For the November Monthly Webinar, we discussed the results to the tests under the November Dossier. He discussed how long it takes for a test environment to settle, the effectiveness of using Rankwyz for CTR, and if traffic can get a page indexed. He also discussed the results on our PBN test and if Google can read link attributes.
At the end, he discussed some of the tests that we have had this year that he has applied to his SEO practice and gave an example of how a page he put up using the results from the tests made it to the top 2 in just 30 mins. Must watch and must read!
Tests this month
Test 153: How Long Does it Take for a Test Environment to Settle?
Test 154: Using Rankwyz for CTR
Test 155: Can You Get a Page to Index with Traffic?
Test 156: Testing PBNs
Test 157: Does Google Read Link Title Attributes?
Using Test Results: Walking Through Some Tests and How We Used Them
November 2017 Dossier
Transcription Below:
How Long Does it Take for a Test Environment to Settle?
This test was conducted by Ryan Hough, one of our regular testers. We started this test just as a normal set up of 30 pages but we were doing them with unique domains because we are trying to set up quasi real environments in order to pit factors against each other.
One of the things we would do very well is to see if something is or not a factor, something more difficult is to see how strong a particular factor may be or if we can figure out how we could overcome certain factors in a quasi real world environment.
As this test was coming together, it refused to settle. The domains themselves refuse to settle and this test turned out into watching and cataloguing how long it takes.
Thirty unique domains and a page per domain. What Ryan did was, went through them and catalogued where they were, how much they moved up or down, and when that movement happened. Some were constantly falling out of the indexed.
Hilariously, we thought it was just going to end up with a lot of things out of the index and nothing was going to work but it finally got there. It took 41 days in order for everything to settle and finally stop moving.
It is an important point for something to think about in your own SEO practices when you put something into play. The pages are identical and there is literally no other factors going on in this test. In the situation where there are 30 different domains, it took over a month and almost a half for everything to settle down.
Think about that when you are doing your SEO and you put things into play. Things could take a long time and when you are putting them into a real environment with many more factors than nothing, it can take even longer.
I wanted to show what is happening in tests and these are insights we could gain simply in running the test. We wanted to show what is going on.
Using Rankwyz for CTR
With this test, the idea is Rankwyz has a new function called WyzTraffic within all the things that they have and do. Something that we wanted to do was isolate the wyztraffic in order to manipulate click through rate through a YouTube video. We wanted to see if we can move that Youtube video up using the traffic from this tool, but will the link through the video and clicking through a test page.
Can we also get the test page to move up as well? This is a very real world scenario where you have a YouTube video sitting on a channel and you send traffic to it and then you drive that traffic to the YouTube video through the target page and the traffic lands there. Traffic appears more clean, is the idea. It might be more difficult for Google to pick it up as a bot. In this tool, these are bots and not real people, not microworkers.
In the end, two interesting things. One thing that happened was that it simply did not work. It took over a month for the software to actually work. Initially, when Karen set it up, everything was correct, nothing happened.
This happens quite frequently with SEO tools and this test almost came up as a trials and tribulations of Rankwyz and kind of using these things in general and just how these things work. You have to monitor. You can’t just set them up and credits were dinged and they are saying that they were going out but we have Google Analytics set up and other tracking tools but there was nothing that went. A whole month of this test was just setting it up.
Then, it finally went and she fired the traffic in. The result here was that, the video that was receiving the traffic moved to the top – up to number 1. However, the page that was receiving the traffic did not get any positive movement. In fact, it got a negative movement, it moved down.
In my mind, it doesn’t tell me to not do this because you will have a negative result. I think that was just more of a dance going on. I think ultimately the page will move back to where it was. The bigger point on this is that using bot traffic through YouTube to the test page would probably not be an effective use of time and resources.
It is also an interesting point that I think YouTube is having more difficulty seeing bots than Google. My guess is that Google’s bot filter is significantly better than YouTube’s eventhough they are the same thing. The bot traffic can improve the video ranking results but the bot traffic going to the test page did not have a positive result. So if that is a technique that you were thinking about doing, I would probably not do it with bots.
I will test it out with microworkers though and a lot of the test we did using microworkers, the results were a lot better. I would consider doing that before spending time and money on bot traffic, especially bot traffic that you are going to send somewhere like the Youtube video to your target page.
Can You Get a Page to Index with Traffic?
