Prospective client: Help! I’ve fallen, and I can’t get up!
Me: Where are you?
Prospective client: Massachusetts
Me: I’m 3,000 miles away! Call 911!
Prospective client: No, it’s my website! We lost 50% of our traffic overnight! Help!
Me: How much can you afford to pay?
Okay, that’s not actually how the conversation went, but it might as
well have, given the fact that I came in and performed a strategic SEO
audit, the client made all the changes I recommended, and, as a result, a
year later they’ve increased organic traffic by 334%!
Now, when you look at that data, you might assume several things.
Ryan Jones saw those numbers and tweeted just such an assumption:
Rather than leaving people assuming what was done to get such
insanely great results, I figured it would make sense to write an actual
case study. So once again, I’m going to ask you to indulge my
propensity for excruciatingly long articles so that I can provide you
with the specific insights gained, the tasking recommended, the changes
made, and the lessons learned.
First Task – Look at the Data for Patterns
One of the first things I do when performing a strategic SEO audit is
look at the raw data from a 30,000-foot view to see if I can find clues
as to the cause of a site’s problems. In this case, I went right into
Google Analytics and looked at historic traffic in the timeline view
going back to the previous year. That way I could both check to see if
there were seasonal factors on this site, and I could compare
year-over-year.
That’s when I discovered the site was not just hit by Panda. The problem was much more deeply rooted.
The MayDay Precursor
The data you see above is from 2010. THAT major drop was from
Google’s MayDay update. As you can see, daily organic visits plummeted
as a result of the MayDay update. That’s something I see in about 40% of
the Panda audits I’ve done.
From there, I then jumped to the current year and noticed that, in this case, MayDay was a precursor to Panda. Big time.
What’s interesting to note from this view is the site wasn’t caught up in the original Panda. The hit came the day of Panda 2.0.
Okay, let me rephrase that. Both of my claims about the site being
hit so far, MayDay and Panda 2.0, are ASSUMPTIONS. Because we work in a
correlation industry and since Matt Cutts refuses to poke this SEO bear,
I doubt he’ll ever give me a personal all-access pass to his team’s
inner project calendar.
Correlation-Based Assumptions
Since we don’t have the ability to know exactly what causes things in
an SEO world, we can only proceed based on what Google chooses to share
with us. This means it’s important to pay attention when they do share,
even if you think they’re being misleading. More often than not, I have
personally found that they really are forthcoming with many things.
That’s why I rely heavily on the work of the good Doctor when it
comes to keeping me informed because I can’t stay focused on Google
forums info. I’m talking about the work Dr. Pete has been doing to keep
the
SEOMoz Google Update Timeline current.
An Invaluable Correlation Resource
The SEOmoz update timeline comes into play every time I see any
significant drop in a site’s organic traffic after they call me in to
help them. I can go in there and review every major Google change that’s
been publicly announced all the way back to 2002. You really need to
make use of it if you’re doing SEO audits. As imperfect as correlation
work can be, I’ve consistently found direct matches between major site
drops and people coming to me for help with entries in that timeline.
The Next Step – Is It Actually a Google-Isolated Issue?
Once I see a big drop in the organic traffic for a site, I dig deeper
to see if it’s actually a Google hit or if it might be all search
engines or even a site-wide issue. Believe me, this is a CRITICAL step.
I’ve found some sites that “appeared” to be hit by search engine issues,
but it turned out that the site’s server hit a huge bottleneck. Only by
looking at ALL traffic vs. Organic Search vs. Google specific will you
be able to narrow down and confirm the source.
In this case, back during the Panda period, organic search accounted
for close to 80% of site-wide traffic totals. Direct and Referrer
traffic showed no dramatic drop over the same time period, nor did
traffic from other search sources.
Curiosity Insight – Google Probes Before Big Hits
Not always, but once in a while when Google’s about to make a major
algorithm change, you’ll be able to see (retroactively in hindsight)
that they turned up the crawl volume in big ways.
