Central Perk: Personalization and SEO

vineri, 22 martie 2013

Personalization and SEO - Whiteboard Friday

Posted: 21 Mar 2013 06:35 PM PDT

Personalization usage data and user data give marketers deep insights into their users' interests and actions. But how can you make the most out of these complex data sets to better serve your SEO campaigns?

In this week's Whiteboard Friday, Rand takes us through the intricate world of personalization and how it affects SEO. We'd love to hear your thoughts and tips in the comments below!

Video Transcription

"Howdy, SEOmoz fans. Welcome to another edition of Whiteboard Friday. This week I'm wearing a hoodie and a T-shirt, so it must be informal. I want to take you in a casual fashion into the topic of personalization user data and usage data, and these are complex topics. This Whiteboard Friday will not be able to cover all of the different areas that user and usage data and personalization touch on. But what I do hope to do is expose you to some of these ideas, give you some actionable insights, and then allow you guys to take some of those things away, and we can point to some other references. There are lots of folks who have done a good job in the search world of digging in deep on some of these other topics.

Let's start by talking about some of the direct impacts that personalization usage data have. Of course, by personalization usage data I mean the areas where Google is showing you or other users specific things based on your usage activities, where they are leveraging usage data, broad usage data, for many users to come up with different changes to these types of search results, and where they're leveraging user personalization on a macro level, taking the aggregate of those things and creating new types of results, re-ranking things and adding snippets. I'll talk about each of those.

In these direct impacts, one of the most important ones to think about is location awareness. This is particularly important obviously if you're serving a local area, but you should be aware that location biases a lot of searches that may not have intended to be local simply by virtue of their geography. If you're at a point, if I'm here in downtown Seattle, there is location awareness that affects the results ordering. I can perform searches, for example for Coffee Works, and I will get these Seattle Coffee Works results.

Perhaps if I was in Portland, Oregon and they had a Coffee Works in Portland, I would be getting those Coffee Works results. Usage history also gives Google hints about your location, meaning that even if you're searching on your smartphone or searching on your laptop, and you said, "Don't share my location," Google and Bing will still try to figure this out, and they'll try to figure it out by looking at your search history. They'll say to themselves, "Hey, it looks like this user has previously done searches for Madison Markets, Seattle Trader Joe's, used our maps to get directions from Capitol Hill to Queen Anne. I can guess, based on that usage data, that you are in Seattle, and I will try and give you personalized results that essentially are tied to the location where I think you're at."

A fascinating example of this is I was searching on my desktop computer last night, which I have not made it location aware specifically, but I did a search for a particular arena in Dublin, which is where the DMX Conference, that I'm going to in a couple days and speaking at, is going to be held. Then I started typing in the name of the hotel I was at, and it's a brand name hotel. What do you know? That location came up, the Dublin location of the brand hotel, even though that hotel has locations all over the world. How do they know? They know because I just performed a search that was related to Dublin, Ireland, and therefore they're thinking, oh yeah, that's probably where he's looking for this hotel information as well. Very, very smart usage history based personalization.

Do be aware search suggest is also affected directly by personalization types of results. If you are doing a search that is going to be biased by some element of personalization, either your search history or your location, those kinds of things, auto-suggest will come up with those same biases as the rankings might.

Next, I want to talk about the semantics of how you perform queries and what you're seeking can affect your search as well. Search history is an important bias here, right? Basically, if I've been doing searches for jewelry, gemstones, wedding rings, those kinds of things, and I do a search for ruby, Google and Bing are pretty smart. They can realize, based on that history, that I probably mean ruby the stone, not Ruby the programming language. Likewise, if I've just done searches for Python, Pearl and Java, they might interpret that to mean, "Aha, this person is most likely, when they're searching for Ruby, looking for the programming language." This makes it very hard if you're a software engineer who's trying to look for gemstones, by the way. As you know, the ruby gem is not just a gem. It's also part of the programming protocol.

This gets very interesting. Even seemingly unrelated searches and behavior can modify the results, and I think this is Google showing their strength in pattern matching and machine learning. They essentially have interpreted, for example, as disparate things as me performing searches around the SEO world and them interpreting that to mean that I'm a technical person, and therefore as I do searches related to Ruby or Python, they don't think the snake or the gemstone. They think the programming language Python or the programming language Ruby, which is pretty interesting, connecting up what is essentially a marketing discipline, SEO a technical marketing discipline, and connecting up those programming languages. Very, very interesting. That can modify your results as well.

