Tip:
Highlight text to annotate it
X
Host: Mike is the founder and president of SpyFu, founder of VelocityScape and SpyFu.
And Mike was awarded Arizona's top entrepreneur under 35 which is really, you know a kind
of a cool thing. And this is a quote I got from Mike's LinkedIn page; success is an endless
cycle of failures that you can manage to learn from. So Mike, we do have your presentation
up in the screen and you are the presenter so, Mike please take it.
Mike Roberts: Thanks. So yeah, again my name is Mike Roberts. I'm the president and founder
of SpyFu. And this is a new presentation. Actually every presentation from me is new.
But this is extremely new contents for me and for us. It's basically the last 2 weeks
I've been spending researching what the ranking factors to go into ranking difficulty. That
is how hard is it going to be for me to rank on this, on this keyword. And the results
of that research are just today as of like noon, live on the site so, I thought it would
be cool to do something extremely timely. It's not exactly just the demo of SpyFu which
is what we sort of talked about. But it's basically, not as cutting edge of research
as you can possibly have, very, very, very new. Real quick, let me tell you about SpyFu.
We were started here in actually Scottsdale or Cave Creek, I was actually up in Tatum
Ranch, since this is a local crowd, you guys may know where that is. And we were founded
in 2005. The original product was actually called GoogSpy. And we changed the time to
SpyFu when we launched the new one. We're actually the very first competitive intelligence
tool.
Prior to SpyFu, you couldn't... If you wanted to know which keyword your competitors are
buying or, on Google AdWords or ranking on, on Google, you would basically do guess and
check. And so when we put that first product together, it's the first of its kind. Soon,
soon you know, we've been in the space, like I said for 6 or 7 years and we continued to
evolve you know, well it's not so much about getting those keywords anymore, is it? It's
about figuring out which ones represent the best opportunity and you know, which one's
are actually going to be profitable on. We want to tell you what you're able to rank
on, which ones you're going to be profitable on before you even start so that you're not
wasting your time and spinning your wheels. One thing to note is that we do no consulting,
we make all of our money on basically people paying $79 a month to use our service.
So from an SEO perspective, that's the audience, we also of course do AdWords and we specialize
in that as well. But from an SEO perspective there's things that we're really good at.
What we really nailed are how to win SEO budget. We have a platform of reporting called spy
for recon reports that basically, we're designed from the ground up to make you, the SEO look
like a million bucks. The idea is to explain the value and highlight the opportunity so,
compellingly the clients won't just love the work that you do but they'll want more of
it, right? And then at the end of our presentation we sort of just say, well, here's what your
competitors are doing, they're acting, now get to work, right? So that's how the whole
spy for recon document is structured. And when we tell them the value and we highlight
the opportunity, we don't talk to them in terms of just ranks and clicks. We actually
translate those things into dollars and we do a really great job of it. So these are
the things that we're currently good at. We can ask you, you know, what's your goal and
that should have something to do with money. I'm going to breeze through these things because
what I really want to get to is the meat which is the subject of my research.
We hope you find your keyword universe and some of the ways that you can do those are
just knowing what they are. You can also steal them from your PPC campaign. This is a great
win because if you look at what your PPC campaign is, the keywords that you rank on all ready,
I'm sorry, the keywords that you're buying but you also rank on, you can easily justify
that to your clients or to your you know, your internal... your boss or whatever. So
we want to look at which keywords you already have traction on. And then we combine those
keywords into keyword groups, silos if you will. So that you're not constantly thinking
about you know, the individual keywords telling your client oh yeah, we lost rank on this
keyword. We want to talk to them about well, this is a portfolio of keywords. So we do
these stuff, right? We calculate and you know, like metrics. Not just on the keyword level
but also on the keyword group level.
Of course we give you everything but what you really want to be talking here, you want
to control that conversation with your customer, talking to them about the value you created
out of macro level. So when we calculate the size of opportunity, the size of keyword group,
we take into account the keyword search volume, the value per click like in terms of you know,
relative to PPC dollars, AdWord dollars, where you're currently ranked and ultimately what
your goal is. When we apply the click through curve, you should be probably familiar with
it, but ultimately the idea is that you get more clicks when you're in the number one
position than if you're in the number 2 or number 3 and so on. We also take into account
you know, number of ads, shopping, video, pictures. Our click through curve is pretty
solid. All these things, we're really good at. When you combine all these things together,
you can actually talk to people in terms of dollars. So I mean, managers or entrepreneurs
or anybody that you're dealing with doesn't really think in terms of ranks or if they
do, if you can get them to think in terms of ranks, they're going to discount your value
because they don't really... they're certainly not going to give you a premium, they're going
to discount your value. But if you can talk to them in terms of dollars, it's a huge win.
