For our July webinar, we talk further into detail on the tests for this month's dossier and how you can apply it to your site or your client's sites to help you move rank.
Tests this month
Test 131: Can You Beat an H1 with H2s?
Test 132: Can You Beat an H2 with H3s?
Test 133: Can You Beat an H1 by Increasing the Keyword Density in the Body Content?
Test 134: Is Dwell Time a Ranking Factor?
Test 135: Does the H1 Position Affect Ranking (Above or Below H2)?
Test 136: Testing the Strength of Internal Links in a 30-page test
July 2017 Webinar
Transcription Below:
Can you beat an H1 with H2s?
What I am focusing in on tests 131, 132, and 133 is what might be called an economy of ranking signals. We know from testing that there are four signals that are continuously strong. You have your H1, meta-title, URL, and the body content and those always win when we are running head to head tests. Then there are all the other signals – your h2s, h3s, h4s, image alt, bold, italic, anchor text, those sort of things. What we are looking at though is, does Google see this as kind of an economy?
In a sense, can you trade one for the other or if you have one H1 and in this case, H2s, can you outrank an H1 with a certain number of H2s? For example, can two H2s beat an H1 or can three H2s? What we are looking at then in these tests is to set up a way to identify what that economy might be or even if this is how it works, in the first place.
This is a common problem that comes up especially when you jump into a website with a client and they have already built it and you are looking at it and you see that there is no H1 on the page and they decide they didn’t want it for design purposes – could you make it up with H2s, for example?
In this situation, what we did is set up five test pages in the normal way and then we added h2s to the test pages and then to the ones ranking 1, 2, 4, and 5 were H1s. Immediately, as soon as the H2s were added in vs the H1s, predictably, the H2 page dropped to the bottom.
At this point, the idea is that we will continue to add H2s until it rises to the top. So that we did not mess up keyword density or word count if those were possibilities here, instead of adding another H1, we wanted to make sure that the word count remains the same, just a line of text was added so in a sense the keyword density and word count remains the same.
As we continued to add H2s, the test page would never rise and ended up sticking to the number 5 spot all the way to the very end. We got to the point where we got to 6 H2s on the page and it simply would not move up. What I think we saw here was we got to the point of over-optimization, in a sense, in actuality.
We had so many h2s on the page as a strong ranking signal that we ended up over-optimizing our page before we could overtake the H1. We could not beat the H1 with H2s. This test was run three times and the results were consistent every time.
What this tells me in this situation is when you are in a situation where you are facing an H1, I do not think you can trade any number of H2s for an H1. H1, the metatitle, the body content, the URL are kind of in a category all on their own and you probably could not overtake that in terms of a straight one to one with a lesser signal.
Can you beat an H2 with H3s?
Then we looked at 2 lesser signals – can we overtake an H2 with an H3? First of all, will they be seen as different categories? Will an H2 always beat an H3? The idea is that once we launch the H3 on a page vs an H2, that H3 will drop predictably and it would rise up again once we have enough H3s on a page.
This test came out exactly the same with the H1 vs H2 test. What ended up happening is that we could not overtake the H2s with H3s. Same exact scenario with when the page dropped to the bottom and as we continued to add H3s, it simply would not move and it stayed in that number 5 spot.
What we are seeing here is I think, when you are looking at a page’s structure, there are separate areas. I do not think that the different areas that Google is looking at are treated as kind of one giant conglomerate, total page optimization and that they are viewed as a separate, individual signals.
As such, you do not want to try to horse trade a number of H2s or H3s for something else, especially something that is higher up on the ladder in terms of a signal. When you are looking at your competitors and they are averaging 2-3 H2s and say 2-3 H3s, that is probably the range that you’d want to get into in order to provide what Google is looking for.
I don’t think that these are an opportunity, necessarily, especially if you are in a competitive niche and things are heavily optimized. I do not think these H tags are an opportunity to gain an edge. I think what you would probably want to do is optimize to the point of what Google is anticipating or what Google is rewarding.