Your social connections. So social connections was a page that existed on Google until last year. In my opinion, it was a very important page and a frustrating page that they've now removed. The social connections page would show, based on the account you were inside of, all your contacts and how Google connected you to them and how they might influence your search results.

For example, it would say randfish@gmail.com,which is my Gmail account that I don't actually use, is connected to Danny Sullivan because Rand has emailed Danny Sullivan on that account, and therefore we have these accounts that Danny Sullivan has connected to Google in one way or another. In fact, his Facebook account and several other accounts were connected through his Quora account because Quora OAuths into those, and Google has an agreement or whatever, an auth system with Quora. You could see, wow, Google is exposing things that Danny Sullivan has shared on Facebook to me, not directly through Facebook, but through this protocol that they've got with Quora. That's fascinating. Those social connections can influence the content you're seeing, can influence the rankings where you see those things. So you may have never seen them before, they may have changed the rankings themselves, and they can also influence the snippets that you're seeing.

For example, when I see something that Danny Sullivan has Plus One'd or shared on Google+, or I see something that Darmesh Shah, for example, has shared on twitter, it will actually say, "Your friend, Darmesh, shared this," or "Your friend, Danny Sullivan, shared this," or "Danny Sullivan shared this." Then you can hover on that person and see some contact information about them. So fascinating ways that social connections are being used.

Big take-aways here, if you are a business and you're thinking about doing marketing and SEO, you have to be aware that these changes are taking place. It's not productive or valuable to get frustrated that not everyone is seeing the same auto-suggest results, the same results in the same order. You just have to be aware that, hey, if we're going to be in a location, that location could be biasing for us or against us, especially if you're not there or if something else is taking your place.

If people are performing searches that are related to topics that might have more than one meaning, you have to make sure that you feel like your audience is well tapped into and that they're performing searches that they are aware of your products getting more content out there that they might be searching for and building a bigger brand. Those things will certainly help. A lot of the offline branding kinds of things actually help considerably with this type of stuff.

Of course, social connections and making sure that your audience is sharing so that the audience connected to them, even if they're not your direct customers, this is why social media strategy is so much about not just reaching people who might buy from you, but all the people who might influence them. Remember that social connections will be influenced in this way. Right now, Google+ is the most powerful way and most direct way to do this, but certainly there are others as well as the now removed social connections page, helped show us.

What about some indirect impacts? There are actually a few of these that are worth mentioning as well. One of those indirect impacts that I think is very important is that you can see re-ranking of results, not just based on your usage, but this can happen or may happen, not for certain, but may happen based on patterns that the engines detect. If they're seeing that a large number of people are suddenly switching away from searching ruby the gemstone to Ruby the language, they might bias this by saying, "You know what, by default, we're going to show more results or more results higher up about Ruby the programming language."

If they're seeing, boy a lot of people in a lot of geographies, not just Seattle, when they perform a Coffee Works search, are actually looking for Seattle Coffee Works, because that brand has built itself up so strongly, you know what, we're going to start showing the Seattle Coffee Works location over the other ones because of the pattern matching that we're seeing. That pattern matching can be a very powerful thing, which is another great reason to build a great brand, have a lot of users, and get a lot of people around your product, your services, and your company.

Social shares, particularly what we've heard from the search engines, Bing's been a little more transparent about this than Google has, but what Bing has basically said is that with social shares, the trustworthiness, the quality, and the quantity of those shares may impact the rankings, too. This is not just on an individual basis. So they're not just saying, "Oh well, Danny Sullivan shared this thing with Rand, and so now we're going to show it to Rand." They're saying, "Boy, lots of people shared this particular result around this topic. Maybe we should be ranking that higher even though it doesn't have the classic signals." Those might be things like keywords, links, and all the other things, anchor text and other things that they're using the ranking algorithm. They might say, "Hey the social shares are such a powerful element here, and we're seeing so much of a pattern around this, that we're going to start re-ranking results based on that." Another great reason to get involved in social, even if you're just doing SEO.

Auto-suggest can be your friend. It can also be your enemy. But when you do a search today, Elijah and I just tried this, and do a search for Whiteboard space, they will fill in some links for you – paint, online, information. Then I did the same search on my phone, and what do you think? Whiteboard Friday was the second or third result there, meaning, they've seen that I've done searches around SEOmoz before and around SEO in general. So they're thinking, "Aha. You, Rand, you're a person who probably is interested in Whiteboard Friday, even though you haven't done that search before on this particular phone." I got a new phone recently.

