January, 2013 - The Internet Marketing Driver

Last week, Facebook unveiled Facebook Graph Search, which is an important step forward for the social network. Many in digital marketing, including myself, wondered why Facebook had never added a serious search feature to its platform. It made no sense… Until now, Facebook’s search functionality has been horrible. I often joked that you couldn’t find yourself via Facebook search… :)

By the way, I also wrote a post in October titled “BeastRank”, after Mark Zuckerberg hinted at a Facebook Search Engine. You should check out that post to find out the various ways that Facebook could rank its social search results (based on the enormous amount of data it knows about all of us).

Now that Graph Search has arrived, Facebook finally has the opportunity to better monetize its massive user-base (eventually). Note, Graph search is only in an early form and will absolutely evolve over time. And no, it is not being monetized yet. But you better believe that’s in the works.

What I’m Going to Cover
Facebook Graph Search is definitely an interesting take on social search. I’ve been testing it heavily over the past few days and will cover several of my findings in the post below. You can currently search for people, places, photos, and interests, and I dug into those areas in detail over the past few days.

Note, I’m not going to cover Graph Search basic functionality, or a tutorial on how to use the new search features. There are plenty of posts covering the basics. Instead, I’m going to cover certain aspects of Graph Search that relate to privacy, adoption, SEO, speed, etc. Let’s jump in.

Autocomplete and the Facebook Graph Search Funnel
Once you start using Graph Search, you’ll notice that there are autocomplete suggestions as you start to type. These are essential for Facebook, since it’s not 100% intuitive how you are supposed to structure queries using Graph Search. All of us are used to Google, where you can basically enter anything and have the search engine return results in seconds. I found myself wanting to enter the exact triggers that would surface the right Facebook data. That could be a challenge for Facebook.

If you perform a search that would trigger Graph Search results, then Facebook will offer its guidance. For example “Restaurants in {location}” will auto-populate with locations based on your profile, and will include the Graph Search icon for restaurants next to the suggestion. If you search outside of Graph Search, Bing’s autocomplete takes over (since Bing provides backup for Graph Search).

After performing hundreds of searches over the past few days, I found that Facebook definitely helps you out and starts to rework your searches to fit the Graph Search lingo. For example, if you simply enter a location and restaurants, Graph Search changes that to the “Restaurants in {location}” format. I think the average person might be confused with what they can search for, and how to best surface the information. I know I found it a little confusing, and I’m neck deep in search… That makes me think my mom will have no idea what she is doing. :)

Speed is Important, and Delays Matter
One thing I noticed is that the speed that Facebook returned the results lagged sometimes. And when you are used to a fraction of a second with Google, a few seconds (or longer) felt like a lifetime. I think it’s just another example of how we take Google for granted. If you spend some time in front of Graph Search, I think you’ll appreciate the speed at which Google moves (and for extremely advanced queries).

There were several times that Facebook just hung after I conducted search… I literally viewed a blank screen for a while until the results were returned. Again, this will kill the user experience for Facebook Search, who will undoubtedly leave for Google’s near-instant results.

Local Graph Search – Places, Restaurants, etc.
I found place search pretty cool, but often ran into situations where Facebook didn’t have enough data to provide thorough results. Just a straight search for “Restaurants near Princeton” returned some good results (along with social annotations). I saw a lot of my favorites on the list, but I didn’t need Facebook to know them… There’s a big difference right now between Graph Search and just using Yelp’s app on my phone. Yelp has a ton of data, reviews, etc. Graph Search seems lacking certain areas, and especially outside of the core categories like Restaurants.

If I searched for “restaurants liked by my friends”, I started to see some results that were interesting. I just might want to take a closer look at several of those restaurants. And that is the power of Graph Search, if it can consistently provide recommendations based on my friends’ data. But, there needs to be enough data, and that information needs to be helpful. That’s different than simply returning restaurants that my friends “liked”.

From the results page, I can like the restaurant, view it on a map, view photos taken by people, and also view visitors of the restaurant. This is where some privacy concerns popped up… Good segue.

Places and Privacy (Photos and Visitors)
When you click “photos in” from the local search results page, you will see all photos taken at the restaurant (or place). By the way, almost all of those photos are from people I don’t know, I’m not friends with, etc. Sure, the photos were shared publicly, but I couldn’t help but question if those people even knew the photos were shared publicly. And now the photos are available to anyone searching Facebook (via Graph Search).

For some restaurants, there were hundreds of photos of people and I could easily see their names, click through to their profiles, like their photo (creepy), view comments and likes, etc. Again, I’m not sure all of those people understand this is happening. Oh yeah, there were many photos of kids too… Way to go parents. You just posted public photos of your kids for the world to see. Ugh.

Visitors and Privacy
If I click “Visitors” when viewing a list of restaurants, Facebook will show me people that visited the restaurant (via check-ins). On that page, I can add them as a friend, message them (creepy), or view their photos, friends, and interests. There is also a faceted navigation that lets me drill into visitors by employer, city, school, hometown, etc. Yes, even more concern over privacy here.

