In recent years, Google have diversified their portfolio of products and services in many ways. Just think of all of the different labels that fall under the Google brand:
That's a LOT of stuff. And there's a lot of other stuff they have going on also. So it's no surprise to learn that Google is closing two of their services: Google Health and the DoubleClick Ad Planner Marketplace .
The question is... Are Google going through a strategic realignment of the business. They have strayed very far from their core competencies with all of these add-on products and services. Some are very noble and interesting, but many are complete failures in market-uptake terms. Focusing on too many things makes it very easy to become scatter-brained and very difficult to achieve success in any one area.
Now that Larry Page is the CEO after taking the helm from Eric Schmidt, it will be interesting to see if and how Google's direction changes. Of course, Google have phenomenal brain-trust. I can't want to see how they use it.
Yesterday I posted a video showing a "prefetch" hit coming from the Google SERPs. The bad news is that this turns out to be nothing new; I thought it may have been related to the recent page previews feature - It wasn't.
The good news is that there's an opportunity here, which I don't think has been discussed before.
To summarize: When you use the integrated search box in Firefox (top right) to search Google, Google will sometimes use a rel=prefetch attribute on the first result link. This causes Firefox to automatically download the HTML source code of the first result page. Along with the request to download the page, the referrer string is sent. This is great, because we have a hit to our server without the visitor ever clicking on our result.
So, what is the opportunity and how can we take advantage of it?
Simple. Because we have a hit directly from a Google SERP (along with the referrer string), we can identify searches where our site is ranking. If we don't get the click, then we know we have a problem.
So, let's go through this quickly so we can all get back to work.
In the Apache HTTP server, tell the server that you want to track only prefetch hits to a separate log file:
SetEnvIf X-moz prefetch Prefetch_Request #Set a variable when prefetch CustomLog /path/to/prefetch_log combined env=Prefetch_Request #Log the request if prefetch
Then restart the server. This will log any requests that come to the server with an X-mox: prefetch header to the /path/to/prefetch_log log file.
You can look through the data in this log file to determine where you're showing up in SERPs. If you compare this data to your normal log file, you will be able to identify where you don't get the clicks from the Google SERP.
What you do with this information is up you. I know what I would do - optimize the title and meta description/snippet to maximize the click-through rate.
I found an odd issue with Google and Firefox today. I noticed that on some occasions, the first result on the SERP was GET'd by Firefox, despite the fact that I didn't visit the site. Cache was cleared, and it only happens when I use the firefox search bar.
Video:
(In listening to it myself again, I see that I messed up my lefts/rights when talking bout the windows... will be sure to pay more attention to that in the future :) )
Postscript: After some more digging and it appears that it is actually Google that is causing the prefetch. They do it by using the rel=prefetch attribute on links they want prefetched, and Firefox then goes and gets them.
That means that hits from Google could be inflated. For those with much higher traffic, the margin of error could be magnified.)
By Darrin J. Ward
Hi, my name is Darrin Ward and I'm addicted to speed. Speedy Websites, that is.
I've always been addicted to making the Internet go as fast as possible. I distinctly remember playing around with the physical positioning of my first 28.8k modem to see if moving it away from the electrical / magnetic interference of the computer and monitor would speed up the Internet, or if tinkering with all kinds of other software and hardware settings sped it up any.
In fact, I love speed so much that sometimes it could actually be considered for a flaw. If you've ever seen a Website that I personally have designed, then you'll know exactly what I mean. Although the Darrin Ward design team does fantastic designs, all of my personal designs tend to be stripped down to almost text-only sites, because I want things to load as fast as possible, and graphics and rich-media only slow it down.
It turns out that we've found a great balance in the team; My team comes up with visually stunning designs, and then I go to work on making the site load stunningly fast, with minimal sacrifice to the original design. It's a win-win.
Anyway. I also happen to practically live inside of Apache httpd.conf and .htaccess files (the configuration files for the Apache Web Server). So when I recently heard that Google released an apache module called mod_pagespeed, I was ecstatic.