The SEO Audit – Process
Here’s where the rubber meets the road in our industry. You can
wildly guess all you want about what a site needs to do to improve. You
can just throw the book at a site in terms of “the usual suspects.” You
can spend a few minutes looking at the site and think, “I’ve got this.”
Or you can spend hundreds of hours scouring thousands of pages and
reviewing reams of spreadsheet reports and dashboards, all to come up
with a plan of action.
Personally, I adopt the philosophy that a happy middle ground exists,
one where you look at just enough data, examine just enough pages on a
site, and spend just enough time at the code level to find those
unnatural patterns I care so much about because it’s the unnatural
patterns that matter. If you’ve examined two category pages, you’re
almost certainly going to find that the same fundamental problems that
are common to those two pages will exist across most or all of the other
category pages. If you look at three or four product pages (randomly
selected), any problems that are common to most or all of those pages
will likely exist on all of the thousands or tens of thousands of
product pages.
So that’s what I do. I scan a sampling of pages. I scan a sampling of
code. I scan a sampling of inbound link profile data and social signal
data.
Comprehensive Sampling
Don’t get me wrong. I don’t just look at a handful of things. My
strategic audits really are comprehensive in that I look at
every aspect of a site’s SEO footprint.
It’s just that at the strategic level, when you examine every aspect of
that footprint, I’ve personally found there’s not enough of a
significant gain when performing more than a sampling review of those
factors to justify the time spent and the cost to clients. The results
that come when you implement one of my audit action plans has proven
over the last several years to be quite significant.
Findings and Recommendations
For this particular site, here’s what I found and what I ended up recommending.
The site’s Information Architecture (IA) was a mess. It had an
average of more than 400 links on every page, content was severely thin
across most of the site, and the User Experience (UX) was the definition
of confused.
Top Half of the Pre Audit Home Page
Note that the screen shot above just displays the TOP HALF of the
pre-audit homepage. Put yourself in a user’s shoes. Sure, some users
might find a homepage with this design gives them a lot of great choices
for finding information. In reality, all this did was confuse most
users. They spent a lot of time trying to digest all the options, and
the overwhelming majority of users never actually clicked on most of
those links.
That’s why click-through data is so vital at this stage of an audit.
Back when I did this, Google didn’t yet have their on-page Analytics to
show click-through rates, and I didn’t want to take the time to get
something like Crazy-Egg in place for heat maps. Heck, I didn’t NEED
heat maps. I had my ability to put myself in the common user’s shoes.
That’s all I needed to do to know this site was a mess.
And the problems on the homepage were repeated on the main category pages, as well.
Pre-Panda Recovery Category Page
Topical Confusion of Epic Proportions
Note the little box with the red outline in the above screen capture.
That’s the main content area for the category page, and all those links
all around it are merely massive topical confusion, dilution, and
noise. Yeah, and just like the homepage, you’re only looking at the TOP
HALF of that category page.
Duplicate content was ridiculous. At the time of the audit, there
were more than 16,000 pages on the site and thousands of “ghost” pages
generated by very poor Information Architecture, yet Google was indexing
only about 13,000. Article tags were automatically generating countless
“node” pages, and duplication along with all the other issues
completely confused Google’s ability to “let us figure it out.”
Site Speed Issues
One of the standard checks I do when conducting an audit is run a sampling of page URLs through
Pingdom and
URIValet to
check on the speed factor. Nowadays I also cross reference that data
with page level speed data provided right in Google Analytics, an
invaluable three-way confirmation check. In the case of this site, there
were some very serious speed problems. We’re talking about some pages
taking 30 seconds or more to process.
Whenever there are serious site speed issues, I look at Pingdom’s
Page Analysis tab. It shows the process time breakdown, times by content
type, times by domain (critical in today’s cloud-driven content
delivery world), size by content type, and size by domain. Cross
reference that to URIValet’s content element info, and you can quickly
determine the most likely cause of your biggest problems.