Faceted Navigation, Graph Search-Style
When using Graph Search, you can often drill deeper into the search results via a faceted navigation. For example, when you search for a place, the navigation is present by default and enables you to fine-tune your search. When I searched for, “Restaurants near Princeton my friends like”, the faceted navigation enables me to choose other types of places like bars, hotels, gyms, parks, etc.

I can also choose which friends I want to focus on (if I know certain friends like similar things). I can also change the location, select “visited by” and then choose a friend.

Cool, But How Many Will Use The Faceted Navigation?
If you like shopping online and enjoy using a faceted navigation to refine your searches, then you’ll like this functionality. I definitely found it useful and quickly began digging into more Facebook data. That said, I’m not sure how many people will use it like I did… When you think of the average user (who is spoiled by Google doing everything for them), they might not want to refine their searches. They might expect Graph Search to be smart enough to refine it for them. Time will tell how many people use the faceted navigation in Graph Search.

Online Reputation Management (ORM) and Graph Search
I’ve worked on a lot of online reputation management projects over the years, so I started to dig into that aspect of Graph Search. If you type someone’s name, autocomplete first tries to display your friends’ names. But as you select a friends’ name or complete their name in the search bar, the autocomplete suggestions focus entirely on that person. You’ll see “friends of”, “photos of”, etc.

“See more”, and I mean more…
So, Graph Search provides autocomplete suggestions for specific people, but what happens when you click the “see more” link at the bottom of autocomplete? Well, when you click the “see more” link, autocomplete suggestions prompt you to search for all sorts of interests from the person in question. That includes music, sports, pages, books, groups, apps, etc.

That’s cool, but what’s that further down the list? Oh, religious views and political views! “Danger, Will Robinson!” At some point in your life, you learn that politics and religion are extremely polarizing topics. Well, how about if the world could easily drill into that data about you? Using Graph Search, they potentially could…

So, if you don’t have your privacy settings set up correctly, the world can now view your political and religious affiliations. If you’re cool with that, fine. If you just fell out of your seat in horror, then go make changes now. :) And to make drilling into the data even easier, there’s a nice dropdown on the page that enables me to find out the religious and political affiliations of any of my friends.

Graph Search and “Not Provided” – Hiding the Referrer
As I mentioned earlier, Graph Search falls back to Bing’s search results when it can’t meet a certain query. But how does that data get passed to the destination site? Will we begin to see searches from Graph Search in our analytics packages, and can we view the keywords people are searching for? That would be logical, right? I mean what kind of search engine would hide keywords from webmasters?? :)

Well, you aren’t going to be happy with the answer. I checked the referrer leading from Facebook Graph Search to websites, and it unfortunately looks like any other visit from Facebook. So I guess we have our own version of “not provided” from Graph Search. :) Come on Facebook! Pass the query along in the referrer… SEO’s will love you for it. Until they change that, you will see a standard referral from Facebook even when that person used Graph Search.

Click the image below to see a larger version of the referrer:

It’s worth noting that Facebook Graph Search almost always keeps you within Facebook. It’s only when the results fall back to Bing that you get standard search results that take you off the site. That’s when I think Facebook should send the referrer.

Summary – Pay Attention to the Details
Graph Search is big news, even if this beginning state isn’t perfect. Sponsored results (advertising) can turn into a lot of revenue for Facebook, as long as the search experience is strong and the results are compelling. But since it’s social in nature, and based on our data, Facebook has to be really careful with privacy. Based on my research so far, I wouldn’t give Graph Search outstanding scores on that front. In addition, the user experience needs to be faster, and improve, in order to meet the Google standard. Sure, it’s a new twist on search, but Facebook can’t take a step back with performance.

Needless to say, I’m eager to see how Graph Search evolves. And I’ll be closely watching adoption, user experience, privacy, the quality of the search results, etc. I’ll be writing more posts on the subject in the coming months, so definitely check back often. Now excuse me while I find a Chinese restaurant near Princeton that my friends like. :)

Redirect Glitch Causing SEO Problems

With the release of index status in Google Webmaster Tools, many webmasters are now questioning why their “not selected” numbers are high. They wonder if those numbers are good, bad, normal, etc? Unfortunately, there’s not an easy answer to that question, since it depends on the site at hand. But, you can definitely look at the ratio of “not selected” to pages indexed to start to understand if there is a technical problem causing a spike in pages being categorized as “not selected”.

For example, if you have 200 pages indexed on your site, and you see 350 categorized as “not selected”, that might be ok. But if you see 25K pages as “not selected” or more, then that could raise a red flag that something may not be right with the site… For example, is there a site structure issue that’s causing thousands of variations of pages with extremely similar content (duplicate content)?

A Recent Example of a Poor “Not Selected” Ratio
During SEO audits, there are times I come across significant problems like the one I mentioned above. And those problems could be inhibiting a company’s search efforts (to say the least). During a recent SEO audit, I came across a very interesting situation. Index status revealed an extremely high number of “not selected” pages (as compared to the number of pages indexed) and I found myself digging into the site to find out why.