So then we add up all that opportunity and we group them into these reports that look
like this. So that's what we're good at. We're currently very capable of figuring out what
opportunity looks like and what value of SEO looks like and we're basically the best that
there is. But the place that we suck is when it comes to trying to calculate difficulty.
This is actually, these are slides ripped from another presentation that I've done.
And you notice that there's nowhere in here where I'm talking about a single metric from
SpyFu. When I'm telling you okay, your next step after you calculate opportunity is look
at the SERP, you know look at the backlinks, look at the SEOMoz, look at the difficulty,
blah, blah, blah, blah, right? None of it has anything to do with SpyFu. But imagine
what we could do if we knew difficulty. If we not only know the size of the opportunity
but also the difficulty, we could come up with, we could tell you look, this is not
just like a big opportunity but it's easy. You will get these ranks. You will get this
traffic. It's a low hanging fruit report that we can produce, right?
There's all kinds of ways that I can you know, revolutionize this stuff if I can do that.
But right now, I can't, okay? So the last couple weeks, my mission has been to figure
out how to quantify this. Okay, so where do we start? Well my theory is the best place
to look and to figure out what affects rankings in the first place, actually that should be
A-F-F-E-C-T-S but ... by the way, Roy, try me if you have any questions. I've got like
a little studio audience here to help me out with sort of queues as to whether or not I'm
explaining things right. But if you have questions, because some of these stuff is going to become
a little bit technical as I go through you know, experiments, correlation ...
Host: Well I'm glad you asked, Mike because what I'd like to ask from our participants
in today's webinar, if you do have any questions, please type them to the question box. Mike
will answer the questions. We have a little area during the end of this webinar for any
questions you might have. And again, the person that asks Mike the best question, Mike's going
to basically tell me who that person is and we're awarding them one free hour of SEO consulting
with our company. So you know, Mike I don't have any particular questions right now but
you know at the end of your presentation, we'll certainly address those.
Mike Roberts: Yeah. I'm just about to get into the crazy stuff.
Host: All right. Here's the crazy stuff guys.
Mike Roberts: All right. So hopefully I've explained it well, I've really tried. I'll
tell you what, when I first wrote an article sort of a draft article for my team. I let
them look at it and they're like 'this is just offensive, I am overwhelmed by this data'
and so this is sort of my first step of trying to make this palatable. I'm pretty sure it
works but you know, I mean if not... All right, so let's see, okay yes. So basically I want
to figure out what affects ranking in the first place particularly the things that I
currently measure, right? I wanted to start out with the stuff that's visible on the SERP
page because people talked about that it's just... the first thing you do as an SEO is
you actually scan the page and there's a lot of what almost seems like, it's expert knowledge
but I feel it feels like instincts. You've looked at it and you're like oh, instinctually
I know that this is a more difficult keyword to rank on but what are those cues and in
my experience when you have this sort of expert knowledge that feels like instincts, if you
can put any amount of that into code and then do it in like a billion times a second you've
got like the makings of a great algorithm.
So anyway, let me talk about my methodology real quick. In order to compare, in order
to figure out what ranking factors, what factors, you know, what metrics affect rankings, what
I do is I take each individual metric like keyword and title and I reorder the top 50
searches as though they're the only factor that Google uses to search the search results.
So if the keyword is in title, twice then you know, okay then that's the number one
result and down at the bottom there's no keyword and title. So that's what we pretend. What
we do is we compare that to the way the Google actually does rank it. So it kind of looks
like this, right? So this is keyword hits and title. You see this column right here.
I actually don't have the actual text here because they'd be really wide. But ultimately
this is the metric for keyword hits in the title. For, basically what I'm saying when
I say keyword, I mean it has one gallon beverage dispenser. 1, 2, 3, 4. That means that there's
not an exact, in this case, it's the number of times that there's partial hit.