Can you beat an H1 by increasing the keyword density in the body content?
The next one, kind of in line with this economy is what about keyword density vs an H1? Can we beat an H1 by increasing keyword density? We know that the point of overoptimization within keyword density within body content is probably 2 times the standard deviation.
The standard deviation is kind of the difference between all of your samples – where we want to aim for is about 1x standard deviation and if that didn’t make any sense to you if you are new to any of this, this is in some previous testing and I would be more than happy to explain it if somebody asks some questions on it. I have gone through this before, happy to go over it again and talk about it later.
The idea is that we know that is the sweet spot for keyword density but if we were to increase the keyword density in the body, can we overtake an H1. The important thing here is, we know that these are the top signals, these signals are kind of in their own category above say and H2 or H3, are they also kind of in the same line that one is better than the other or stronger than the other and you could not overtake them or are all of these approximately the same signal between those 4 things.
In this scenario, what ended up happening is that by adding two keywords to the keyword density – the keyword density at this point was 2.6%, all the other densities were at 2%, we are right about 1.5 of the standard deviation, roughly. Low and behold, the test page moved up to the number one spot.
Kind of putting all of these together, what I think we are looking at is when you are looking at H tags, specifically H2s and H3s which are supporting metrics, that is not an opportunity to gain an edge and I would not do that. However, how you can try to gain an edge is within body content. That is where I think you can try to determine what the average is and manipulate your text to get slightly above that average and gain an edge on your competitor.
Do not try to outdistance your competitors with H2s and H3s, that is not the way to do it. What you’d want to do with those things is get them to were Google wants them to be and that is basically the average that Google is rewarding, looking at your particular keyword. Where you can get that edge and maybe move up against other stronger signals is within keyword density, using the exact keywords in a specific number of times.
One thing to note is that 2% is a high percentage and you probably won’t have that in real life as an exact match for your keyword so don’t use that percentage as the golden rule. Look at what your competitors are at for exact matches for body content and then determine things from there.
Especially if you have longer pages that you are going up against, your keyword density will be remarkably low on the exact match. This doesn’t take into account LSI, of course, it doesn’t take into account what might be considered matches, if you are kind of into the CORA software or those concepts, this is just using your keyword exactly against your competitors.
Is Dwell Time a ranking factor?
Dwell time is an interesting concept. It is basically time on site but Google Analytics and other tools that measure user engagement cannot give you time on site unless you get another click. For example, if somebody comes to your page and they stay for 25 minutes, they do not click on anything and then they leave, that is a bounce and then their time on site is not recorded.
In order to get actual time on site, they need one more click. Google needs that extra click to another page in order to gauge the time on site. Dwell time would be that 25 minutes, click or no click. The idea here is that, can Google measure that, theoretically, yes - but is it a ranking factor? That is what we are going after with this test.
Can we send users to a page, have them site there for a decent amount of time to show that this is a good content versus other test pages that aren’t getting that dwell time even though we are not talking about a click – so bounce rate is not really what we are talking about here, we are just looking at coming to a page and sitting on it. And if you do a decent research on this, you will see that a lot of people thing that yes, this is a ranking factor.
This test was put together by Ryan Hough who did a really good job at this. Dwell time is a really weird concept. There is a good content here and we took some concepts here – feel free to go over the research if dwell time is a little bit fuzzy in my explanation there.
What we wanted to do was to send visitors just to a page, not clicking on it, in a sense, it is a bounce but they stayed on the page a fair amount of time. Some people are pointing in that dwell time is an actual thing is a feature that used to exist – it does not exist anymore but Google used to say – would you like to block these results from the serps or would you like more of this results, learn more about this. Where people thought where that determination or feature was coming from was from dwell time – if you went to a page and bounced immediately and your dwell time was low.
Google was asking would you like to take this out from your personal results vs when you are spending a lot of time even if you were not clicking but your dwell time was high, Google would offer you more options to see more from that site. That is where speculation came from, that dwell time was being taken into consideration. These features do not exist anymore. It went away maybe 2015 or so.