In this situation, there is the idea of building a page and wanting it to get out in the interwebs, should you submit it to Google or should you let things happen organically or is there maybe a more organic way to allow indexing to happen and indexing through sending traffic. The idea being the page becomes interesting to Google or Google is following where this traffic is going and then it sees the page and index it for a particular keyword.
In this test, we used crowdsearch.me – these are microworkers being used for real traffic, real people going to the page. This was a Ryan Hough test, he set this up. I thought this was absolutely great. There is the idea that you want to be real but then you also want it to happen quickly. You want it to be in a way that would benefit the page. It makes perfect sense that if the page is popular, Google will index it.
Five test pages were set up and traffic were sent to each page and each page is optimized for the same keyword and the pages were checked. After a few days, there were still no indexing. In the end, just as of two weeks ago, there was nothing.
What we are seeing here is that traffic will not index a page. If you have a page that is getting visitors say from Facebook or some other referral traffic – referral traffic may get there because of the link, but the traffic itself is not going to index the page. Google might get there and index a page based on a particular link but direct traffic is not going to index a page. Google will not see it because it is becoming popular.
Testing PBNs
This was an excellent test to show how these things work. We wanted to see if we can test to see if PBNs pass juice – as an indicator to see if this is a good PBNs or not a good one. All of these PBNs came from SIA member Mads Nordskov. Mads is a linkbuilder out on Denmark. He is one of the linkbuilders that I know. He actually runs his business as a linkbuilding service but one of the things that he does is that he builds PBNs.
In addition to guest posts, outreach, those kind of regular links that you think of that are a little more whitehat, he also dabbles in PBNs as well. To my knowledge, since he started building PBNs, I do not think he even lost one. All of the PBNs that he ever built are still indexed.
Has anybody really taken the time to look at a group of PBNs to see how they all worked or if they all work and they are effective and can pass juice? The best environment to do this is in a test environment where nothing else is going on so you can see if the PBN is a positive factor, negative factor, or no factor.
If you have been in SEO for more than 4 minutes, you have heard of PBNs and you probably heard of a large mix of what people thing about PBNs – some will say that they are super effective and you definitely want to build them, some say they are not effective, some say Google has been sniffing them out since 2012 and you are going to get your self penalized. It is one of the things that goes across the board. Interestingly, most large sites actually build versions of PBNs.
They build small satellite sites they own and control and they end up linking back to themselves. Just as kind of an aside to the idea of PBNs, big sites do them. They do not call them PBNs, they call them their satellite sites or sister sites but they are essentially building PBNs. That is something to keep in the back of your mind when thinking should I build PBNs, is Google going to penalize me, my thought process is no.
If you are doing something gross where you are selling mass quantity of links to all kinds of different sites, that is a different kind of thing from what we are talking about. We are talking about controlled sites that have real content of them, you are never going to get into trouble with that. That is my opinion.
What we are doing here is we are looking at can we measure these things as passing juice. Mads very kindly donated 19 PBNs to be tested. From each of these PBNs, there was 1 link sent to a test page and then we chronicled what would happen. I initially thought that this would take 3-4 days or 5 days at the most, we will see the juice pass and we are good to go. Quite the opposite happened.
On the table – yellow indicates a change in rank either up or down, green means a positive movement up, red means a negative movement down, white indicates no movement at all. You can see that initially, once the link was fired, nearly every test page dropped which stopped my heart.
We let this test go and you can see that eventually, we are talking after two months, all but two are not positive. One is negative and the other one is in the no change. This is why I think people has such varied opinions on PBNs.
Just in a test environment, these things took forever – two months to figure out what was going on. You can fire a link, you can pay for a link, you could build your own PBN and you send that link and nothing happens for two weeks then you get a drop and it stays down and down and the decline continues and you are freaking out. You take that link away and tell your friends that PBNs are being penalized and Google found me out. I let it sit for a month and all I did was negatively impact my target page.
Conversely, you could be like the first row where I fired a link and the second day it stayed at number one the whole time. This is where in my mind, those polar opposite opinions come from, from the same basic technique. It takes a long time and you can see transitional rank. That transitional rank can be negative for a long time – in this situation, more than half were negative for quite some time before positive movement came out of those links. So, moral of the story, PBNs pass positive juice majority of the time.
