You&A with Matt Cutts
At first, Matt comes in and says that Danny, who introduced him to the session, is not dressed casual enough. Danny, therefore, strips into a Disneyland shirt, cargo shorts, and funny shoes.
Michael Martinez asks the first question. A lot of people have found that supplemental pages don't come up. Is this the way that they'll be in the future, or will they be found in search results if there are links in the near future? Matt: Both of these are true. If you get more links in supplemental results, they can get in the main index. We parse pages differently that are in the supplemental index (than in regular SERPs). That's what distinguishes the main index from the supplemental index; supplemental index pages are compressed more often. The phrase relationships are handled a bit differently. You shouldn't worry too much about how many pages you have in the supplemental index. I have hundreds of pages in there; that's fine.
Question: You're doing more and more to figure out where the user is to deliver relevant results. What percentage of your results do you feel are completely relevant? Matt: That's a little outside my area of expertise. We can target people geographically.
Question: Pat from feedthebot asks about the Google Guidelines. I'm curious to see where you see the future of the Google Guidelines and why are they so undescriptive that I have to write about them? Matt: People should be able to think. It would be really good to have examples and the folks, like Vanessa, who are on the Google Webmaster Central, are working on making this clearer. Having more details is a pretty good way to go forward. Pat: Feel free to use feedthebot.
Question: Speaking of link schemes, I came across an ad looking for a SEO manager and they were looking for "expertise in buying links." What would you say about this? Matt: When people advertise for that, you have no idea how many people write to Google and say they are offended. We get those job descriptions forwarded to us. I wouldn't be surprised that we have something like this in the Webmaster Console in the future, just like we had the Spam Report form (which we had since November of 2001). We try to approach things algorithmically and also take people into account. We consider buying links to be outside of our guidelines and we might take strong actions on that in the future. If people want to ignore that, we as a search engine might take action on that because we want a high-quality index.
Question: Since we're talking about links, the more links you get, the higher you get ranked. What about links that come out of your site? Matt: I think it's good for your users, and therefore it's good for search engines.
Question: I want to talk about search pages and indexing of search pages. We have a catalog of many items and we want these indexed. How can you make it user friendly and search engine friendly? Matt: We basically said that you should avoid search result pages for good results for your users. However, that's not in the quality/spam guidelines; that's in the technical guidelines. It's more of a "best practices" guideline. We do reserve the right to remove result pages that don't add value to the user. What is the value at? If it just looks like search results that are available anywhere else, you have a problem. If you have unique content, that's much better. Categories are great. Pretend you're a competitor and ask "would this be a good page?" It will help you think of search in new ways.
Question: What is the impact of click-through on authoritativeness (personalized search aside) in organic results? Matt: We haven't talked about whether it will affect general web search. What I will say is that if you were to use that as a signal, it would be very noisy. If usage metrics are involved, everyone will jump onto it. I'm afraid of using metrics like that. I think MSN has actually said, "yeah, we use that." But because of the people who would try to optimize, we haven't talked much about those signals and to which degree.
Question: Why does Google love Wikipedia? When will you break up with him? [Everybody laughs.] Matt: That's an interesting question. Let me put it to you differently. By definition, people who go to Google are not regular users. [Added by Barry: Matt said those that are attending the SMX conference are not like normal Google users. I.e. SEOs and SEMs are not the average Google user.] This question actually came up on SEOmoz where they asked "why do you rank Wikipedia over accurate sites?" Regular users do like Wikipedia a lot. That said, it's not always the right answer. It's a fairly good result most of the time, but for expert results, sometimes it's not the most accurate. We change the algorithms to make the most accurate result #1. Followup: I work for Edmunds.com and Wikipedia bounced our results down. Matt: I'll take that into account.
Question: At Pubcon, the first thing you did was look up every domain everybody owns. So I have a question - what business is that of yours, and can other domains affect this? Matt: If you have 10 sites and 6 of them are catalog sites and 4 of them are hardcore porn, that doesn't bother me. But if a webmaster has two sites versus 50 sites, there's something to talk about. That's a different webmaster than the one who has 2 sites. My goal was to determine if this guy was a power webmaster or if he's a beginner webmaster. Followup: Algorithmically, it has nothing to do with it. Matt: [silence] [Laughter] Matt: I would consider that fair game. Another site by itself though might not be bad. But if you have 200 sites that we think are spammy, that may not be the best.
Question: Jason Calacanis has said that Yahoo and Google have too much spam in their results and he thinks that his results in Mahalo have more relevant results. Matt: It's too early to say because his site was just launched. We have all these other sites - chacha, Sproose - I support these. Let a thousand flowers bloom; let's try these approaches. The whole idea is that humans have nothing to do with search at Google is not true. PageRank is based on hyperlinks. We have toolbar voting, we have report spam - we look at our webmaster guidelines and try to remove the wording focused more on "algorithmic" results to allow for possibilities to include more human involvement. It's good if Google changes things for things like social search. We've used humans in many different ways.
Question: I have a followup on categorization and permutations. We have millions of products and our category pages are in trouble. How do I put this against my competitors? Matt: Suppose your product is in 3 categories - shoes, sneakers, "best" category. Find the best-applicable category instead of having the same object show up 30 different times. That's why you're outside bounds. Followup: But if you have color, shape, etc - you'll still have trouble. Matt: It's difficult to slice and dice that. Color as a category may not be as important. Take a look at your category - talk to your users, and only work with the categories that are most important. As far as competing with your resellers, that's tough. A lot of these companies are savvy - look at what they're doing. You can learn a lot from competitor analysis.
Question: I wanted to follow up on the algorithmic vs. human powered in terms of a recent Google Bomb that was diffused. Everyone seemed to agree that it was not a human intervention, but now you're saying that there is human intervention. Please clarify. Matt: The Googlebomb algorithm is completely algorithmic. My understanding is that the algorithm was not changed. It doesn't run every day. It runs every 2/3/4 months. I think Google does reserve the right to use humans in a scalable way and to take manual action on spam, and to the best of my knowledge, it has remained that way. Danny asks: What about Bush and failure? Did you keep pushing the button? Matt: By putting the word on the page, it comes up as a valid match. There is a thing that determines potential Google Bombs, but in that instance, they were engaging in discussions ("failure" was on the page).
Question: I have a question about image results in the search results page. How is this going to evolve? Matt: We always have a Onebox. When people look for "sunset," they are often looking for an image of a sunset. That's what our Onebox did. With Universal Search, we're looking for the right result. If you're looking for "fix a sink," it might be appropriate to have a video result. We'll check out the ones you mentioned ("George Bush" shows Jimmy Carter image; "retarded" shows people's photoshopped image). We're getting better at image analysis but by no means are we perfect. People can come and say "the search for my name is incorrect" and we'll try to fix it.
Question: A few years ago there was a University of Washington video where they put similar keywords in results (LSI, themes). How has that technology evolved? What about a theme across a website -- people think that it will dilute results. Is that true? Matt: My bottom answer is "try it and see what works for you." You might have "bio," "biography," "discography," etc. Work these in a very organic way. You don't need to do it artificially. Work these synonyms in a natural way. There are people who are fans of LSI. Google does a lot of work behind the scenes to do good semantic matching. If people are searching for bios, they are generally looking for biographies. But Apple and apples is not the same thing, for example. If you can do it on your own, that's fantastic, but we as Google will try to do semantic understanding ourselves under the hood.
Matt asks: What do you guys want from the Webmaster Console? People answer: Penalty reports, real time information, accurate reports, errors without having to go into each domain, spider traps, shared logins, RSS, 404 referrers, more data on a query.
