.

So where might social search be headed? Perhaps you befriend a search engine, and when you are looking for recommendations, then the search engine, like a friend, can make recommendations in your social network as well.

To read this research and other research projects from Meredith and Microsoft, you can download the reports here: http://research.microsoft.com/en-us/um/people/merrie/publications.html.

Brian Theodore – Yahoo!

Brian was up next and first gave some stats about Yahoo! and explained why they focus on “mass market” aspects. Brian indicated that while Twitter has mass market appeal, the content often doesn’t have mass market appeal. Some mass market consumption patterns are emerging, including:

  • “trending now” aggregation
  • algorithmic curation of linked content
  • easily identifiable and filterable service auto tweets and retweets

But XLST (style sheet) for Twitter is missing and ought to be the norm rather than the exception.

Changing consumer behavior is the barrier for mass adoption, according to Brian. He feels it only offers the user limited value like looking through a pinhole. There’s a fire hose of information. You need thematic clustering and time-series analytics to give a better user presentation.

What Yahoo has learned about using tweets on Yahoo:

  • Celebrity tweets have higher engagement than news
  • Needs to be an easier way to find/cover and expand in the SERPs
  • Influencers rule
  • Signal/noise is sub-optimal

Some leading versus lagging indicators of realtime buzz were that realtime buzz can be derived from multiple signal sources, but realtime buzz detection needs to include a local even baseline.

Share this article

Share on LinkedIn Share on Twitter

Receive Monthly Digital Marketing Tips

Subscribe to monthly updates from the Marketing Mojo Blog to get the latest digital marketing tips, best practices and insights - hot off the presses and straight to your inbox!

 

Blog Search