Here is what I would do if you are interested in doing PBNs, you should test them. You should send them to a test page and see if that page moves up in a test environment then you can see if it passes good juice. There are other things that you could do as well. You could submit them to Search Console. I do not have any tin foil hat issues showing Google my links.
The idea is Google can see them anyway, that is the whole point of them. It would be weird if you have a whole bunch of sites that did not have Search Console and Google Analytics on them, in my opinion. That might be more of a footprint than having them on. You can submit them on Search Console when you buy and expired domain or a domain in auction, I would put it in Search Console because it would immediately tell you if there was a penalty on the site or not. You will see if there is some manual action. I will put it in there and see what is going on.
The next thing that you could do is to see if this will rank for long tail keywords. You should know what niche that particular site was in if you are buying an expired domain. Put some long tail terms. It should be able to rank for them. Third thing would be this test – fired to a test page and see if it passed juice and goes up.
Once you got those three things, then you know that you got a real site that will pass juice and Google will love it and I think you can safely use it. Mads has done this with these sites and you can see that the results of what he has done is most of his are excellent. If you are purchasing PBNs or PBN links, it might be something that you would want to get into.
This is another reason why it is difficult in my mind to get a link from somebody that you do not control. If you have no control over the particular link, you can’t do these things and you are opening yourself up to more risks. I would do these things to my PBNs.
It is also helpful even in just linkbuilding in general – it could take months before your links have any positive value.
Does Google Read Link Title Attributes?
This is one of those things that are just like everything – you do a quick search: link title attributes are an excellent hack – people don’t think about it, it will boost your pages, others will say no and it is a negative factor and pages drop, others say it is not a factor and Google does not read them. Link title attributes are very similar to image title attributes – not the image alt but the title.
If you hover over an image, you would see a little pop-up box that is the image title, you can say the same thing for links. You can also do this for links inside pages. Hilariously, in that wp article, it talks about some SEO experts believe that title attributes are not useful while some claim they are useful for SEO and provide more context – so even in the article that explains how to do that they throw in that people are split on this.
That is why we are here. That is the whole reason for SIA. There is no reason to be split. This is an easily testable thing which we could test and check out. That is what we did.
I ran this test three times. This is in WordPress, in the text editor. We created the link and we put the title which was a fake keyword and an anchor text. This fake keyword did not exist on the page. The idea is, in the first test, we want to see if Google is even reading these things and if they are in the game at all.
If Google is reading this as an element, if this is a factor that Google is considering, this page should index by putting that keyword in that spot. Test one, it did not index. We ran it again with the same set up and it was still not indexed, the same with test three.
This is pretty close to a double mythbuster where people say that it is good and that it is bad. Actually, it is looking like it is not a factor because the page should index if it is something that Google is reading. This doesn’t do it 100% because it is possible that there are some factors that just simply aren’t part of indexing.
I find that difficult to believe but it is possible so we are running this test again right now where we have it in a test environment to see if our number three page will move up, down, or not at all. But I am anticipating based on this that it is not going to move but we will test it to be double sure
The bigger thing here is, it is something that is not worth your time. At best, it would be an infinitesimal factor but since we are not seeing any movement, this leads me to believe that it is not a factor, possibly a negative factor, if Google is really reading it and caring. So, it is not something that I would spend my time doing when we think about how we only have so much time and budget, this isn’t where I would be looking to get those slight edges.
Using Test Results: Walking Through Some Tests and How We Used Them
I had a very good speaking SEO year. I spoke in several conferences including SEORockstars a couple of weeks ago. People know that I do tests, people are in SIA and brought me to speak there and they are familiar with this concept but something that I would stress is we don’t just do these tests and leave them.
These tests create a platform in order to add and subtract things on how I do SEO. So when I do these tests, I do not just leave them, I do have an SEO practice and I have 50 clients in my agency and I do all kinds of SEO. The biggest reason why I started doing these tests was so that I could actually apply them. I want to walk through some of the testing that we did this year and how I have applied them and how they changed the way that I do SEO.
The first one was using H tags in Titles. This came in September – test 142. In a Whiteboard Friday, Rand said that using a particular h tag for your page title doesn’t matter whether you use an H1 or H2, etc. That is an absurd statement. We know from our testing that that is simply not the case because the H tags are a factor and they are also weighted differently.
But, giving Rand the benefit of the doubt, I double checked this. The test page with the H1 moved to the top. An important point is what it doesn’t disprove what Rand said. We need to run the test the other way because it is possible that maybe it was a uniqueness factor that moved that H1 to the top because all the others were H2s.