The apache httpd module is for version 2.2 and higher, and it takes care of some of the items that Google's Page Speed Firefox plug-in addresses.
I haven't had time to experiment with mod_pagespeed yet, but it will be interesting to see how it holds up under high load. Because of our success with SEO and Internet Marketing, some of the sites we manage get a LOT of traffic (top site peaks at about 2,000 hits per second!). The other issue is that it's not yet listed as compatible with FreeBSD, which is my server OS of choice (I love the HTTP accept filter), but I'm sure it can be made to work.
The long and the short of it is this... Here at this SEO company, we are 100% committed to making Websites fast because we know it's important for visitors, and because it helps Google rankings now that page loading speed is part of the Google algorithm. We're happy to see that Google is providing Webmasters with tools to help make the Web faster (and save bandwidth).
Every now and then we are reminded how much we rely on Google.
Over the past couple of years, Google has done some experimentation with using AJAX search engine results pages. In doing so, such experiments have broken the keyword-tracking functionality of analytics tools because a full set of referrer data was not being sent (anything after the # in a URL is not part of the referrer string). Once Google were made aware of the issue, they rolled back and things went back to normal, where analytics packages could continue tracking referring keywords.
The visual change of the AJAX implementation was unnoticeable to an end-user, but the impact for Webmasters was immense. Earlier this week, it was noticed that Google was again not passing referrer strings. Matt Cutts of Google commented to let us know that the changes were an error on Google's part due to repurposing of old code. However, ever since the first implementation of the AJAX results pages, my Firefox installation on my Mac continues to show me AJAX results pages, not the regular ones. So, I have no doubt that there is still some Google traffic out there that is not sending referrers/keywords correctly.
It's just another reminder that if Google did want to make this change and not listen to us, they could really cause a big nightmare for all of us SEOs out there!
I sometimes use Google as a dictionary replacement, as I suspect a lot of people do. I search for the word on Google and then click on the "definition" link beside the word in the horizontal blue information bar. Google links to answers.com, which gives the definition:
What I don't understand is why Google hasn't licensed the content from the Oxford dictionary or some other dictionary and made their own dictionary function. Probably a better idea would be to license the content from multiple dictionaries to make sure they have all of the right definition variants, including those that are regionally specific.
Granted, Google does have the "define:keyword" operator that attempts to define words by scraping content from pages across the Web. But, anyone that has experimented with this function to any degree will tell you that it can be horridly inaccurate. I've often seen it pull definitions from adjacent words on pages, yielding a completely irrelevant definition.
It should be noted that, according to compete.com, Google is responsible for 61.19% of answers.com's traffic:
It's not clear how much of this comes from the definition links and how much comes from regular organic listings. Either way, that's a pretty significant share of the traffic.
Hey, Google... If I set up a dictionary site, will you link to me instead?!
I change the number of results-per-page setting quite a lot, and I just noticed that the "Preferences" link on the Google homepage that I normally use to change my results-per-page setting is missing. Instead, it is now located in a "Settings" drop-down menu up at the top, named "Search Settings".
Before, the "Preferences" link was to the right of the search box, I believe under "Language Tools". In the words of Stewie Griffin... "I don't like change".
Take a peek:
Edit: The "Settings" link at the top will be a drop-down to include both "Search Settings" and "Account Settings", if one if logged into an account. If one is not logged into an account, the "Settings" link will change to read "Search Settings", and it will go directly to the settings/preferences page.
I'm not sure if this is a new thing or not, but I just searched on Google for "staples.com" and I noticed that the Staples paid listing at the top was sporting sitelinks... something that I don't recall seeing before. Upon refresh, they were gone.
Take a look:
Here's a very small but interesting change. Google has changed the text label beside the checkbox on the Google account login form that keeps the user signed into their Google account. It's changed from "Remember me on this computer" to "Stay signed in".
I wonder if the previous label was confusing people. Good to see their experimenting with little usability things.
Here's how the login form looks now:
And here's what it used to look like:
Start your persistent cookies, get set. GO!
A thread over at WebmasterWorld has some reports of a PageRank (PR) update going on with many people seeing new PR values, just one month after Google did their last PR update.