When Advertising Costs You More Than You Earn from It
In the case of this site, they were running ads from four different
ad networks. Banner ads, text link ads, and in-text link ads were on
every page of the site, often with multiple instances of each. And the
main problem? Yeah, there was a conflict at the server level processing
all those different ad block scripts. It was killing the server.
Topical Grouping Was a Mess
Another major problem I found while performing the audit was not only
was there a lot of thin content and a lot of “perceived” thin content
(due to site-wide and sectional common elements overcrowding the main
page content), but the content was also poorly organized. Links in the
main navigation bar were going to content that was totally not relevant
to the primary intent of the site. Within each main section, too many
links were pointing to other sections, and there was no way search
engines could truly validate that “this section is really about this
specific topical umbrella.”
Topical Depth Was Non-Existent
Even though the site had tens of thousands of pages of content,
keyword phrases had been only half-heartedly determined, assigned, and
referenced. Match that factor to the over-saturation of what was
ultimately links to only slightly related other content, a lack of depth
in the pure text content on category and sub-category pages, sectional
and site-wide repetition of headers, navs, sidebars, callout boxes, and
footer content, and the end result was that search engines couldn’t
really grasp that there really was a lot of depth to each of the site’s
primary topics and sub-topics.
Inbound Link Profile
A review of the inbound link profile showed a fairly healthy mix of
existing inbound links; however, there was a great deal of weakness in
volume specific to a diversity of links using anchor text variations,
especially when comparing the site’s profile to that of top competitors.
Putting Findings Into Action
Another critical aspect of my audit work is taking those findings and
putting them together in a prioritized action plan document. Centric
to this notion is knowing what to include and what to leave out. For a
site with this much chaos, I knew there was going to be six months of
work just to tackle the biggest priorities.
Because of this awareness, the first thing I did was focus my
recommendations on site-wide fixes that could be performed at the
template level. After all, most sites only have one, two, or a few
primary templates that drive the content display.
From there, I leave out what I consider to be “nice to have” tasking,
things a site owner or the site team COULD be doing but that would only
be a distraction at this point. Those are the things that can be
addressed AFTER all the big picture work gets done.
Tasking for This Site
After the rest of my audit review for this site, I came up with the following breakdown of tasking:
Improved UX
- Eliminated off-intent links from main nav
- Reordered main nav so most important content nav is to the left
- Reduced links on home page from 450 down to 213
- Rearranged layout on home page and main category pages for better UX
Addressed server speed and script noise issues related to ad networks
- Removed in-text link ads
- Reduced ad providers from four networks down to two
- Reduced ad blocks from seven down to five on main pages
- Reduced ad blocks on sub-cat and detail pages from nine down to seven
- Eliminated ad blocks from main content area
Refined keyword topical focus across top 100 pages
- Rewrote page titles
- Modified URLs and applied 301s to old version
- Rewrote content across all 100 pages to reflect refined focus
- Reorganized content by category type
- Moved 30% of sub-category pages into more refined categories
- Eliminated “node” page indexing
Inbound Link Efforts
- Called for establishing a much larger base of inbound links
- Called for greater diversity of link anchors (brand, keyword and generic
- Called for more links pointing to point to deeper content
________________________________________________________________
Here’s a screen capture of the top half of the NEW, IMPROVED home page.
Post Panda Home Page top half
Yeah, I know. You probably have an opinion on colors, aesthetics, and
whatnot. I do, too. However, the issue wasn’t about certain aesthetics.
That is exactly the kind of issue that can be dealt with AFTER all the
priority work happens.
Massive Changes Take Time
Given how much work had to be done, I knew, as is usually the case
with action plans I present to audit clients, that they were in for a
long overhaul process that would last several months. I knew they’d
start seeing improvements as long as they followed the plan and focused
on quality fixes, rather than slap-happy, just-get-it-done methods.