That usage data and personalization is affecting how auto-suggest is suggesting or search suggest is working. Auto-suggest, by the way, is also location aware and location biased. For example, if you were to perform this search, whiteboard space, in Seattle, you probably would have a higher likelihood of getting Friday than in, let's say, Hong Kong, where Whiteboard Friday is not as popular generally. I know we have Hong Kong fans, and I appreciate you guys, of course. But those types of search suggests are based on the searches that are performed in a local region, and to the degree that Google or Bing can do it, they will bias those based on that, so you should be aware.

For example, if lots and lots of people in a particular location, and I have done this at conferences, it's actually really fun to ask the audience, "Hey, would everyone please perform this particular search," and then you look the next day, and that's the suggested search even though it hadn't been performed previously. They're looking at, "Oh, this is trending in this particular region." This was a conference in Portland, Oregon, where I tried this, a blogging conference, and it was really fun to see the next day that those results were popping up in that fashion.

Search queries. The search queries that you perform, but not just the ones the you perform, but the search queries as a whole, kind of in an indirect, amalgamated, pattern matching way, may also be used to form those topic models and co-occurrences or brand associations that we've discussed before, which can have an impact on how search results work and how SEO works. Meaning that, if lots of people start connecting up the phrase SEOmoz with SEO or SEOmoz with inbound marketing, or those kinds of things, it's very likely or you might well see that Google is actually ranking pages on that domain, on SEOmoz's domain, higher for those keywords because they've built an association.

Search queries, along with content, are one of the big ways that they put those topics together and try to figure out, "Oh yeah, look, it seems like people have a strong association with GE and washer/dryers, or with Leica and cameras or with the Gap and clothing." Therefore, when people perform those types of searches, we might want to surface those brands more frequently. You can see this in particular when you perform a lot of ecommerce-related searches and particular brands come up. If you do a search for outdoor clothing and things like Columbia Sportswear and REI and those types of brands are popping up as a suggestion, you get a strong sense of the types of connections that Google might build based on these things.

All right, everyone. I hope you've enjoyed this edition of Whiteboard Friday. I hope you have lots of great comments, and I would love to jump in there with you and suggestions on how you people can dig deeper. We will see you again next week."

Video transcription by Speechpad.com

Sign up for The Moz Top 10, a semimonthly mailer updating you on the top ten hottest pieces of SEO news, tips, and rad links uncovered by the Moz team. Think of it as your exclusive digest of stuff you don't have time to hunt down but want to read!

Announcing the Just-Discovered Links Report

Posted: 21 Mar 2013 02:00 AM PDT

Posted by The_Tela

Hey everyone, I'm Tela. I head up data planning at SEOmoz, working on our indexes, our Mozscape API, and other really fun technical and data-focused products. This is actually my first post on the blog, and I get to announce a brand new feature - fun!

One of the challenges inbound marketers face is knowing when a new link has surfaced. Today, we're thrilled to announce a new feature in Open Site Explorer that helps you discover new links within an hour of them going up on the web: the Just-Discovered Links report.

This report helps you capitalize on links while they're still fresh, see how your content is resonating through social channels, gauge overall sentiment of the links being shared, give you a head start on instant outreach campaigns, and scope out which links your competitors are getting. Just-Discovered Links is in beta, and you can find it in Open Site Explorer as a new tab on the right. Ready to learn more? Let's go!

What is the Just-Discovered Links report?

This report is driven by a new SEOmoz index that is independent from the Mozscape index, and is populated with URLs that are shared on Twitter. This means that if you would like to have a URL included in the index, just tweet it through any Twitter account.

One note: The cralwers respect robots.txt and politeness rules, which would prevent such URLs from being indexed. Also, we won't index URLs that return a 500 status code.

Who is it for?

Our toolsets and data sources are expanding to support a wider set of inbound marketing activities, but we designed Just-Discovered Links with link builders in mind.

Getting started

You can search Just-Discovered Links through the main search box on Open Site Explorer. Enter a domain, subdomain, or specific URL just as you would when using the Inbound Links report. Then select the Just-Discovered Links beta tab. The report gives PRO members up to 10,000 links with anchor text and the destination URL, as well as Domain Authority and Page Authority metrics.

One important note on Page Authority: we will generally not have a Page Authority score available for new URLs, and will show [No data] in this case. So, when you see [No data], it generally indicates a link on a new page.

You can also filter the results using many of the same filter drop-downs you are used to using in other reports in Open Site Explorer. These include followed and no-followed links, and 301s; as well as internal or external links, and links to specific pages or subdomains. Note: We recommend you start searches using the default "pages on this root domain" query, and refine your search from there.