In this test set up, Ryan was using BrowSEO, if you are not familiar with BrowSEO, it is a pretty tool where you can essentially create fake personas and have people that live within your BrowSEO browser and you can switch between them without closing your browse.
Within the platform itself, it will maintain the passwords, IPs, other user data that a real person would have, and you can bounce between different personas and they are seen as different people online. So, using BrowSEO with 3 different personas and we are sending them to the test page to sit on the page and leave so that the dwell time would go up.
The concept here is that, if dwell time was a factor, the test page would go up even though essentially they have the same bounce rate and not really recording time on site.
Interestingly, what ended up happening on this test which was totally unexpected. The test page hit the bottom and stayed at the bottom. Very surprising results. Do I think that dwell time is a negative ranking factor? I don’t but there are some concepts as to what might have gone wrong here. It is entirely possible, and this is the scary part, is that Google identified this as not real traffic.
Ryan was doing this on a Friday, they all came from the same Amazon IP, since he was the user even if there were three different BrowSEO personas, he was basically typing in the same way, the amount of time spent on all the pages were similar, and they weren’t logged in users. It is possible that Google has identified these as fake.
That, to me, is I think the most unlikely result, I think the last one would be possible – they are not logged in as users and it is conceivable in my mind that Google would track it if these people were logged in to Chrome. That is where the user engagements might come from, where it has the ability to actually track these things is through logged in users.
That is a little bit more difficult to emulate in a test environment but that to me tells me that in my mind, that is where the money is at – that Google wasn’t tracking these things, it didn’t see these as different dwell times because the users were not logged in. That is where I think where things went wrong. I have a hard time believing that dwell time is a negative factor and I also have a hard time believing that Google has identified them as fake users and punished the particular page.
I do not know why they would do that because all of the pages were being seen as fake views. Where I think maybe the issue lies is in they weren’t logged in which then did not give that page the benefit of the dwell time.
How many visit?
The visits weren’t that many. I think there were 5 to each page. Not a ton of visits but they were spaced out in a certain amount of time. I think 2-3 weeks. Actually, I might have done it over a month, now that I think about it. So it was not an incredible amount of visits and what actually might happened to that test page because it dropped was simply it could have been an algo shift. Since all the pages were the same, Google just recalculates and reshuffles the test pages, kind of within the time that the test started and completed.
Does the H1 position affect ranking (Above or below H2)?
The idea is, we are putting h tags out of position. In the first tests that we talked about, we are looking at can you beat one signal with another signal. In this situation, what if we put these signals out of order? I am a proponent of a roman numeral outline set up. The idea is what will be your roman number 1 will be your H1, you’re a, b, c will be your H2s and your 1, 2, 3, will be your H3s and other things. That is kind of the structure that I think Google wants to see. What happens if you start putting things out of order? Can you harm yourself when you do that?
This test was done by Derek Booth, who in this test put pages out of order. For the number 3 page, when the H tags were added, in this situation, the H2s were placed above the H1. It is not in the right order. Our initial thought when we were discussing this test set up was that 1 H2 out of order isn’t going to do anything. You see it all the time too, in design, for some reason there is an H4 and H5 at the top of the page and it ranks well.
But in this environment where we are going to bring it down to a single variable, are you going to harm yourself or are you damaging yourself in any way. This might not be the hugest signal but is it possible? Our thought that nothing was going to happen here. For a little bit, the page didn’t move. Hilariously, the test page dropped to the bottom and it stayed at the bottom.
Should you run out now and freak out and light everything on fire and change all of your pages? No, you shouldn’t do that but when you are consulting with your clients about setting up a page from the beginning or hey, this page isn’t ranking very well, what should we do? – look at the order of your signals. Do you have them in proper order? That would be one small signal that I think you could send to Google that you know what you are talking about, you know what you are doing, you are setting things up the way that Google wants to see it.
At the end of the day, I think it is death by a thousand cuts, why shoot yourself in the foot when you do not have to. Put your H1 at the top. Use only one H1. Your supporting content should be structured in a way that makes sense so your H2s then next are your H3s, and you would cycle through – H2, H3, H4, H2, H3, H4, or how you’d do the page.