So then we sort them, right? And then we compare these differences. And the way that we compare
the differences is using what's called a Spearman's Correlation Coefficient. And so technically
if you know anything, if you know about Spearman's, this sort of order isn't exactly right. But
it gets the point across. You actually have to account per ties, and we do. So when I
apply this methodology, I can see a correlation between certain metrics. And these are not
super strong correlations, I should explain real quickly what a correlation coefficient
looks like. It ranges between negative one and one. And one is basically, that is the
exact same data. One means that is a perfect, perfect match. It's not necessarily the exact
same data but we go the Google search results in the exact order that Google put them, we
would have a correlation coefficient of one, okay? If we have our data completely backwards,
then we would have a correlation coefficient of negative one. If we have a correlation
coefficient of zero, that means it's pure white noise, right? So these correlation coefficients
are relatively small but they also have a small margin of error.
So what we're looking at here is when I say point one, that means it's not like a very
strong signal. Google doesn't rank their search results exactly based on result is homepage
or number of keywords hit and title which we know intuitively. But what I can say is
with a plus or minus of this top one could maybe range between .09 and .11. There's actually
a smaller margin of errors on that, because I'm looking at 65,000 rows of data. So it's
pretty huge. So anyway, what we found is that of the things that you could see on the SERP
result is homepage is the biggest. Keyword hits and title is the number 2. But what I
found interesting here was that keyword hits in the URL was very small. I've always heard
that keyword hits in URL is a big deal, I'm sorry, putting the keyword in the URL is a
big deal and what I want to point out is that keyword hit is... this particular metric,
keyword hits in URL actually also contains exact match domains, because I'm looking at
the full URL. So if it's an exact match domain, that's also in there. So this thing should
be dragging in there up. So I wanted to drill down deeper into that and see like is that
true? I mean I'm double checking my work. I've always heard that that's actually kind
of an important thing.
So I drill in, right? And here's the domain only. It's the exact match domain only. And
this is the URL without the query string so really not a big difference between the full
URL. So this was, see this is .05 and this is .05. Those are the same ones, okay? And
number of keyword hits in URL. This would be number of keyword hits in URL. But then
when you drill down deeper it's like putting it in the path is as close to white noise
or putting them into subdomain or literally not even the first level of the path but like
anywhere in the path or page name or the querystring or anything after the domain, is nearly white
noise. Because remember, this might be a plus or minus .01. So it's very, very close to
zero correlation and... But there's nothing to say that it has a negative impact. It's
just that refactoring your URLs so that they have the keyword in them probably isn't the
first order of work. You could spend a lot more time getting the right titles, I mean
or whatever. But there is some evidence that putting the keyword in the query string could
potentially, I mean it's a negative correlation with it, right? It's very, very small negative
correlation. But it's a negative correlation nonetheless. So what that means when I say
negative correlation is it may actually negatively impact your ranking. Probably not going to
negatively impact your rankings, but it could, it's associated with low ranking keywords
or low rank results.
Okay, so I wanted to find out, and this is like you know, sort of what I do, is I want
to figure out whether or not I can combine any of these factors, these on-page visible
things together to make a super-metric, right? And in my experience, I've done a lot of algorithm
and stuff like that, usually you actually don't do much good by combining things together.
Usually it's, and if it is good it's like 10% better. So it doesn't usually pay off
really well. But you know, I search for those needles in the haystack. That's what I do
to try to make everything that we do better and so I was actually able to figure out a
way to combine. And I forgot to put this on the slide exactly what I combined. It's like
keyword and title and results at homepage, and... I'll try to publish this later, I need
to put that on the slide. But I ultimately was able to, you know what, I'll do this after
the presentation. I know where it is. I have the formula some place. If anybody wants the
formula, I'll happily disclose it. But anyways, I was actually able to combine several of
these on page signals together to create one that actually correlates better. 30% better
than the best one. So that's cool.
So what we're doing here is not exactly what Google does. Google may have millions of pages
that it could push to the top result. And what we're doing is reordering the top 50,
right? So we're basically trying to reconstruct exactly the identical tip of the iceberg without
taking any of the rest of the iceberg into account, right? And the truth is, I mean we
could go a thousand results deep but like you can't get you know, there are certain
levels past which you can't get... you can't get all of Google's results. So you know,
we're doing this top 50 which is more results than anybody has taken into account and doing
this sort of study but, I think that the growth actually matters, right? Metrics that get
better as you go from the top 10 to the top 20 to the top 30 to the top 40. The deeper
you go into the SERPs, the more fundamental that is of a ranking factor, right? Because
before you get to the top position, you have to get to the top 1000. And you know, before
you get, and then you have to get to the top 100 and then you have to get to the top 10,
right? But if you really want to understand the way that Google ranks, you need to try
to predict you know, deeper down the results than you know, the top 10 or the top 30 or
the top 50 or whatever.
