luni, 22 septembrie 2014

Searchmetrics Ranking Factors 2014: Why Quality Content Focuses on Topics, not Keywords

Searchmetrics Ranking Factors 2014: Why Quality Content Focuses on Topics, not Keywords

Searchmetrics Ranking Factors 2014: Why Quality Content Focuses on Topics, not Keywords

Posted: 21 Sep 2014 05:15 PM PDT

Posted by searchmetrics

Searchmetrics recently launched their yearly Ranking Factors Study that bases numbers on rank correlation and averages of top 10 SEO rankings, and this year's analysis shows that content on top-performing sites is much more holistic and less keyword-focused.

Everybody talks about how "content is king." People are advised to "create quality content for users," and not ever since keyword (not provided), some have said "the keyword is dead." Though these phrases may convey somehow understandable approaches, they are often nothing more than empty clichés leaving webmasters alone with without any further information.

Making relevant content measurable

What is quality content? How can I create relevant content for my users? Should I still place the keyword in the title or use it seven times in the content?

To understand how search engines develop over time and what kind of features increase or decrease in prevalence and importance, we analyze the top 30 ranking sites for over 10,000 keywords (approximately 300,000 URLs) each year. The full study with all 100 pages of details is  downloadable here.

In a nutshell: To what extent have Panda, Penguin, and not least Hummingbird influenced the algorithm and therefore the search results?

Before we get into detail, let me—as a matter of course—point out the fact that correlation does not imply causation. You can find some more comprehensive information, as well as an introduction and explanation of what a correlation is, here. That is why we took two approaches:

  • Correlation of Top 30 = Differences between URLs within SERP 1 to 3
  • Averages = Appearance and/or extent of certain factors per position

The "Fall" of the Keyword?

Most keyword factors are declining. This is one of the major findings of our studies over the years. Let me give you an example:

The decrease of the features "Keyword in URL" and "Keyword in Domain" is one of the more obvious findings of our analyses. You can clearly see the declining correlation from 2012 to 2014. Let's have a look at some more on-page keyword factors:

What you see here as well are very low correlations. In other words: With regard to these features, there are no huge differences between URLs ranking on positions from one to thirty. But there is more than that. It is also important to have a look at the averages here:

Explanation: X-Axis: Google Position from one to 30 / Y-Axis: Average share of URLs having keyword in description/title (0.10 = 10%). Please note that we have modified the crawling of these features. It is more exact now. This is why last year's values are likely to be actually even a bit higher than given here. However, you can see that relatively few sites actually have the keywords in their headings. In fact, only about 10% of the URLs in positions 1-30 have the keyword in h2s; 15% have them in h1s. And the trend also is negative.

By the way: What you see in positions 1-2 is what we call the "Brand Factor." It is often a big brand ranking on these positions, and most of them differ from the rest of the SERPs when it comes to classic SEO measures.

Actually, taking only correlation into consideration can sometimes lead to a false conclusion. Let me show you what I mean with the following example: 

The correlation for the feature "% Backlinks with Keyword" has considerably increased from 2013 to 2014. But the conclusion: "Hey cool, I will immediately do link building and tell the people to put the keyword I want to rank for in the anchor text!" would be a shot in the dark. A glance at the averages tells you why:

In fact, the average share of links featuring the keyword in the anchor text has declined from 2013 to 2014 (from ~40% to ~27). But what you see is a falling graph in 2014 which is why the correlation is more positive with regard to better rankings. That means: the better the position of a URL is, the higher the share of backlinks that contain the keyword (on average). On average, this share continuously decreases with each position. In contrast to last year's curve, this results in the calculation of a high(er) positive correlation.

Conclusion: The keyword as such seems to continue losing influence over time as Google becomes better and better at evaluating other factors. But what kind of factors are these?

The "rise" of content

Co-occurrence evaluations of keywords and relevant terms is something we've been focusing on this past year, as we've seen high shifts in rankings based on these. I won't go into much detail here, as this would go beyond the scope of this blog post, but what we can say is that after conducting word co-occurrence analyses, we found that Proof and Relevant keywords played a major role in the quality and content of rankings. Proof Terms are words that are strongly related to the primary keyword and highly likely to appear at the same time. Relevant Terms are not as closely related to the main keyword, yet are still likely to appear in the same context (or as a part of a subtopic). These kinds of approaches are based on semantics and context. For example, it is very likely that the word "car" is relevant in a text in which the word "bumper" occurs, while the same is not true for the term "refrigerator."

Proof and relevant terms to define and analyze topics

Let's have a look at an example analysis for Proof and Relevant Terms regarding the keyword "apple watch," done with the Content Optimization section of the Searchmetrics Suite:

The number behind the bar describes the average appearance of the word in a text dealing with the topic, the bar length mirrors the respective weighting (x-axis, bottom) and is calculated based on the term's semantic closeness to the main keyword. Terms marked with green hooked bubbles are the 10 most important words, based on a mixed calculation of appearance and semantic weighting (and some further parameters).

As you can see, the terms "iphone" and "time" are marked as highly important Proof Terms, and "iwatch" is very likely to appear in the context of the main keyword "apple phone" as well. Note that simply reading the list without knowing the main keyword gives you an idea of the text's main topic.

The above chart shows an excerpt from the list of Relevant Terms. Note that both the semantic weighting and the appearance of these terms is somewhat lower than in the previous chart. In contrast to the Proof Terms list, you won't know the exact focus of the text just looking at these Relevant Terms, but you might probably get an idea of what its rough topic might be.

