Below is live coverage of the Getting Personal, Part 2: How Google & Bing Personalize With Search History & Geography panel from the SMX West 2012 conference. This coverage is provided by Avi Wilensky of Promedia Corp.
Disclaimer: The coverage is brought to you in real time, using a custom live blogging tool. Feel free to ask questions or leave comments for inclusion into the live coverage. During the live event, live notes will auto-scroll with newest entries at top. After the session is complete the archive version will have the oldest entries at the top. We ask you to please excuse any typos, as these are live notes.
Finished
Auto-Refresh: Off
Avi Wilensky:
10:38:37 am
Moderator: Danny Sullivan, Editor-in-Chief, Search Engine Land (@dannysullivan)
Q&A Moderator: Michelle Robbins, Director of Technology, Third Door Media, Inc. (@MichelleRobbins)
Speakers:
Rangan Majumder, Principal Group Program Manager, Bing
Jack Menzel, Product Management Director, Search, Google (@jackm)
Q&A Moderator: Michelle Robbins, Director of Technology, Third Door Media, Inc. (@MichelleRobbins)
Speakers:
Rangan Majumder, Principal Group Program Manager, Bing
Jack Menzel, Product Management Director, Search, Google (@jackm)
Avi Wilensky:
10:49:11 am
Horst from Searchmetrics starts off.
Avi Wilensky:
10:51:00 am
Steve Boymel is sitting next to me - shout out.
Avi Wilensky:
10:51:58 am
Check out Searchmetrics free new open beta tools - just register.
Avi Wilensky:
10:53:11 am
Danny: In this session we will look at the user personally and search history, welcomes Jack Menzel to podium.
Avi Wilensky:
10:53:44 am
Jack is a product management director at Google Search and will talk about personalized search.
Avi Wilensky:
10:54:52 am
All about most relevant results. Understand you. SMX is a good example of ambiguous nature. SMX can be a motorcycle, home theatre, etc. Talking about things more personal and how google can understand query to get you to most relevant results.
Avi Wilensky:
10:58:07 am
If you type in Pizza, location is key. Query is from San Jose, not very relevant to user. Type taxes, and IRS comes up. In UK - not so relevant. It's the wrong result, need the result for taxes in the country you are in. Another example is at the metro level - "when's the next train" - local intent to give best results.
Avi Wilensky:
11:00:31 am
If search Google for lawnmowers and click on amazon, Google will see you are more likely to visit Amazon again. If you type in a generic term like "dominion" - it may be a power company, or a card game, so Google needs to know the user to give them the right result. Based on your search history.
Avi Wilensky:
11:01:07 am
Also use web history to get you back to sites that you like. Jack likes a scrabble dictionary that is not as popular as others, so he will get served that more frequently.
Avi Wilensky:
11:01:45 am
How can webmasters help? Make content users want. Make content that users want to come back to.
Avi Wilensky:
11:04:14 am
What Google struggles with to get people to your content is basic - for a business listing - need a valid phone number, hours, for example. If you have a local presense, go to Google places and just enter the right info. So important and easy but mundane. Not sexy but necessary. Make it simple for google to find your content. Whatever you can do to make your phone number, address clear and simple and readable.
Avi Wilensky:
11:05:48 am
Make your website work on mobile phones! Don't have to be complex, just make it simple.
Avi Wilensky:
11:06:28 am
Go into GWT and help us out using Geotargeting and other features to get Google to understand sites better.
Avi Wilensky:
11:07:24 am
Also, if you want to see what results look like without personalization, can turn it off. Languages are another great way to connect right users with content. That's all!
Avi Wilensky:
11:07:39 am
Avi Wilensky:
11:08:10 am
Avi Wilensky:
11:08:29 am
Rangan from Bing is up now, first time speaking @ SMX - Welcome!!!
Avi Wilensky:
11:09:57 am
If you search on Bing for like Harry Potter - you see lots going on. He works on the 10 blue links. Not the related searches, not the answers...works on ranking of blue links. Also worked on autosuggest. Do a bit of personalization there.
Avi Wilensky:
11:10:57 am
Bing search: 4 components. Query, component, index, "ranker magic" (query intent)". Rangan's team owns that algo. Does not work on finding pages, crawling, spam, etc.
Avi Wilensky:
11:11:32 am
Search used to be about query text match to document text. What is user intent? Made things more complex but gave more opportunity to create a more relevant engine.
Avi Wilensky:
11:12:16 am
What is context? social - friends and interests. who are you? where are you? when are you searching? many others.
Avi Wilensky:
11:13:08 am
Long term personalization - who are you? Short term - what are you doing now? Location - where are you? Privacy - cant abuse user trust. Make it clear that changing the experience and controls.
Avi Wilensky:
11:14:31 am
Bing's rank 5 years ago was authority and keyword match. New factors are personal and social preference. Personal - how much will user like the content based on history. Behavior with content and similar to yours. Combine signals from Facebook - if have the users social graph, will leverage it to improve relevance.
Avi Wilensky:
11:17:08 am
Examples: With autosuggest - type in S - will see a mix of popular sites and local results. Can change location to see how autosuggest is personalized.
Avi Wilensky:
11:18:52 am
Also gives examples of search history influencing rankings. For "Yahoo" most times show Yahoo.com. For some people just want Yahoo Mail.
