Everything you need to know about SEO, delivered every Thursday.
Live Blogging: Interview with Amit Singhal, Google Fellow
Danny Sullivan and Chris Sherman are on the stage at SMX London to interview Amit Singhal. Amit is a Google Fellow, a honorary title reserved for Google’s most accomplished engineers, and he has spearheaded Google’s core ranking team since 2000. He’s also a key influencer of Search Plus Your World, Google’s search experience centered around people, that lets you find personal results from your world — your photos, your friends, your stuff — in search.
Chris Sherman is on stage to introduce Amit Singhal, Google’s Vice President and Google Fellow. Chris provides Amit’s background showing a dynamic visualization using Google Maps with the countries/cities that Amit has lived on. Amit got a M.Sc. degree in University of Minnesota. He worked for AT&T (Bell Labs), and from where he headed to Google.
Amit thanks for the intro and talks about his child memories and how he grew up watching Star Trek. He dreamed about creating robots that we can talk to. As an academic for many years he worked hard on language software that he believed would help him with his dream. In 2000 he went to Google and said to Sergey: “Your engine is excellent, but let me re-write it!” This became the new ranking system.
Amit then talks about how he got involved with the challenge of moving beyond keywords to tackle the problem of having the same words that have multiple meanings: how can you understand user intent? Apple (software) vs. Apple (fruit) was the first generation of search. The next leap in search technology will be when computers will understand the difference between those Apples. That’s what excites him.
In the last 5 years Amit feels he is very close to building his childhood dream. Even though there are many things to be done before achieving this dream, he feels Google is in the right direction and they will be able to achieve that. “Computers don’t understand things, they understand strings” and it is Google’s jobs to teach computers how to differentiate between different intentions.
Danny is speaking about Universal Search and how it evolved Google to Search Plus Your World. How is Search Plus Your World impacting Google? Amit says the key motivation behind Search Plus Your World is to have a secured search, it is the first baby step to achieve Google’s dream, and data shows that Google users like the personal results. It also gives the user one click removal from their personalized results. Google is currently analyzing and improving their personalization engine.
Chris mentions that personalization can be narrowing, as it gives people the same results and they do not discover new things. Amit answers that there should be different points of views in any search results, and Google is aware of that and they balance between personalized and non-personalized results.
Danny mentions a Pew research that concluded that people do not want personalization. Amit says “I am a scientist, when I look at researches I look at how the question was asked.” He discussed the specific research, and said that personalization is valuable for Google users. Danny asks: can you tell what percentage of personalized searches are clicked? Amit says people are clicking more than before on searches and it is lifting CTR from search pages.
Chris mentions Bing Social efforts and how it is different from Google’s. Amit says: “the key challenge with personalization is that no one can judge a personalized search for someone else.” That’s why Google looks at the data about how users like their results. Search Plus Your World is the same approach as Universal Search, people have to find what they intend to find on their results.
Danny mentions the integration Bing did with Twitter and Facebook, and how this might be good for users. Will Google do that in the future? Amit said that their contract with Twitter expired. Google cannot add Twitter and Facebook right now as their information is hidden behind a wall. It has been tough to build an integration in this terms.
Chris asks Amit how is the evolution process at Google with so many updates; how does Google decide about which update goes live? Google has an internal system where every flawed search result is sent to Amit’s team. Based on that engineers are assigned to problems and solutions are tested on a sandbox. Then the engineer will show how the results will show after and before the update and the update is tested using an A/B test. They discuss the results and this loop runs several times until they find a change that is better in all aspects. After this process the change is send to a production environment for a very low percentage of real user traffic and see how the CTR is changed. Based on this, an independent analyst (that works for Google) will generate a report. Based on that report the group discuss and decides if the change is going to be launched or not. That’s how scientific the process is. There are some videos available on some of these sessions: check them at this post.
Danny talks about Penguin and asks how it is going from Google standpoint, are search results better? Amit says that in the end of the day, users will stay with the search engine that provides the most relevant results. Google’s objective was to reward high quality sites and that was a success with Penguin. One of the beauties of running a search engine is that the search engines that can measure best what the users feel is the one that will succeed more.
From Google’s perspective they use any signal that is available for them, more than 200 of them. They have to make sure they are accurate and good. They will use any signal, whether it is organic or not. Chris discusses the link graphs and how it is common sense now, but what about the knowledge graph? Google wants to return to users the answers that they are looking for, and that’s what drives them. Google is increasingly investing in understanding the real meaning of each query so that they can return the right answer.
Danny asks about Paid inclusion in vertical products, which was against Google’s policy in the past. Amit says that a class of searches could not be answered organically and they realized that they would have to establish relationships with data providers to get that data. To be super safe and honest with users, they make sure that these results look different, and they also started calling it sponsors to be even more clear about that.
Chris asks about the dream of creating a communicating machine and asks how this will change the way we relate with Google. Amit says that these changes come in baby steps, and it won’t be an overnight change. Amit gives the example of spoken search, and how this data is still scarce, and Google will adapt according to data.
Amit was asked whether search results are measured by Google’s revenue or by relevancy to users. Amit firmly states that revenue is not measured at all, only relevance is taking into account when defining search quality.
Amit says that if you build a great search engine for users, they get more curious because they expect to get great results, so they ask more questions. Giving relevant results will give more time for people to search more and free them time.
Postscript: Here’s video of his response, which came to a question about how publishers might potentially lose traffic if Google provides more direct answers:
Chris asks: with the scope that Google have reached, is there anyone that still knows all of Google? Amit says that there are senior executives that each can understand very well their own “entities” such as Search, Advertising, and other big groups, but no one understands everything.
Danny asks which funny search Amit has came across. Amit says that once he read a query along the lines “do my ear make me look fat?” Amit laughs: “why are you asking Google that? Go figure it alone!”
Amit concludes that he couldn’t have a better job, he gets to influence search quality and also to improve the world in some ways.
Some opinions expressed in this article may be those of a guest author and not necessarily Search Engine Land. Staff authors are listed here.