Ben Gomes on Google’s user interface

This summer several people in Google’s quality group have pulled back the curtain on how people think about search quality at Google. We’ve had Udi Manber give an overview of search quality and the groups that work on it. Then my office-mate Amit Singhal discussed some of our principles of core ranking. Amit followed that with a post about how we understand pages, queries, and users that revealed that Google does much more sophisticated semantic processing than just keyword matching.

Today, Ben Gomes steps out from behind the curtain to discuss Google’s user interface for our search results. Ben is another office-mate, he’s been at Google longer than I have, and I think he’s got quite a knack for blogging:

A common reaction from friends when I say that I now work on Google’s search user interface is “What do you do? It never changes.” Then they look at me suspiciously and tell me not to mess with a good thing.

Ben goes on to reveal a bit of the philosophy behind Google’s search interface, which might seem counter-intuitive at first glance. For example, a big goal of our search results is to get you off of them and to your destination quickly. That’s one reason why we usually put query refinements (which are a somewhat distracting feature) toward the bottom of the search page. If you get to the bottom of the search results and still haven’t found what you’re looking for, then you’re more likely to actually want those refinements to modify your search.

Every feature on Google’s search page has to defend its pixels in terms of usability, and Google tests a ton of changes that most people never notice. For example, Ben points out that we know “Arod” and “Alex Rodriguez” can be the same thing. Instead of hitting people over the head with that, we just subtly highlight the words “Alex Rodriguez” if you search for [arod]:


Get the skinny over on Ben’s post to read more about how Google thinks about our search interface. Ben, thanks for writing this–I’m glad that several search quality folks are working to be more open about how Google works and how Google thinks about search quality.

25 Responses to Ben Gomes on Google’s user interface (Leave a comment)

  1. Google’s search interface has influenced the interfaces of the other 4 major search engines due to its simplicity.

    There are just two suggestions that might be considered as an experiment.

    Put all of the news and video results in a separate area at the bottom or top of the SERPs – but not mixed in with the organic Webpage results.

    Also for the homepage, could you spruce it up on important holidays by adding a nice easy to load background image instead of the special logos.
    The backgrounds could have the Google logo embeded so as to load fast.

    In an interview with the NYtimes – Amit Singhal indicated that the algos are tweaked several times per week.

    If this is true, you probably repeat many of the experiments after re testing them. But how is it decided with tweak stays and goes or is modified upon?

    You cant really tell in such a short time by the user behavior?

    Also has there been a decline in the status of Wikipedia results? They do not appear to be a high on the rankings as they have been for the past year?

  2. This statement:
    “a big goal of our search results is to get you off of them and to your destination quickly”

    me just wish it to be true. Sometimes your SERP is “so good” that user do not need to click in order to get information that is looking for. Yes, all we that use Google would say awesome, but there is but… you are becoming the biggest “scrappers” of the web, hijacking users from one site using our information and our content to create new your content and keep surfer around your properties and your ad space.

    Examples are infinite.

  3. Silicon Valley, I believe some experiments run for weeks or even months, while others can reach a conclusion in hours or days.

    Zoran, I thought someone would ask about that. A more nuanced explanation is that we want to get someone the information that they need as quickly as possible. The vast majority of the time, that’s done by sending someone to a web site. But earlier today I was using the calculator feature of Google. That’s a good example where we can answer a user’s information need immediately, and I think we should. But the vast majority of the time we’re going to be sending people away from Google directly to other sites. And I think that we send most users away faster (for an average session) if you compare us to other major search engines.

  4. Cool, ranking is based on what’s out there in the web, all Google does is try to improve its interpretation of what’s out there. That’s pretty much clear.

    How does Google deal with new developments online? The way to interprete what’s out there should change with what’s out there, no? Does a link today represent the same thing as it did 5 years ago? Internet culture changes too, doesn’t it?

    Here’s a question that I find very interesting. I hope you´re allowed to answer it. What’s your opinion in how far Google is doing Information Retrieval? If you’d have to put it on a scale of 100% What percentage is data retrieval and what percentage is information retrieval?

  5. Nice reading on both yours and Google blog. Google Rocks!

  6. How come Google doesn’t have sitelinks for itself? 🙂

  7. @Matt Cutts
    Thanks for response. I do not object against calculator feature, me personally use it every day and 5634/67 is not proprietary information. Even currency converter is very close to “using somebody else data” as own without giving credit to the source. Even if you post link next to converted value “currency exchange rates from … at (time) retrieved.” 99.99% of users will ever click to go to that source. What about airline data? Weather? Products and Base box…

    Well, once I have seen, I think Philipp Larsen have made mock-up of Google SERP, that looks like “Wikipedia” page, built out of content retrieved around the web. That is what I worry about. One day search engines may stop to be search engines but more and more final destination.

    I just want to say that your user interface from day to day becomes more and more Google centric and we are getting less and less traffic.

    I think there has to be balance and one day when you cross that line I guess humanity vill react and some alternative will show up democratizing access to information… like you once lead that at the beginning of this century.

  8. It is a pity that google doesn’t have the same quality control on their back end systems.

    I used the country targetting option in webmaster tool 4 months ago. I set the country to the UK and then my website dropped out of (5th page), whilst still being 1st page for queries on every other google country server. It has now been 3 months since I switched the targetting off and still no sign of my site.

    I wonder if anyone will ever fix

  9. Hi Matt,

    Couple of months back, there was a report saying Click throughs on yahoo search results page is more than that of Google (I think you commented saying, Google displays the required answer in the results page and searchers don’t need to click on any result). So Yahoo! is sending more traffic to other websites than google.

  10. Here is one cool feature for Google toolbar (that the most probably you will not ever implement even as advanced option) that I would like to have. (Maybe I will develop my own FireFox extension).