So I look at growth, okay? And you can look at this. And this is basically, what this
is the correlation coefficient plotted based on, okay so this is the top 10, top 20, top
30, top 40. So ones that are, these guys that are going up into the right are actually getting
better at predicting Google's rank, or Google's results as you consider more and more of Google's
documents. And the ones that are going down to the right are getting worse and worse.
Makes sense? Yeah? Okay. So what this means, and this sort of should correspond how you
feel, like this is like if you have an SEO instinct, I'm saying the keyword in title
is more important than exact match domain. Because you see exact match domain here in
purple becoming less and less important as you go down in ranks, I'm sorry, as you go
deeper into the results. And then you have for example, number of keyword hits in title
going up pretty rapidly. Not actually very important at all in the top 10 or the top
20. But as you go deeper and deeper, it becomes more and more important. So I think, my argument
is that's a more fundamental ranking factor.
So here's another way of looking at these things. And you can sort of visually see the
heat map how these each of these on page are in-SERP metrics sort of grow and how they
are relative to one another. So we are talking about exact keyword in title, or I'm sorry,
we're talking about number of keyword hits in title. And see how it goes from orange
to green. And sort of if we are to rank this list here, it's ordered by top 50 rather than
by top 10. But if they were ordered by top 10, number of keyword hits in title would
be very near the bottom, it would be third, I believe from the bottom whereas it sort
of it makes a recovery but at top 50, it actually is third from the top.
All right, so this is, I don't know, maybe a little bit of in depth stuff here. Okay,
but what you can do is actually plot a linear, you can linearly regress or you could do a
regression of any of these. It doesn't have to be linear. And essentially figure out well
what's the flow. You know, how fast is this thing getting better or how fast is it getting
worse. And if you figured that out, it's essentially the growth rate. It allows you then to predict
into the future, predict deeper into the results. So if we wanted to predict well, what's it
going to be in the top 100 or the top 200 or the top, well, I think these things probably
aren't linear forever. So we don't want to go to the top thousand probably without like
maybe a better model or something. But I think it's fair to like project into the top 100
or top 200 or something.
So when we do that, so what I ended up doing is taking this 47X and this 39X, those are
essentially, that's the growth rate. That's like, if you remember like, I don't know,
7th grade Algebra, that's the slope. The slope in this case is .0047X. And so I multiply
that by like 100,000, to make it... like make you annoyed every time you saw it, so I may
show you, I'm not sure whether or not I'll show you, I'll show you these number again
in like an article that I publish. But these are the growth rates. And so you can see faster
growing versus slower growing. It's a simple number. Anyway, here's this growth rate, here's
what happens when we use that growth rate to project the top 100. Okay.
So you can see keyword hits in title has a very fast growth rate and we predict that
it's going to become more significant. So it's sorted out here, down here at .08 which
would have put it you know, somewhere in here in the list. And because of its growth rate.
By the time you get into the top 100 it's a much more significant metric. Conversely,
you have results in homepage and does that makes sense, what results in homepage is?
That means that the search results is actually the homepage. On a gut level when you're just
looking for SERP and you see a bunch of homepages there, you're like oh this is kind of a difficult
SERP. And so you know, that's like sort of common sense, SEO common sense or SEO instinct
that sort of proved to be true here. It's actually a pretty strong factor. But it becomes
less important as you go down... as you go deeper and deeper into the SERP.
Similarly, you got exact match domain. Exact match domain is actually the slowest, or it's
the fastest growing... it's the fastest, what's the opposite of growth? Losing? You know,
negative growth. It goes down into the right the fastest. So I wanted to contrast those
and give you another way of sort of visualizing this growth. So we've actually taken some
of these metrics, these new things and integrated them onto SpyFu. And this is brand new today
as a coincidence in a way or why don't I just say that I did it for you guys? But we actually
happen to launch this thing at about 11:30 today. And so you can go to spyfu.com. This
is all free stuff that I'm showing you right now. You can go to spyfu.com and type in any
keyword and you'd be able to see this sort of analysis. So you can see on a roll up level
we're like okay, there's two homepages in results.