It will be interesting to see how this plays out during the course of the day... is this an incremental update or a full PR update? Are Google returning to their monthly PR update cycle from many years ago? It's not impossible... as more new content is generated, people want to see the PR values for those pages. Leaving PR updates on cycles of just 6 months leaves huge gaps in the number of pages for which Google has missing PR values. That can reflect poorly on Google.
Google Blogoscoped has some screenshots of Google showing product results on the right side of the SERP underneath the Sponsored Links. This is almost certainly related to Google experimenting with commission-based advertising. I had seen screenshots of the product links up top, but not at the side.
I should note that Google has experimented with CPA (Cost Per Action/Acquisition) advertising before with AdSense Publishers back in 2006, but they stopped that program. My thinking at the time was that it turned out to be ineffective because the ads weren't targeted enough, so CPC probably yielded a larger income for Google and their publishers. But with direct products integrated into search pages, I'm thinking that they should actually work quite well. It will be interesting to see where this goes.
There has been a lot of talk this past week about how Google Voice is set to open up to the public soon. At the moment it's only available to Grand Central users - Grand Central being the name they have been using for the last few years. Fortunately, I've had a Grand Central account since the very beginning, so I'm now testing out the new Google Voice system. Here's a screenshot of the homepage / inbox: when you first join:
As you can see, this layout is very similar to GMail. You've got an Inbox, Starred, Spam, Trash and some other items specific to Google Voice like Voicemail, SMS, Recorded, Placed, Received and Missed.
The first thing I wanted to do was check out the transcript feature for voicemails, so I called my Google Voice number to leave a voicemail. This is what I said:
And here is what Google delivered to my Inbox:
Google also delivered a transcript of the message to my email Inbox:
Obviously, this is not a perfect transcript, but I have to say that it is still pretty accurate. I was disappointed however that Google didn't deliver a .wav file of the message like Vonage does.
Clicking on the "SMS" button in the top left will allow you to quickly send an SMS text message:
This message came through on my phone (BlackBerry) very quickly, but what I liked most was that it appeared to have come directly from my Google Voice number and the message did not have any labels or advertising indicating that the message was from Google Voice. It looks just like it came from another mobile phone:
When I replied to the text message from my mobile phone Google alerted me that I had 1 new message. This was a link which I had to click to see the reply in my Inbox:
I never use SMS messages, but I do use the BlackBerry PIN service. So I'm a little disappointed to find that this is not an option on Google Voice. I frankly don't understand why people use SMS messages. There's a character limitation (or you break up into multiple messages and get charged multiple times). BlackBerry PIN is free with an internet connection and there is no limitation, or at least none that I have ever run into.
Google Voice is painfully slow for me right now. Pageloads are taking upwards of 15 seconds. However this problem may be on my end; We have a new fiber connection in the building that doesn't have a reverse DNS yet, and this seems to cause huge delays with sites that require reverse DNS lookup.
Google recently acquired 1 million new phone numbers from Level 3, so they are obviously expected a surge in numbers soon. I think Google Voice is a cool service, but it doesn't add anything that I don't already have. You will still need another phone to make outgoing calls and the SMS functionality for me is useless (though I might use SMS more if I can proxy it through email or BlackBerry PIN so I don't have to pay.)
Please share your thoughts on Google Voice!
Barry Schwartz at the Search Engine Roundtable has some screenshots of Google SERP listings where some of the long URLs are truncated to be only one line.
It doesn't seem to be happening for all URLs on all SERPS, but it's definitely happening for some of them. A quick look seems to indicate that Google are more likely to truncate part of a URL path (the folder names) instead of the filenames. They always seem to keep the beginning of the filename (whether it's located in the root or in a folder/directory), but they sometimes truncate the end of the filename.
Gotta love Google with incremental changes!
China is at it again. They have disabled part of the Google search engine that they claim displays "pornographic and vulgar content", according to the New York Times:
I'm not clear on whether Google disabled it themselves voluntarily or if China disabled it by force in some way.