    After every keyword instead of suggest to show top 10 results in the form Title (domain) and click to go straight to website.

    That would save zillions of seconds to mankind and increased speed we access to information… but as I said it would never happened and most probably my extension would get ban “you have spyware installed”… famous page.

  11. Speaking of semantic processing. Yesterday I was searching for something related to “private messaging”. I noticed that Google was highlighting PM in the SERP. I thought, that the search engine is getting smarter and is doing some semantic analysis of my query.

    The problem was that the highlighted PMs were those appearing in 2:00PM and such. So at the end, this semantic analysis was more a burden and it actually affected in a bad way the quality of the search results. I don’t know if there is away to turn this semantic processing off.

  12. You guys should add “PayRod” as a synonym for Alex Rodriguez…and for that matter, “bush leaguer”. Howie Clark sure won’t forget PayRod anytime soon…nor will most diehard Jays fans. (Someone had to mention that).

  13. Matt,

    I don’t know if it falls under Ben’s purview, but I can’t get the “Books” results from Google Books to appear with any kind of consistency. It doesn’t appear to have any dependency on whether or not matching (or closely) matching title are in the program, or even how popular the search phrases are. It seems more like the search people established a list of reserved words for which somebody made the decision that the user probably isn’t looking for a book, so book results aren’t displayed. Just try Googling something like

    Java Book
    Internet Book

    and book results appear at the bottom, while

    Computer Book
    Laptop Book

    draw a blank.


  14. Well, I can say that the very first reason I switched to Google was the speed. There were no pictures on the main page.

    I work on computers remotely all the time and when I need to open Internet Explorer through an RDP session, and all those dynamic images, I can never get the thing to open. But, if Google in the home page, it opens right up.

    Perfect, like it was designed for me…

    ~ Jim

  15. Google surpasses other SEs because of its indescribable less response time for a search query. Imagine when i was using a dial-up connection back to old days, its speed clings me on it to use it every time i need to search for something on the net without suffering from heavy net load. Simplicity is beauty, that’s Google.

  16. Dave (original)

    Matt, IMO, “Advanced Search” should read “Refined Search”.

    Also, how one be signed into their Google acount WITHOUT having SERPs skewed to personal search? Would be a nice *option*.

  17. This is off topic but why does still come up so strongly (with expanded navigation and all!) in google now that the news is well and truly out that it is a total scam site that has swindled people for millions over many months?

  18. Dave (original)

    Matt, on the above valid gripe by Justin. IE informs the surfer that the said site is Phishing, while Firefox gives NO warning and takes the users directly to the site.

    While M$ and not as good as Google at search, their Browser is head and shoulders above that of Firefox in many areas.

  19. And what about the people who want to find information on the horse named Arod? Now they have to type in “arod horse” or some qualifying thing like that.

    That leads me to wonder if the algorithms are moving toward favoring the most popular subject associated with a keyword (I’m pretty sure A-Rod is more popular than a fictitious horse). If that’s the case, what are you going to do for people who share name spaces?

  20. Sean from Ontario

    This can be fixed by the user typing in “+arod -alex -rodriguez +horse”. Those results shouldn’t have anything about the baseball player, and everything about a horse.

    For some reason, I like to play with Google’s search and I search about searching with Google. While searching for this, I’ve never come across a page listing all of the helpful commands for Google, BY Google. Either I’m not looking hard enough, or it doesn’t exist.

    I’d like to see a page made by Google that has tips and a complete command list. If it exists already, I’d be grateful for a link.

  21. Ben goes on to reveal a bit of the philosophy behind Google’s search interface, which might seem counter-intuitive at first glance. For example, a big goal of our search results is to get you off of them and to your destination quickly.

    No, it’s not counterintuitive at all. It is what you have been saying and doing for 10 years now. I very much understand what you mean, when you say it.

    But to me, this philosophy is your biggest problem. It’s a problem for two reasons: (1) It’s the wrong goal for a search engine. As search engine should be to help meet people’s information needs. If you go back to what “relevance” is, and how it is defined, it is defined in terms of someone’s information need. It is not (necessarily) defined as “1 of the top 10 links”. And if you place a higher priority on getting people off of your site than you do on meeting their information need, that’s not a good thing. Your goal instead should be to get them off of your site, if and only if, when and only when, they have finally had their information need met. But that’s not what you do, and it is very frustrating to some of us.

    This ties in to the second reason why your philosophy is a problem: It is not user-driven. It is a philosophy that you have, which leads to a strategy of imposing certain constraints on the user of your system. You pre-pattern the user to act in certain ways, because of your philosophy. If instead you placed the user at the center of the equation, you would not (necessarily) have this same philosophy. Because some users have different information needs than others, which information needs are better satisfied with richer, more information-filled interfaces.

    Thoughts? Do you get where I’m coming from?

  22. This is the first I have ever heard about the “+arod -alex -rodriguez +horse” search options. I really doubt that most people know how to search like this.

    Also this example shows how much Google loves wiki. The few times I’ve searched for Arod (I’m a Yankee Fan), I’ve been looking for his MLB page, not his wiki page, but his wiki page is the first hit in Google. You would figure that Arod’s page would be first, but it’s not…

  23. Dave (original)
  24. Kenny A. – this is the first time I’ve heard about this search format too. I agree that most people are probably unaware of these operators that can alter your search query.

    I also think that if someone is looking for a horse named “Arod”, and knows nothing about baseball and Alex Rodriguez…than even knowing how to manipulate the search results by using these operators – is useless.

    There needs to be a simpler way to eliminate non relevant results.

  25. I don’t like that feature. More than once I have been doing an obscure search and it has come back with the search I didn’t want it to do. I sure would like an option to run the search the way I want to run it rather than what your algos think I want.