Earlier this year, EggHeadCafe hired an agency to review the entire site. It had
been through all sorts of changes over the last 10 years but noone other than
myself or Peter had reviewed what we were doing from a search engine optimization
as well as a usability perspective. In 2011, a lot has changed in the search
engine world especially with Google. They have pretty much turned the world upside
down and sideways via changes to their algorithms most noteably through updates
code named Panda. Changes in Google Adsense and the general advertising market
have also been felt here at EggHeadCafe. Some due to the economy and how it affects
Information Technology departments and vendors as well as how visitors interact
with ads.
We figured with all of these changes going on, it was time to conduct a long overdue
external review. What we have found is that the hundreds of thousands of backlinks
from other forums, articles, rss feed publications, etc. have drastically changed
in ranking value from previous years. The whole landscape has changed. In fact,
if you look at the whole tech forum landscape, you'll see a negative trend. EggHeadCafe was also provided with a great deal of information as to how social
media such as Twitter and Facebook play a much bigger role in search rankings
than they did even one year ago. We do utilize Twitter and Facebook but have
yet to form a strategy for getting the most out of them.
As part of the external review, I also started to look closely at a typical daily
visitor to EggHeadCafe. Our visitors almost always come from a search engine
query (usually from Google - very little love from Bing) where an IT professional
has a problem or question to address and they are looking for a specific answer
to that question. Once they find their answer (hopefully), they quickly return
to their task and leave the site. It was this simple concept that has driven
much of the new user interface to EggHeadCafe. Make the pages more readable and
have the content focus only on the issue at hand. This meant making better use
of screen real estate and either moving or removing extraneous things like site
navigation, biographies, and author photos. As Joe Friday used to say on Dragnet,
"just the facts ma'am". In our first release, there have been some
successes and some failures that I'll touch on more below.
To help visitors out, in 2010 we introduced a well received feature for related items
at the bottom of content pages. We created these by using dtsearch to manage
an internal search engine with snippet extraction capabilities on all EggHeadCafe content pages. On those pages that our algos did
a good job of determining keywords, this feature worked extremely well. On others,
not so much. We received numerous comments from members as well as non-members
that this was a useful feature because they could quickly find other code samples
that referenced a similar method, event, property, or functionality that originally
brought them to the page. This was particularly useful with rarer keywords. Unfortunately,
in 2011, Google's algos started seeing these related item sections as on-site
duplicate content. So, we were forced to remove a useful feature to serve the
Google gods. Insert colorful adjective of choice here to express my feelings
on that...
In an attempt to attack the related item concept from a different angle, I started
tracking all of the keyword search phrases and words used to find a given page.
When the page is rendered, my code picks up the top 7 phrases or words based
on visitor usage and displays them at the top of the page in the form of links
to our own search engine. Essentially, I'm letting our visitors tell use what
is important on the page. My original intent was to display the total number
of pages on EggHeadCafe that contained that word or phrase but the performance
just wasn't fast enough on the fly. So, the counts you see are visitor counts.
I may revisit this in the future as I think the total page count on a given subject
is of far more use to the visitor.
As more and more visitors find the page from a Google search, the more accurate the
keyword targeting. While it has only been out for a week or so, our page views
versus visits is up about 10% even though the site navigation has been removed
from the view of non signed in members. I expect that to improve as our search
keyword history improves. One interesting thing that has come from this endeavor
is that people are horrible spellers. However, Google's algos seem to understand
visitor intent. The keyword history also gives you a feel for how people search
specific subjects. Over time, we hope to write content that targets these behaviors
better.
Internal Site Adjustments
The internal factors that needed to be addressed focus largely on url seo optimization
across the site, reduce on-site content duplication, drastically reduce the total
number of urls indexed in the search engines, and reducing our html code to human
readable text matrix.
Reduce Total Urls Indexed
We've identified via Google Webmaster Tools that our listing pages (and their myriad of different filter options) have created
millions of urls that are indexed. In previous years, this really didn't have
much of an affect on the ranking of the overall site. In 2011, that has changed.
To compound the problem for us, we had dynamically created forum post summaries
and used them on these listing pages. The net affect of that is that Google now
sees those summaries (and they myriad different ways they can be found) as on-site
duplicate content. That means that millions of urls on EggHeadCafe were seen
as problem pages for Google. To address this issue, I've done a couple of things.
First, remove the summaries from the listing pages and alter the page layout.
Second, take steps to suggest that the search engines not index the listing pages
at all by using a meta tag for no index.
Improve Url SEO Optimization
In some places on the site, we were already adding the topic to the url but not all.
That has been changed. As a general rule, you want to get as many keywords that
actually exist in the page in your url while still keeping the url as small as possible. We do a combination
of keywords as folder names as well as dash delimited keywords as filenames.