We had to run it the other way – the idea is that the H2 should drop to the bottom. If Rand is correct, maybe it would stay where it is at. This aren’t important, there is something wrong with the first set up.
Again, kind of validating what Rand was talking about. But of course, it dropped to the bottom. This is not a surprise. If you have been in SIA for a week, you would probably have picked this out but it is good to see what Rand was talking about. Maybe he had some inside knowledge, maybe something has recently changed, maybe we are just off page, so running this test was good. This validates what we know.
We also know the most important things – if you want to rank for something, you need that keyword in the meta title, body content, H1, URL, there are other factors – H2, H3, H4, anchor text, image alt, bold, italic. Btw, sidenote – anchor text is really powerful right now. I would bet it is as strong as an H2 or H3 so I would make sure to get at least one link in my page and that could be an internal or external link but it contains your target phrase.
Just throwing that aside – it comes from the testing and what we have seen. We know that these are the places to put your keywords in, there are other places but these are the big ones.
Looking at this and knowing that one factor is stronger than the other in Rand’s situation, that an H1 is stronger than an H2, this leads to the question – is there an SEO factor currency? Can you exchange one factor for another? If you have an H1, can you beat that H1 with 2 H2s or 3 H2s or 4, is there some sort of factor currency? So we tested that,
Tests 131 and 132. In the first one it was H1s vs H2s and the second one, H2s vs H3s. We can’t beat an H1 with multiple H2s. After we added 2 H2s to the page, it dropped to the bottom which was not surprising but then as we added H2s, we couldn’t get it to move. Same with the H3 test, same with a couple of other tests.
These particular concepts really change the way I think about SEO. Even in the middle of this year, in June or July, I did an on-page course and I was talking about page optimization and was looking at the page more as a total entity. What I realized after these tests is that the signals are in their own verticals. You need to look at your own signals individually.
A signal on its own can be underoptimized or overoptimized and the implication there, the biggest take away is that means that a page can be both overoptmized and underoptimized at the same time and I think that this is something that holds a lot of people back or it is something that they haven’t thought about when they are doing their SEO.
They can walk into a page and see just some things are wrong based on overoptimization and you make the declaration to the client that it is overoptimized. IT very well may be. But then, what happens is that you then reduce everything on the page in all the factor areas. What you have probably done is brought some of the factors that were over into the realm of good but you took some of those that are in the realm of good and dropped them to under.
As a result, the page does not perform as well as they should. Conversely, you take it the other way – oh no, the page is underoptimized so you start adding the keyword into certain places. Those factors that were underoptimized are now in a good zone but the ones that were in the good are now overoptimized and you end up sliding back and forth with the same page. You probably get stuck on page 2 or 3 and can’t figure out what is going on.
The important point here is that you need to figure out individual factors if they are under or over optimized and then tweak them accordingly. Since I have been thinking this way for the last several months and applying it to my own clients, I can’t tell you how many times I have seen a page that is both over and under optimized at the same time.
A scarier thing is that looking at variations of your keyword, concepts that are coming from Cora where we are looking at exact match keywords but we are also looking at variations – what Ted calls “matches”. I’ve seen signals now where it is perhaps under optimized in variations but over optimized with exact match uses or vice versa. Not only can a page be under and over optimized but you can have a signal that is also under and over optimized and that can be extremely frustrating.
If you look at the different places that you optimize on a page an you evaluate them individually, you are going to create a much better page and you are not going to fall into the traps that majority of SEOs are falling into. I was falling into them until looking at these tests and realizing what they were saying – you can’t trade one factor for another, the factors are in their own vertical, and you need to optimize the factors according to what they should be for that individual thing.
Another thing that we got into was looking at Google’s ability to read. We got into how content is king, you need unique content, good content, and the people and rankings will come. If you have been in SEO for a bit, you know that that is nonsense.
The first thing that we looked at was readability and it is how difficult a content is to read and it is usually corresponds to a particular grade level. You can actually see the grade level score that Google was giving a page a couple of years ago. That was a filter that you could look at, once you got results, you could filter by their grade level or how difficult it was to read.
Google then took that away and SEOs started flipping out and saying they took it away because it is a part of the algorithm, what is a good number, what is an appropriate reading level, and people fell on seventh grade. The idea there is that an underperforming adult can still handle it and a teenager can handle it. If you have a website, you probably are not targeting children so seventh grade is the right level, you should have seventh grade content.