Amid the ongoing Iranian election madness, Google has added Persian (Farsi) as an available language to their Google Translate tool, according to the Google Blog. The language is currently in "Alpha" status, meaning in the very early stages of testing (pre-beta).
Rand Fishkin of SEOmoz recently posted on the issue of rel=nofollow changes at Google, which prevent Pagerank being saved and sent through to the other links on the page (sculpting). The post comes after a blog entry by Matt Cutts on the topic.
I agree that it can be frustrating when Google reverses direction, but I've always maintained my belief that using rel=nofollow to sculpt Pagerank was a bad idea and counter to the idea of Pagerank in itself. I just plain don't like rel=nofollow, unless it used as a purely protective measure for potentially spammy UGC (User Generated Content).
Rand says one or two things that I disagree with:
"I'm saddened to say that given this change, we, as SEOs, are going to have to also recommend the best practice that comments (in all forms of UGC) no longer accept links ... Comments that contain links, unfortunately, will actively detract from a site's ability to get pages indexed (as they'll pull away link juice from the places that need it)"
The web has changed, and the notion of preserving Pagerank that existing in the early 2000's is gone. What matters today is not so much the Pagerank of pages, but how pages are intertwined, themed or fused with pages on the same topic or closely related topics.
A blanket statement that links in UGC should be ignored is a terrible idea. UGC often introduces some quality links that can actually add to the ranking potential of the page. Granted however, there is a risk of spam links causing a negative impact on rankings, but sites should be sufficiently moderated to prevent such links. If quality control and moderating isn't in place, then how trustworthy is the site anyway? Besides, rel=nofollow can still be used. Worrying about saving the Pagerank is not a big thing in my view.
"From now on, if you wish to sculpt PageRank, you'll want to use one of the following classic PR sculpting methodologies:"
And Rand goes on to list some of the old-school methodologies to prevent links being displayed directly on the page. There are a couple of problems with these trickery ideas:
I agree with Rand in that this a pretty big shift in how Google handles PR flow, but I would certainly say that it was a very predictable move. In hindsight, I'm very glad that my team has never adopted the notion of PR sculpting with rel=nofollow. We have other more practical methods.
PS. My blog CMS is a custom solution and I haven't gotten around to implementing links in comments yet, but I do plan on doing it. When I do, the links probably won't use rel=nofollow (I haven't thought much about it yet), but I will be removing spam links, so I'm not too worried.
by Darrin J. Ward:
Although this is not a new topic, we still see a lot of people attempting to trick search engine robots by using JavaScript "include" files in order to perform nasty redirects or to set a particular element's visibility:none (via JS or CSS) - thus making it invisible to users but visible to search engine robots. Similarly, we see a lot of people that use CSS to set the H (H1, H2, etc) family of elements to a much smaller font size than that of their natural appearance.
The fundamental premise of such implementations is normally that the search engines do not actively look at these include files, thus the "tricks" will remain uncovered. Such an assumption would be incorrect.
It's not exactly "new news" that search engine crawlers do indeed look at these files. In fact, I distinctly remember posting about Google's crawler making hits on .js (JavaScript) and .css (Style sheet) files years ago, literally. To prove it: Here are some hits taken from the raw access log files from this very blog (which has only on this new domain for a number of days):
66.249.72.20 - - [16/Jul/2007:18:27:44 -0400] "GET /inc/js/share-this.js HTTP/1.1" 200 1178 "-" "Mozilla/5.0 (compatible; Googlebot/2.1; +http://www.google.com/bot.html)"
66.249.72.20 - - [16/Jul/2007:18:28:40 -0400] "GET /inc/js/prototype.js HTTP/1.1" 200 14471 "-" "Mozilla/5.0 (compatible; Googlebot/2.1; +http://www.google.com/bot.html)"
66.249.72.20 - - [15/Jul/2007:19:15:47 -0400] "GET /inc/css/styles.css HTTP/1.1" 200 2180 "-" "Mozilla/5.0 (compatible; Googlebot/2.1; +http://www.google.com/bot.html)"
66.249.72.20 - - [17/Jul/2007:22:22:24 -0400] "GET /inc/css/styles.css HTTP/1.1" 200 2264 "-" "Mozilla/5.0 (compatible; Googlebot/2.1; +http://www.google.com/bot.html)"
How do I know that these are actually from Google and not a fake? Simple. I take the IP 66.249.72.20 over to the ARIN IP Whois tool and see who owns that IP. OK, so how did I actually get to see these entries? Well, even though SpyderTrax is a great tool for checking on robot activity at the page level, it doesn't show details on hits to these .css and .js files. So. I logged into my server via SSH and performed the following command on my access log:
grep 'Googlebot' FILENAME | grep '.js|.css'
This command shows me all of the hits that contain "Googlebot" along with either ".js" or ".css". If I only wanted to see one or the other, the I would use only '.js' or '.css' for that last part (not escaping backslash).