Another thing that I sort of didn't put in this research is what happens with WDU domain.
But that's sort of a historically interesting thing to look at, right? Especially if we're
like looking for medical terms. Google is probably going to almost give a brand effect
to those government and educational domains. So we want to look for that. It gives us a
sense of the difficulty. In this case, I think I'm looking at like Lance Armstrong, lots
of keywords in title meaning lots of results. This is the number of results with the keyword
in title. Lots of results in keyword in URL. So it's a relatively well optimized page.
I see well, when you're actually scanning the SERP, we bolded the things that you need
to be looking for. You got the homepage, you got an exact match domain here, you got the
keyword in the title. So you can sort of just look and see how much bold there is too, it
gives you a general view or it improves your ability to scan. You got the keyword in the
URL here. Actually I didn't actually call that one out but you see them here. In cases
where you got 2 SERP, 2 results in the same SERP or more than two, you can see that here.
And we also added to this thing the position based click through curve. I think that's,
it's funny that we've never done that before because everyone's always like take all these
keywords and put them into a spreadsheet and do this really annoying calculation that's
not even, you know that's based on 2006 AOL data, which is the most annoying thing you
ever have to do with keywords. And it doesn't take into account universal results or ads
or anything like that. So we decided we'd put that underneath keyword page. One other
thing that's sort of, of note here is this domain diversity. And that basically shows
you how many different domains are on this individual SERP.
Okay, so the next thing that I want to look at after looking at the stuff that's visible
on SERP is domain level metrics, right? So you know, you got like your backlinks, you
got you know, you got page rank, so on that type of stuff. So when we did these, I looked
at... I pulled in a bunch of majestic SEO stuff, I was also going to pull in SEOMoz,
the same metrics from SEOMoz and the same metrics from like AHF. But I did, just because
I wanted to see how hard I wanted to work, I actually pulled in like a sample of those
and compared them against each other. And those like metric by metric even for like
trust flow versus Moz Trust there's like a .91 correlation coefficient or very, very
high correlation coefficient between pretty much all of those metrics. So I was like well
then I will just pull in one and see how those metrics perform and if I need to do the other
ones later then I guess I will. But save me a little bit of time. So the best metric is
trust flow from majestic followed by this interesting, just straight up domain age.
That's a metric that we pulled from the SpyFu database.
At SpyFu we have like 79 months of history. So you can actually look and see... well actually
you currently can't but I can, you can see every keyword that any domain ranked on 6
years ago. And anyway, this domain age thing is actually just the first time that we saw
a domain show up in the ranks, show up in any search result even if there was an advertisement.
And so that domain age was almost as good as trust flow in predicting Google's rank.
I thought that was really... I was surprised, almost annoyed because it was so simple. But
yeah, I guess, I don't know, you just don't really want to believe in whole domain age
thing. And it's not exactly the same as how long ago it was registered, right? It's like
actually how long it's been kind of trusted by Google if you like, if you want to think
of it that way.
Okay, so we also looked at backlinks. And then you see that domain page rank is the
lowest performing. If you pay close attention to this stuff, that's not a surprise. But
I mean everybody uses page rank, well not every I mean, you find yourself doing it even
though you don't want to. Anyway, there's a lot of metrics even domain age that beats
page rank, okay? So you could happily probably replace your domain page rank dependency with
damn your anything.
Okay, I wanted to throw something into the mix. I, before I even sort of thought of doing
this experiment, I came up with a metric my own and I wanted to see if I could predict
ranking results based on actual performance in the SERP without taking into account backlink
data. My idea is that this would be a good metric to combine with other metrics. So I
was like it's going to be great, I look at a page... I'm going to look at these you know,
backlink data points and actually eventually combine a SERP performance metric at the domain
level with those things to make a better super-metric. This domain strength is not like a straightforward
equation that I can just layout for you. It's actually algorithm based which means that
there's a whole bunch of branching, you know statements and stuff like that to figure out
what an actual domain strength becomes. But I use a lot of the tricks that I used to calculate
SEO value, the opportunity side of things that I was talking about earlier, that click
through curve that we do that takes into account universal search and ads and stuff like that,
ultimately we're looking at how many keywords a domain ranks on, what positions the keywords
are in, how many times it ranks with the same keyword, how many searches that keyword gets,
how much the keyword costs, the competitiveness of the keyword. A lot of different factors
go into this but I'm kind of, that's kind of what we do and so I thought I'd compare
it.