Results Are Epic
The results that have occurred over the course of the past year have
been truly stunning to say the least. I knew the site would see big
improvements in organic data. I just had no idea it would be on this
scale.
Here are three screen captures of the actual Google Data, just so you can see I’m not making all of this up.
The Panda Effect
Post Panda – Audit Work Starts Rolling Out
What A Difference Sustainable SEO Makes
Takeaways
As you might venture to guess, the recovery and subsequent, continued
growth has been about as “best case scenario” as you can get. It truly
is one of my proudest moments in the industry. The fact that not one
single other site had been reported as having yet recovered at the time I
performed the audit just means more than words can describe.
Sustainable SEO Really Does Matter
The biggest takeaway I can offer from all of this is that sustainable
SEO really does matter. Not only has the site rebounded and gone on to
triple its organic count, it has sailed through every other major Google
update since. What more can you ask for in correlation between the
concept of sustainable SEO best practices and end results?
UX – It’s What’s for Breakfast
I drill the concept of UX as an SEO factor into every brain that will
listen. I do so consistently, as do many others in the search industry.
Hopefully, this case study will allow you to consider more seriously
just how helpful understanding UX can be to the audit evaluation process
and the resulting gains you can make with a healthier UX.
IA – An SEO Hinge-Pin
With all the IA changes made, Google now indexes close to 95% of the
site content, more than 17,000 pages versus the 13,000 previously
indexed. Their automated algorithms are much better able to figure out
topical relationships, topical intent, related meanings, and more.
Ads as Shiny Objects
This was a classic case study for many reasons, one of which is the
reality that a site owner can become blinded to the advertising revenue
shiny object. Yet there comes a point where you cross over the line into
“too much.” Too many ads, too many ad networks, too many ad blocks,
too many ad scripts confusing the server…you get the idea. By making the
changes we made, overall site revenue from ads went up proportionally,
as well, so don’t let the concept that “more is always better” fool you.
Not Provided Considerations – Long Tail Magic
This observation applies to just about any site, but it’s more of a
comfort for medium and large-scale sites with thousands to millions of
pages.
If you read the first spreadsheet I included here, you’ll note that
it appears that the total number of keywords used to discover and click
through to the site spiked and then leveled off. However, if you were to
look more closely at that data, you’d also observe that the plateau is
directly related to both “Not Provided” and “Other,” with “Other”
actually accounting for more keyword data being in one bucket than “Not
Provided” (approximately 13% vs. 8%).
The more critical concept here is even though you may in fact be
missing opportunities due to those two buckets, just look at the other
aspect of the keyword count.
First, by cleaning up and refining the topical focus on the 100 most
important pages and instilling proper SEO for Content Writing into the
client mindset, we’ve taken the long tail to a whole new level, a level
that people may have difficulty comprehending when I talk about REDUCING
the number of keywords on a page.
Yet it’s that reduction and refinement, coupled with truly
high-quality content writing that is GEARED TOWARD THE SITE VISITOR that
allows search engines to actually finally do a really good job of
figuring out that all the variations for each page and each section of
the site really do match up from a user intent perspective.
If you were to unleash the “Other” and “Not Provided” phrases into
the total count, you’d see that the site now generates visits via more
than a million different phrases over the past six-month period. A
MILLION PHRASES. For all the potential opportunities in “Not Provided,”
how much time would YOU be able to spend analyzing over a million phrase
variations looking for a few golden nuggets?
Personally, my client and I are quite content with not being able to
dig into that specific metric in this case. So while it’s a bigger
challenge for much smaller sites, when you’re at this level, I’m more
concerned that Google says, “You have so much data we’re just going to
group hundreds of thousands of phrases into ‘Other’.”
In Conclusion
Well, there you have it, my very first Panda Recovery Audit a year
later. And for the record, the findings and recommendations I detailed
in this case study were presented at SMX Advanced 2011. At the time,
some of the feedback I got was “That’s nice, but it’s all speculation.
Nobody’s told us this is for sure how to fix Panda.” #WIN