How does it work?

When a link is tweeted, we crawl that URL within minutes. We also crawl all of the links on the page that have been tweeted. These URLs, their anchor text, and their meta data (such as nofollow, redirect, and more) are stored and indexed. It may take up to an hour for links to be retrieved, crawled, and indexed.

We were able to build this feature rapidly by reusing much of the technology stack from Fresh Web Explorer. The indexes and implementation are a little different, but the underlying technology is the same. Dan Lecocq, the lead engineer on both projects, recently wrote an excellent post explaining the crawling and indexing infrastructure we use for Fresh Web explorer.

There are a few notable differences: we don’t use a crawl scheduler because we just index tweeted URLs as they come in. That’s how we are able to include URLs quickly. Also, unlike Fresh Web Explorer, the Just-Discovered Links report is focused exclusively on anchor text and URLs, so we don’t do any de-chroming as that would mean excluding some links that could be valuable.

How is it different?

Freshness

Freshness of data continues to be a top priority when we design new products. We have traditionally released indexes on the timeframe of weeks. With this report, we have a new link index that is updated in about an hour. From weeks to an hour - wow! We'll be providing additional details in the future on what this means.

URL coverage

This index includes valuable links that may be high-quality and topically relevant to your site or specific URL but are new, and thus have a low Page Authority score. This means they may not be included in the Mozscape index until they have been established and earned their own links. With this new index, we expect to uncover high-quality links significantly faster than they would appear in Mozscape.

I want to clarify that we are not injecting URLs from the Just-Discovered Links report into our Mozscape index. We will be able to do this in the future, but we want to gather customer feedback and understand usage before connecting these two indexes. So for now, the indexes are completely separate.

How big is the index?

We have seeded the index and are adding new URLs as they are shared, but don’t yet have a full 30 days worth of data in the index. We are projecting that the index will include between 250 million and 300 million URLs when full. We keep adding data, and will be at full capacity in the next week.

How long will URLs stay in the index?

We are keeping URLs in the index for 30 days. After that, URLs will fall out of the index and not appear in the Just-Discovered Links report. However, you can tweet the URL and it will be included again.

How long does it take to index a URL?

We are able to crawl and include URLs in the live index within an hour of being shared on Twitter. You may see URLs appear in the report more quickly, but generally you can expect it to take about an hour.

Why did you choose Twitter as a data source?

About 10% of tweets include URLs, and many Twitter users share links as a primary activity. However, we would like to include other data sources that are of value. I’d love to hear from folks in the comments below on data sources they would like to see us consider for inclusion in this report.

How much data can I get?

The Just-Discovered Links report has the same usage limits as the Inbound Links report in Open Site Explorer. PRO customers can retrieve 10,000 results per day, community members can get 20 results, and guests can see the first five results.

What is “UTC” in the Date Crawled column?

We report time in UTC, or Coordinated Universal Time format. This time format will be familiar for our European customers, but might not be as familiar for customers in the states. The time zones for UTC are ahead of Eastern Standard Time, so US customers will see links where the time-stamp appears to be in the future, but this is really just a time zone issue. We can discover links quickly, but can’t predict links before they happen. Yet, anyways :)

CSV export

You can export a CSV with the results from your Just-Discovered Links report search. The CSV export will be limited to 5,000 links for now. We plan to increase this to 10,000 rows of data in the near future. We need to re-tool some of Open Site Explorer’s data storage infrastructure before we can offer a larger exports, and don’t have an exact ETA for this addition quite yet.

export search results

This is a beta release

We wanted to roll this out quickly so we can gather feedback from our customers on how they use this data, and on overall features. We have a survey where you can make suggestions for improving the feature and leave feedback. However, please keep in mind the fact that this is a beta when deciding how to use this data as part of your workflow. We may make changes based on feedback we get that result in changes to the reports.

Ready to find some links?

We’ve been releasing new versions of our Mozscape index about every two weeks. An index that is continuously updated within an hour is new for us, too, and we’re still learning how this can make a positive impact on your workflow. Just as with the release of Fresh Web Explorer, we would love to get feedback from you on how you use this report, as well as any issues that you uncover so we can address them quickly.

The report is live and ready to use now. Head on over to Open Site Explorer’s new Just-Discovered Links tab and get started!

vineri, 22 martie 2013

Personalization and SEO - Whiteboard Friday

Personalization and SEO - Whiteboard Friday

Video Transcription