High Percentage of "Not Selected" Pages in Index Status

I found several issues causing the problem, so there wasn’t just one issue pumping up the number. That said, the problem I’m going to cover today was causing thousands of duplicate pages to be created, and without the site owner knowing. The more pages I checked, the more duplicates I found. And this is a problem that can easily slip through the cracks for many webmasters. And that’s especially the case if a small or medium sized business is handling all website development on its own. Below, I’m going to cover what I found, and more importantly, how you can avoid the problem in the first place.

The Danger of an Extra Character
As I was analyzing the site manually, and via a number of test crawls, I came across some URL’s that contained an extra character. Specifically, the extra character was being appended to each canonical URL. All of those URL’s were from one specific section of the site (which contained thousands of URL’s). After digging into that section of the site, I found out that this problem was happening to almost every URL being linked to from a certain element within each page. So, I honed in on that element within each page to find out how the duplicate pages were being created. And by the way, it just so happens that the section of the site contains nearly 200K pages. Yes, this was a huge problem that was uncovered.

How One Extra Character Created Thousands of Duplicate Pages

The Result – Duplicate Content Ad Infinitum
The core problem is that the extra character created a new URL, but that new URL was an exact duplicate of the canonical URL (the URL that should be resolving). And as you can guess, both pages can be accessed on the site. One part of the site links to the canonical version of the pages, while this problematic section linked to the duplicate versions of the pages.

So, right off the bat, we are dealing with at least 200K duplicate pages. In addition, as more content is added to this section, more duplicate pages will be created over time (based on the extra character being added to each URL). Also, the canonical URL tag was not being used on the duplicate pages, so that wasn’t helping this specific case. And on that note, I wouldn’t advocate using the canonical URL tag to fix this problem… Technical problems like this should be addressed at the code or structure level.

1 to 1 Ratio of Duplicate URL's to Canonical URL's

So, if this was left in place, this problem could generate an unlimited number of duplicate pages. If 500K pages ended up there, then there would be 500K pages of duplicate content. Not good, so I dug deeper to find out exactly what was causing the problem.

The Root Problem – Faulty Redirects
Let’s face it, we all need to implement redirects at some point. And that introduces the possibility of a poor implementation, which can be catastrophic SEO-wise. It’s one of the reasons that website redesigns and CMS migrations are so risky. On that note, to learn how to avoid SEO disaster during a redesign or migration, you should check out my Search Engine Journal column on the subject.

For example, using 302’s versus 301’s, using meta refresh redirects, redirecting to the wrong pages, or having the redirect code bomb the URL’s. For this situation, we ran into the “bombing of URL’s” problem. The redirects were faulty, and were redirecting to URL’s with an extra character.

The Solution – Fix the Redirect Code!
So, hundreds of thousands of duplicate pages were being generated, and it was due to one piece of redirect code on the server. The 301 redirects being generated simply added an extra character to the destination URL. That’s it. The fix will be implemented soon, and once the new redirect code is rolled out, the correct URL’s will resolve.

This situation underscores the fact that even one small piece of code could have serious implications SEO-wise. If this situation was left unchanged, it could have ended up generating an unlimited number of duplicate pages. Knowing the content on this site, my guess is the problem would have generated 500K-750K pages of duplicate content over the next 2-3 years.

How To Avoid This Situation
After reading this post, you might be scared that this could happen to you, or worse, that it’s happening right now. I’m going to provide a short list of things you can do to make sure this doesn’t happen. Of course, if you feel you are having problems already, you should have an SEO audit performed.

First, whenever you create redirects, make sure you have a system for testing those redirects before they launch. You can do this a number of ways, including on a local server or test server prior to releasing the final code to production. If you thoroughly test the redirects, you could nip serious problems in the bud.
Second, make sure your xml sitemaps contain the canonical url’s for the pages at hand. Making sure you are feeding Google and Bing the correct URL’s can help them understand which ones should be considered the canonical url’s.
Third, you should develop a strategy for using the canonical URL tag on the site. If the tag is present, then you can ensure that any duplicate pages pass their search equity to the canonical URL’s. Note, I’m not saying that you should leave a technical problem in place! Instead, I’m saying that having the canonical URL tag in place will make sure the engines pass any search equity to the correct pages on your site while you figure out solutions to your technical problems.
This final bullet assumes you are already experiencing problems with duplicate content from a technical problem. If you are, and you cannot determine what’s going on, then invest in having a technical SEO audit completed. To me, the provide the most bang for your SEO buck. It’s a great way to find out what’s truly going on with your site (beyond just the problem I listed here).

Summary – Know Your Site
This case emphasizes something I’ve said a thousand times over the past few years. It’s incredibly important to have a sound site structure in order to perform at your highest level SEO-wise. Coding problems, site structure issues, flawed redirects, etc. can kill your SEO efforts. It’s one of the reasons that I believe SEO audits are critically important. They can catch all types of SEO issues and provide ways to remedy those problems. You know, like generating hundreds of thousands of duplicate pages. :)

Archives for January 2013

Faulty Redirects, Duplicate Content, and SEO – How a Redirect Glitch Created Hundreds of Thousands of Duplicate Pages