I wouldn’t just kind of toss my H tags willy nilly and say that Google will figure it out. I think people want to give Google a lot more credit than Google deserves at this point. We know that Google can’t pick out a keyword from LSI, if you say sink, refrigerator, stove, Google doesn’t know you are talking about a kitchen, you would need to put the work kitchen on that page or something that Google thinks is the word kitchen but the LSI is not it.
Should you throw things out of order just because they look pretty? I think no, I think this shows that you should have it in order that Google wants to see and that is your H1, H2, H3, in that order. Then from a previous test, what does it matter, my competitors are using H2s and I am just going to use an equal number of H3s and maybe some more and we will be fine because the over all optimization is the same, it is not. I don’t think it is.
I think that you want to make things in the right order and I think you want to have things in the amount that Google is looking for.
Is it best for the three headers to be H1, H2, H3, then you can change it around?
How I personally structure my pages are let’s say I am talking about tacos in Phoenix, that would be my H1. My H2 would be hard shell, my next H2 would be soft shell. Those would be my 2 supporting concepts for tacos in Phoenix. My H3s might then be what goes into making my hard shell, my h3 down below would be what goes into making my soft shell. That is how I would structure my page.
I would not necessarily move things around. I would kind of keep it in that structure so that my most important supporting points would be my H2s and then the points that support my H2s would be my H3s. I almost always put an H4 at the bottom, when it is all said and done.
Testing the strength of internal links in a 30-page test
We have looked at internal links and we have found that it is 2-3 internal links will equal one external link of the same value. Just throwing out random numbers, if your trust flow is 20 and you’ve got 3 internal links pointing to a page and you’ve got an external link that would equal an external link of 20.
Just a real basic example. Two to three internal links is one external links. This was done in a smaller testing environment – a 10-page test. Ryan wanted to see this in a 30-page test. In a sense, will we see this same movement in a 30-page test because there are so many pages, will we see that these internal links while they might have that same value, maybe they won’t push all the way to the top. Maybe you started at the 30 spot and you’d only get to page two or bottom of page one and not all the way up.
We are really playing with a lot of larger field test to see how these techniques respond and we are also kind of looking for a way to see if we can see what is stronger than other things.
In this test he has 30 pages and creates silo’d posts for that last result and points them in. After pointing one link, he moved from 30-18 which is a pretty significant jump, moving 13 spots. Once the second link hits, he moved all the way to the top – from 18 to 1. Contrasting this to earlier tests where we were playing with internal links.
In the earlier test, we were actually just using pages with a typical page structure. With these particular pages, we were using more of a silo’d structure and pointing them in – kind of as you would with WordPress blog post. In this situation, it just took two links to move up. I think the reason that I liked this test the most is that my internal linking structure appears to work a lot better.
This is something that I think everybody could do. When you comment to a new client’s site, especially if they have a blog, they will find that they have a lot of content that they have written that is probably great content and it is probably not well structured in order to support a ranking page. It might be about the content of a ranking page but they probably structured it poorly.
It is a real opportunity to manipulate the url structure to get it into a silo or a virtual silo and then get a link on that post and point it to your ranking page to get the value of that post. Two or three of those, you are going to get the same external value.
When I presented in Rockstars in 2015, I showed that I moved the site that wasn’t ranking at all just on the basis of three years’ worth of blog posts and ended up with 7 or 8 first page rankings. You can do that. I am actually doing that right now with another client’s site where we are pushing things up where they are stuck in 7,8,9 position.
Now, we are moving them up to 3,4,5. Those last bits of slug when you are really fighting for the page to move up, you can still get the benefits of these internal links. It is also one of the best ways to get a good bang for your buck coming in. You don’t have to build all these links. If the site has value, take what you have. Put it into a silo structure and start pointing it out at the pages that you need to rank and you could get some really good movement. This test is a nice validation of that technique in a test environment.