And I will say that I was extraordinarily surprised that it beats all these other metrics
because I have a lot of, I basically have not predicted that in any case but it beats
them. And of course I'll release data and we should probably do a follow up study to
make sure that I know what I'm talking about. I'm pretty sure I know what I'm talking about
but we'll have to pull in a whole new set of data and re-compare and so on. But at this,
especially domain strength, does not suck. It actually in my opinion a very, very solid
metric. Here's how it looks as it grows, okay.
So all of these things actually have a pretty solid growth rate. None of them, you see have
a negative growth rate like we saw with the other, with the on-page metrics. Everything
here goes up into the right. Here's how it looks as a heat map. So you see domain strength
on the top and actually what's interesting is that page rank has an interesting growth
rate. It starts out actually as a negative correlation and then results as falls still
the lowest rank but it makes a little bit of a comeback. The next best metric is the
trust flow. I've seen trust flow and citation flow be very similar. What I think is pretty
interesting is that there's not that much... you can look at a raw number like class C
blocks and know what that's derived from. And be really, really, basically within the
margin of error of accurate. And that's backed up by basically SEOMoz. SEOMoz will compare
their number to the class C block numbers that they have and it's basically the same.
The other value of looking only at class C blocks is you're going to look at backlink
data is that it is not scaled to, it's power curved just like search volume is power curved.
So you can actually divide. If you want to stick with using backlink data to estimate
keyword difficulty or whatever you should look at like class C blocks because you could
divide search volume over class C blocks to get basically a really solid metric versus
dividing something that's, dividing into something that's like called to zero to a hundred. You
talked to me about dividing into a power curve into a linear curve and that's just, it's
just going to make the big keywords always win. And so that sucks. So yeah, anyway, moving
along. Getting like weird looks from my inside audience. They're like what the hell. I will
not talk about that anymore. All right, so here's the growth rate. SpyFu domain strength
actually has the fastest growth rate also. And you'll note that, that like 186 is scaled
to the same as those, as these guys. So you have like 47 and 109, and 186 is fast. 186,
it says 155 is fast. Even 135 is a strong growth rate. So let's talk about how we're
using these in SpyFu. Back at that same keyword result page when we're analyzing the organic
rankings, we have the domain strength right there. So you can take all of these things
into account. All at once in one spot.
All right, moving on to page rank stuff, okay? Because this is essentially the holy grail
of estimating difficulty or predicting rank, is how many backlinks does this page have.
Not like the entire domain but this individual page. And in fact you see that these are like
the best correlation coefficients that we've seen. Class C blocks linking to URL is the
best metric there is. Oh you don't know this, I could have actually maybe skipped that.
But anyway, I pushed forward. Here's the issue, is that I was only able to get page level
metrics for about 40% of the search result. So 60% of them basically returned N/As. So
it was actually like... I was like what do I do here? Everything else I've compared against
has been you know, the full data sets, 65,000 rows or whatever. And now I can only get a
very small amount of data like I mean, that's a significant loss. 60% of the URLs didn't
contain this page... I couldn't get any of the page metrics back. So that's sort of the
weakness of it, when it's there it's super strong, very, very good. Page level metrics
are great. Class C blocks, linking of URL is outstanding.
Even URL backlinks, anything's really good there. But if it's not there, and you want
to try and apply it universally which I do, right? Because I ultimately want to come up
with a keyword difficulty, I mean I ultimately end up having to just take these things into
account and I don't know, come up with what is an old value. But it makes it a little
bit more difficult especially it makes it a lot more difficult when you're kind of trying
to do this without the benefit of an algorithm and you're just sort of looking at a SERP
and have some numbers. But anyway, here's how this worked out, right? The growth rate
was also very different. The top tier we have you know, this is the best case, right where
we filtered out all of the rows where we didn't have page level metrics and down below is
the full data set and you see that you're going from yellow to green or red to orange.
Basically you're getting more... you're getting colder or whatever it is. You're getting better
as you go from top 10 to top 50 whereas the opposite is true when you're doing this with
the full data set... And it actually makes sense, right? The further you go down the
search results, the less likely there to have to page level metrics because those are like,
those pages are just not as popular probably. I mean statistically speaking, you will end
up having less data points on those. So here's how the whole thing looks, okay? And what
I did here, this is actually the best case scenario for the page metrics. So what I did
was I tried to combine several of these data points to come up with like really the foundation
for my keyword difficulty stuff. I wanted to say, okay, what can I actually do to really
you know, do the best job I can. And you see there's a big jump here between the best individual
metric and then the best... and then just a set of super-metrics.