The one area I wanted to change was to remove the Guids from the article and
faq urls but the level of effort was just to great. Whenever you request a content
page on EggHeadCafe, our code compares the url used with what the url should
be and issues a Response.RedirectPermanent(validUrl, false); call when they don't match. So far, this seems to work properly. We've noticed that
it takes awhile for Google WebMaster Tools to pick this up though.
Site Speed
We've moved our page logging from the page load to an offline process. We significantly
reduced the amount of css and total number of images used and moved virtually
all of our fixed images to static.eggheadcafe.com where no logging occurs. Moved
css and scripts to folders where we could disable IIS logging. Tuned a few major
queries. Google Web Master Tools already shows a steady trend towards faster
page load times.
Html Code To Human Readable Text Matrix
Ultimately, this meant moving away from our knarly use of nested tables within nested
tables that have been used for 10+ years now. Tables work really well for the
wide variety of screen resolutions, browsers and browser versions, and constraining
the effects of html pasted into our forum posts from members. It was strongly
suggested to move our content pages (forum threads, faqs, and articles) to a
DIV tag based approach and try to use as few of them as possible. Our articles
and faqs were the easiest to change. The forum posts were an entirely different
animal.
The DIV tag based approach exposed serious issues with individual builds of IE compounded
by differences with Compatibility View settings, and differences between older
and newer versions of FireFox and Chrome. One of the bigger issues that affected
numerous areas of the site was how different builds of IE (even differences across
builds of IE 8) rendered DIV tags. In particular, hidden or collapsed DIV tags
with HTML in them. DIV tag height and width properties in a "float"
type layout scheme are important. Not all browsers will automatically render
the width to 100% or the height appropriately by default. If you see wrapping
of text where you wouldn't expect it, this is likely the issue.
The other major issue with a DIV tag approach is that it exposes unknown problems
with pasted in HTML content in our forums. We clean out hazardous stuff but do
not attempt to address the thousands of different things that could hose up a
page. This is most common when a member pastes in html from another site (which
we strongly discourage) to answer a question. Tables that contained this same
HTML constrained the rendering much better in more browsers.
In an IT shop setting, one would need a couple of dozen virtual machines configured
to test all of these. We don't have that here at EggHeadCafe and to be honest,
I was just too lazy to do it. So, over 300 posts later from our EggHeadCafe members,
I was able to get most of the issues fixed within a few days. Some of these issues
(namely the Telerik editor we use for our forum posts) are not going to be addressed
for older browsers. It is crap like this that drives me to Silverlight for web
based business applications.
Advertising (got to pay the bills)
To improve the performance of our image and text ads, I created a page layout that
is considerably different than what you see in most forums. Today, that view
is only available to non signed in members. The idea is to remove all extranous
links, names, dates, buttons, merit scores, etc... from the vertical view of
the thread. This makes it read much like a piece of paper. I've found over the
years that advertising performs much better when the eye is not distracted by
other content, graphics, links, etc... This is especially the case with text
ads.
On EggHeadCafe forum threads, you'll see that we have a large block of text ads (up
to 10) in the form of a list (signed in members don't see these ads). You'd think
this would have the potential to really irritate visitors. However, the click
thru percentage is twice what a typical block of 4 or 5 ads would be. This is
reflected across the board regardless of the subject matter. In fact, the larger
the base of related/applicable/on target ads for a given subject, that percentage
goes up. The really irritating part of using contextual based ads and their algo
is that you often get unrelated ads show up higher in the list. So, if you restrict
your ad block to 4 or 5 ads, you often get a block of completely irrelevant ads.
Thus, your click thru percentage goes through the floor. By showing up to 10,
you've got a much better shot that even a lower paying ad will get clicked because
it is relevant. In the first couple of weeks, the click thru percentages are
up in the new layout.
Our image based ads have largely been restricted to our new fixed header. This brings
the readable content closer to the top of the page and reduces wasted screen
real estate. In most months, we run banners for giveaways in this space upwards
of 25% of the time, so I wanted to keep this visible as much as possible.
Forum Issues
Our signed in members quickly told me that just because one layout makes reading
posts easier doesn't mean it makes the forum easier to use. During development,
I didn't see the new layout as an issue but our members felt differently. Since
they are more important, I quickly put out a different view for signed in members.
Going forward, serious consideration is being given to permitting members to
customize how threads are viewed. They would only see there own view. So, if
they screwed it up, only they would be affected. EggHeadCafe has such a diverse
member base that coming up with a single view that works for people is almost
impossible.
Logo
I got so many emails from an old image I did with Dr. Dotnetsky drinking a martini
and one of my own personal catch phrases "will code for booze" that
I just had to do something with it. Neado Designs was given the simple assignment
of creating a logo with an egg and a martini. What you see here is the first
crack. Peter and I loved it so it stuck.
Summary
Well, this gives you some insight as to why I did the things I did. Still got some
work to do but it has given us a bit of a boost in the areas I wanted to address
quickly. Thanks again for all the help from our dedicated members! I tried to
reward alot of you with forum content points for taking the time to submit bugs
and screen shots.