Content features on the rise

By the way, the length of content also continues to increase. Furthermore, high-ranking content is written in a way that is easier for the average person to read, and is often enriched by other media, such as images or video. This is shown in the following charts:

Shown here is the average text length in characters per position, in both 2014 and 2013. You can see that content is much longer on each and every position among the top 30 (on average) in 2014. (Note the "Brand Factor" at the first position(s) again.)

And here is the average readability of texts per position based on the Flesch score ranging from 0 (very difficult) to 100 (very easy):

The Flesch score is given on the y-axis. You can see that there is a rather positive correlation with URLs on higher positions featuring, on average, easier-to-read texts.

But just creating more (or easier) content does not positively influence rankings. It's about developing relevant and comprehensive content for users dealing with more than just one aspect of a certain topic. The findings support the idea that search engines are moving away from focusing on single keywords to analyzing so-called "content clusters" – individual subjects or topic areas that are based around keywords and a variety of related terms.

Stop doing "checklist SEO"

So, please stop these outdated "Checklist-SEO" practices which are still overused in the market from my perspective. It's not about optimizing keywords for search engines. It's about optimizing the search experience for the user. Let me show you this with another graphic:

On the left, we have the "old SEO paradigm: 1 Keyword (maybe some keyword variations. we all know the " An SEO walks into a bar joke") = 1 Landing Page – Checklist SEO. That's why, in the past, many websites had single landing pages for each specific keyword (and those pages were very likely to bear near-duplicate content). Imagine a website dealing with a specific car having single landing pages for each and every single car part: "x motor," "x seats," "x front shield," "x head lamps," etc. This does not make sense in most cases. But this is how SEO used to be (and I must admit: the pages ranked!).

But, to have success in the long term, it's the content (or better, the topic) that matters, not the single keyword. That is why landing pages should be focused on comprehensive topics: 1 Landing Page = 1 Topic. To stick with the example: Put the descriptions of all the car parts on one page.

Decreasing diversity in SERPs since the Hummingbird update

How these developments actually influences the SERPs can be seen in the impact of Google's Hummingbird. The algorithm refactoring means the search engine now has a better understanding of the intent and meaning of searches which improves its ability to deliver relevant content in search results. This means search engine optimization is increasingly a holistic discipline. It's not enough to optimize and rank for one relevant keyword – content must now be relevant to the topic and include several related terms. This helps a page to rank for several terms and creates an improved user experience at the same time.

In a recent analysis on Hummingbird, we found that the diversity in search results is actually decreasing. This means, fewer URLs rank for semantically similar ("near-identic") yet different keywords. Most of you know that not long ago there were often completely different search results for keyword pairs like "bang haircuts" and "hairstyles with bangs" which have quite a bit of overlap in meaning. Now, as it turns out, SERPs for these kinds of keywords are getting more and more identic. Here are two SERPs, one for the query "rice dish," and one for the query "rice recipe," shown both before and after Hummingbird, as examples:

SERPs pre-Hummingbird


SERPs post-Hummingbird

At a glance: The most important ranking factors

To get an insight of what some of the more important ranking factors are, we have developed an infographic adding evaluations (based on averages and interpretations) in bubble form to the well-known correlation bar chart. Again, you see the prominence of content factors (given in blue). (Click/tap for a full-size image.)

The more important factors are given on the left side. Arrows (both on the bubbles and the bars) show the trend in comparison to last year's analysis. On the left side also, the size of the bubbles represents a graphic element based on the interpretation of how important the respective factor might probably be. Please note that the averages given in this chart are based on the top 10 only. We condensed the pool of URLs to SERP 1 to investigate their secrets of ranking on page 1, without having this data influenced by the URLs ranking from 11 to 30.

Good content generates better user signals

What you also notice is the prominent appearance of the factors given in purple. This year we have included user features such as bounce rate (on a keyword level), as well as correlating user signals with rankings. We were able to analyze thousands of GWT accounts in order to avoid a skewed version of the data. Having access to large data sets has also allowed us to see when major shifts occur.

You'll notice that click through rate is one of the biggest factors that we've noticed in this year's study, coming in at .67%. Average time on site within the top 10 is 101 seconds, while bounce rate is only 37%.

Conclusion: What should I be working on?

Brands are maturing in their approach to SEO. However, the number one factor is still relevant page content. This is the same for big brands and small businesses alike. Make sure that the content is designed for the user and relevant in your appropriate niche.

If you're interested in learning how SEO developed and how to stay ahead of your competition, just download the study here. Within the study you'll find many more aspects of potential ranking factors that are covered in this article.

Get the Full Study

So, don't build landing pages for single keywords. And don't build landing pages for search engines, either. Focus on topics related to your website/content/niche/product and try to write the best content for these topics and subtopics. Create landing pages dealing with several, interdependent aspects of main topics and write comprehensive texts using semantically closely related terms. This is how you can optimize the user experience as well as your rankings – for more than even the focus keyword – at the same time!

What do you think of this data? Have you seen similar types of results with the companies that you work with? Let us know your feedback in the comments below.


Sign up for The Moz Top 10, a semimonthly mailer updating you on the top ten hottest pieces of SEO news, tips, and rad links uncovered by the Moz team. Think of it as your exclusive digest of stuff you don't have time to hunt down but want to read!
Publicat de 1 la 09:34

Niciun comentariu:

Trimiteți un comentariu

Abonați-vă la: Postare comentarii (Atom)