So I was actually able to come up with 4 super-metrics that all outperformed everything else. Even
basically the page level metrics on their best day, right? I can actually apply on page
signal plus domain strength, right? Or basically it's like on page signal plus domain strength
taking no page level metrics into account or whatsoever and actually beat the page level
metric. So the very best, the holy grail, the perfect metric, right? The page level
backlink data is not as good as on page signals plus domain strength which is actually a fairly
huge surprise for me. Of course we can make it incrementally better by taking on main
page signals, domain strength and then these page level citation flows that I actually
use which is right... someplace. Where's citation flow? Oh there it is. So I may be should try
that with class C blocks. That would be kind of cool and it occurred to me ... at the time.
But I should probably do that, see if I can actually really do well. But yeah, anyway
so, that's kind of the big surprise for me is that I was able to beat those backlink
metrics.
Here's what this thing looks like as a big old heat map. So you see that there's pretty
good growth in all of these top metrics. They're not like slow growths. Here's what that growth
looks like if we were to project 200 results deep. Because this growth here for these guys
is actually stronger than the growth here for this guys, the separation becomes nearly
50% once you project further down to the SERPs. So how are we using this? Basically, so what
I did is I took these metrics and effectively used them to calculate the keyword difficulty.
This is the end of my mission, right? There's actually a few steps in between but I'll save
you the horribly boring math part there if you're not already horribly bored, so you
look at the same search page I was talking about and we got ranking difficulty right
here front and center. We also have it integrated, probably seen it before, into the organic
search SERP analysis. One thing that I didn't mention before but this is really interesting
for the Martha Stewart one, is that we have these social domains and so surprise, surprise,
Martha Stewart is on Pinterest, I would be interested to see how this affects, I've actually
looked to see how this affects the SERPs and I can't find an exact pattern yet. I've actually
looked to see what social in SERP actually does. But it's not like having Pinterest.
Pinterest doesn't necessarily rank high on the SERPs and Twitter doesn't necessarily
rank high on the SERPs. Actually I believe there is, I did look and see, oh yeah, it's
results from Wikipedia. So this isn't that strong of a signal and it trends downward.
I use that as a baseline. So I want to figure out a way to algorithmically incorporate social
and domain, or social and SERP but I can't quite. But you could, you know you got that
gut level instinct as an SEO looking at this and oh, Marthastewart.com, Marthastewart.com,
Marthastewart.com, Wikipedia, Martha blog, Martha Stewart Weddings. And then Pinterest
and Twitter. It's like okay, it's going to be hard for me to get above those, right?
And then you look here and there's 30 Marthastewart.com in SERPs. See your domain diversity is like
39.6%. So that 245 is really high keyword difficulty. And we captured it well but I
feel like you know, looking at these social signals and domain diversity could potentially
improve. You could use that to hone your skills. But you know, we provided, it's just that
I haven't captured them in algorithm yet. But it's sort of what I'll be working on.
Anyway, so I've got a little bit of time for a quick demo of the other place we're integrating
this. And that's in our keyword research which is keyword smart search. So I just put in
like an interesting niche here, the ultrasound technician and what you can do here is actually
filter by SEO difficulty, right? So I want something that's less difficult than a 100.
And that's going to filter for me. And really I want to make sure that there's a difficultyn
so I want it to be greater than 25. And then I also want this to let's see, I definitely
don't want it to be zero. Let's go for 5 or more searches a day and see what we get. Okay,
so you can see what we're doing, I've actually try and figure out like cost per click. Let's
see, maybe I could do $2 and more. I haven't actually done $2 and more. It'd be interesting
but you can sit here and do this with your smart search all day. Oh look at that, there's
number one results that I was looking at. We're still pretty good. Diagnostic ultrasound
program, easy difficulty 56.9 and a high cost per click. Not a bad search volume. I mean
for the cost per click. So you can now use this. We just now made it so that you can
start refining your keyword research based on these stuff. And we'll be integrating keyword
difficulty pretty much all over the site in various different ways you know, going forward.
But that is all I got. Did I go over? Did I stay inbound? I don't know. Do we have any
questions?