In this particular test using the flesch readability score, we created a page with a more and less difficult content and pitted it against each other. This isn’t just something that is done in a vacuum though, there is a particular plugin that is quite popular that gives a red dot for content that is difficult to read.
If you were coming into a situation with an unsophisticated client and they have wordpress, there is a good chance that they have a plug in like this and it is giving them a red dot and they are going to ask you if they should be spending their time fixing their content and making it more readable – is that something that Google wants? It is an important question because you need to be able to answer.
In pitting the two – seventh grade against sixteenth grade content and vice versa, there was no movement. Google is not factoring readability. Then after learning about what readability is, it is really a number that is quite arbitrary, it is looking syllables divided by words, words divided by sentences. It doesn’t take anything into consideration like what niche the content may be in, cultural considerations, grammar, or spelling.
Grammar and spelling is then something that we could look at. We can see if that is a factor and that is what we did. Grammer and spelling are not ranking factors, as well. So is there anything that is a ranking factor within reading that Google cares about and yes, test 135 – your H tag order. Your H tags need to be in semantic order – H1, H2, H3. If you start going willy nilly and throwing your lower H to the upper ones without semantic order, you are putting yourself in trouble, you are holding your self back and that is something that you don’t need to do.
I have found that the best way to handle this is to mimic a roman numeral outline. That is what Google wants to see. The other thing that you’d realize is that humans want to see this as well. Humans want to read this way – at least English speaking or western country humans, this is how most put it together and you have to keep in mind that Google was written by English speakers and geared towards English speaking audience, that makes sense to me that this is the semantic order that you should follow.
So now we got all these things, let us start putting them together. Some other things that we know from last year, word count is a ranking factor. We know that LSI are a ranking factor. We have played with Cora tons. It is not something that we specifically tested but the correlations looked really really good.
When I am doing my onpage, I am looking at four factors – exact match keyword placement in those factors that we looked at above (Metatitle, H1, body content, URL, H2, H3, H4, etc.). I am going to calculate my exact match in all those places. I am going to look at the variation of my keywords in all those places. We have not tested but we talked about strong correlations. I will look at my latent semantic index in all those places, word count for the body.
When I do my content, I create a sheet with the factors and what I have for my variations, exact match, and my word count is here, as well.
In this example, I did this for motorcycle course novo scotia. I created a page on my site and this is on my agency’s site. This page has nothing to do with motorcycles or courses. I did is an an orphan page so there is no internal linking and it is not going to get any domain help as it is an orphan page. I didn’t build any backlinks. This is completely on the strength of this page alone and building this page on the site following what I think I should do, it indexed and ranked in 30 minutes.
For the main keyword – motorcycle course novo scotia – it hit page 2 and then motorcycle training course novo scotia: page two, novo scotia motorcycle balance test: page 3 . These were related courses and these show up at the bottom of the results as related. I put them as H2 and H3 just one time on the page.
If you properly optimize on the top level page, you can win those related searches just by having them on the page. Both of those hit page 2 and 3 as well. These are not bad. We are talking 30minutes getting to page two – not to shabby. I didn’t use real words. The content that I grabbed from Lorem Ipsum. I knew how many words I would need on the page, I just grabbed that many words as lorem ipsum and they I crassly copied and pasted the exact match keywords in the factors, variations, lsi, into the lorem ipsum.
This goes back to what we are talking about that Google cannot read. This page should not rank and it shouldn’t matter that I copied and pasted keywords in particular places, lsi in particular places, variations in particular places because Google wants good content right? Google wants unique content. We will rank it if it’s really good – this is nonsense and I figured out the factor areas from our testing concepts.
The big takeaway here is that Google could not read. Google is looking at particular onpage factors clearly. Those onpage factors have different strengths. Each factor need to be measured and implemented individually. Where all these came from were from the tests that we have done, in particular, a lot of the tests that we have done this year and how they have impacted how I do SEO and how I think about SEO.
I thought this was a lot of fun doing this kind of nonsense page and then showing if you are looking specifically at the factors and what you can do with SEO now if you would take real content and a real page that you are firing links at, you’ve got internal link structure, and you’ve got a silo set up and all those other things we have talked about that you can do. You can do that much better.