Avi Wilensky:
11:20:43 am
Key Takeaways: Personalization more complex but better results. Users get what is best for them. The new SEO formula is not just about links and KW's - people have personal and social preferences. Think about social authority not just link authority. Get people to follow your brand.
Avi Wilensky:
11:21:48 am
Help Bing understand your page, of course. Use SEO best practices....and that's all!!
Avi Wilensky:
11:25:35 am
Q+A: Danny asks - lets talk about GEO. Down to more than just the metro. Outside the US is it also metro? Goog: Yes, but certain countries the data is more spotty. In Manhattan can find something on same block, in Turkey, don't have the data. Danny: How do you decide location? Always search for SEO and always get a local irrelevant match. G: We do our best to TRY and target. Don't always get it right. A whole suite of tricks used to tease the query. Much fancier than just "words" to trigger local. B: The more generic the term, the more likely personalization will kick in.
Avi Wilensky:
11:26:43 am
Danny: 180 days later after the cookie, if I 'm signed in do you ignore that and use the web history? G: Yes, if signed in, will use web history. Danny: If I'm signed in and don't have history, is there a cookie history. G: I believe we do not. Danny: Would you swear to it in a court of law!
Avi Wilensky:
11:31:08 am
Danny now pools from the general questions: Q: If you were an ecommerce international company, how does geo help or hurt? Should I create an office for every office? G: Do not do unnatural things for the sake of Google. Use GWT to target. Should't have to do anything bizarre. Different language is good. B: Same as G, but it's a challenge. Tricky.
Avi Wilensky:
11:33:31 am
Next Q: How does this not turn into an Amazon issue where you buy something for someone else and you get served relevant content. How do I get unlocked out. G: You shouldn't see irrelevant content just because you've been there before. When thinking about the problem, we give you choices! Do try to value both the new content and the content you are clicking on all the time. Balancing game. B: Good thing about search is we still have the keyword history. Not as powerful as the search query. Amazon doesn't have that advantage.
Avi Wilensky:
11:35:00 am
Q: Do you make use of Geo-location services - G+ checkins, 4SQ, etc. Does it play a role? G: Alot of this content is private and we don't have access to it. Try our best to find out when businesses are closed, and the physical world - try to use any observations we can. B: Better question is will we use it? If It's valuable, then we will try to use it.
Avi Wilensky:
11:37:36 am
Danny: Lots of questions about national trying to reach global audience, etc. G: We know that people buy stuff online and in store, and we try to balance that. B: Like Jack said, we try to do our best, it's a problem were trying to figure out. Schema.org may come up with a solution to help with that challenge of locations served.
Avi Wilensky:
11:41:06 am
Next Q: Can you give us some sense of the % of personalization? Do you geotarget 90% of the time, etc...G: No target percentage, we try to tackle the problem by analyzing these loose percentages and vary by query, varies by the ability to do it and data available. No target percentages. B: Agree, but lots of opportunity. Last year about 5%, now it's less than a 1/3 but cant give exact number. Certain queries get impacted more than others.
Avi Wilensky:
11:43:33 am
Next Q: Bounce rate - you can tell someone comes back to the result sets, is that a factor? G: We try to figure out if search results are good. Many ways. Not so simple. We look for queries to improve by looking for info people aren't finding what they want. B: IF you bounce, probably not good - will look at signals to see if users are satisfied. DAnny - can be fooled with Mturk - B: no.
Avi Wilensky:
11:48:27 am
Danny: Can the bots tell if the sites are not mobile friendly? G: Tough, we don't want to hide good content because of browsers compatibility issues. Danny: There was a time there was mobile content and search engines. G: Nothing bad will happen if consumers can get to your content. B: Mobile is definitely a context. In general, match context of the query to documents optimized to that.
Avi Wilensky:
11:53:40 am
Q: If you have a dense geo - for example SF and the valley - does population density affect search results? G: We try to give you the right local results given your area - how far people are going to get local content - if you are in NYC vs. Sunnyvale - how far you are willing to travel is different in different areas of density and we do try to tune it. Not as simple as population. B: In terms of web search, if an area is more dense, you need to be more precise. G: Bodies of water are so mean, because it's close but you wont go across the bridge to get pizza, etc.
Avi Wilensky:
11:55:33 am
Danny: This is really screwing up ranking reports - makes it difficult - what are we supposed to do? G: We apologize! At the end of the day we want users to get to content they want. We measure EVERYTHING to improve UX. Sometimes it hurts ability to measure. B: Trading complexity and user experience.
Danny: Ranking should have been tossed out years ago because you cant guess at all your terms. Look at overall traffic increasing from search.
Danny: Ranking should have been tossed out years ago because you cant guess at all your terms. Look at overall traffic increasing from search.
Avi Wilensky:
11:56:37 am
Danny: Clicking on ads, is that recorded: G: Yes, B: Don't know. That could be a useful tip to get a user to click on an ad and then it becomes part of their history.
Avi Wilensky:
11:57:16 am
Danny: Any Q's for the audience to the panel? B: All ears for feedback. G: Big challenge.
Avi Wilensky:
11:58:43 am
That is all. Thank you Michelle Robbins for QA moderating. Expo hall is now open!