So, in writing this, I was wondering when all of this activity actually started. I know it's been going on for years. Lucky for me, I'm a fanatical Analytics fan (not Google Analytics), and I know the value of being able to retroactively look at Key Performance Indicators - So I have log files dating back to the start of 2003. And, not just on one site - but on enough sites to actually take a peep and see when Google's activity might have started. So I did.
64.68.89.138 - - [23/Mar/2004:22:49:08 -0500] "GET /includes/js/nav/menu_com.js HTTP/1.1" 200 21960 "-" "Googlebot/Test"
64.68.89.138 - - [24/Mar/2004:01:05:20 -0500] "GET /includes/js/functions.js HTTP/1.1" 200 1085 "-" "Googlebot/Test"
64.68.89.167 - - [25/Mar/2004:08:32:50 -0500] "GET /includes/js/nav/exmplmenu_var.js HTTP/1.1" 200 3406 "-" "Googlebot/Test"
64.68.89.182 - - [25/Mar/2004:21:31:06 -0500] "GET /includes/js/nav/menu_com.js HTTP/1.1" 200 21960 "-" "Googlebot/Test"
el64.68.89.182 - - [26/Mar/2004:04:09:04 -0500] "GET /includes/js/functions.js HTTP/1.1" 200 1085 "-" "Googlebot/Test"
These are the first hits that I have tracked from Googlebot. Admittedly, I only looked at 2004, because I started with that year and the log files were so large they took forever to process. Obviously, you can see that they were using the "Googlebot/Test" User-agent then. But an IP Whois confirms that it's Google's IP block. So - it would appear as though there was a three day test or so going on at that point. One week before my birthday.
I had intentions on stripping out all valid hits from a Googlebot over the last few years and plotting a graph to show activity levels, but there's work to be done and I'm not sure something like that would have all that much value, even if it is super-interesting.
So what is the take-away from all of this nonsense? Simple. Take a look through your JavaScript and CSS files to make sure that they validate and that there's no functions that might accidentally perform redirects in what might be considered a sneaky way. I'm not worried necessarily about re-styling H1 or H2 tags with CSS - I do that myself. However, I wouldn't ever have them 100 pixels off screen or invisibly small, because that's obviously very easy to detect. Of course - only the big G know's that they do with those files for sure!
by Darrin J. Ward:
As my mind wandered the other day (as it does from time to time), I started thinking a little bit about Google's "Master Plan". One of the notions that struck me was that Google Checkout is probably a very critical step in Google's plan to dominate the web - more than I had previously acknowledged. So it probably should be no big surprise to see it being promoted as heavily as this on their Results Pages (RPs):
Google's RP with multiple Google Checkout Buttons!
I remember spotting that blue button way before it was talked about in any of the forums, so it's not as though I've never seen it before or that I'm not accustomed to it. But having that many on single page is a little bit too much for me. I think the green little speedy shopping cart was subtle enough - but Google Checkout hasn't exactly taken off as I'm sure was planned - so it must be time to give it the additional push. Shameless Self Promotion!!
edit: A couple of refreshes later, and I managed to get 6 buttons as opposed to the 5 above. I wonder if there is a hard limit, or if it will gladly show a checkout button for every ad?!