How Google Hummingbird Really Works: What We Learned by Analyzing 9.93 Million Words of Content

hummingbird

Search engine algorithms are super complicated.

Our task as SEOs is to figure search algorithms out. It’s not easy, especially since Google’s search engine uses more than 200 ranking factors. And, Google isn’t going to just tell us how they rank sites.

So, we have to figure it out.

How do we figure it out? With data — lots of data. What you’re reading now is one of the biggest data-driven studies on the subject that I’ve ever done.

My goal was simple. I wanted to figure out exactly how Hummingbird works.

Why Hummingbird? Because it’s the biggest search algorithm change ever. Every single Google search is controlled by Hummingbird.

If I could crack the Hummingbird code, I could get my websites to kill it in the SERPs.

But, I didn’t want to keep all this data a secret! It cost a lot of money, and I had a lot of help from researchers, engineers, data scientists, programmers and data providers from MarketMuse.

This article tells you exactly what we found after analyzing hundreds of thousands of pages and millions of words of content. More importantly, we analyzed the semantic groupings of words, organized them into topic clusters and analyzed this against a database of more than 31.5 million relevant content items.

This study is the most comprehensive of its kind and reveals how Hummingbird works.

If you look at our data and read our takeaways, you’ll understand exactly how to crack Hummingbird wide open.

Here’s how we crunched the data

We observed that Hummingbird ranks web pages that have comprehensive and thorough content on a narrow topic higher in the SERPs.

In other words, if a website has a lot of really good, deep content, it’s going to rank well in the search results.

But, what kind of content exactly?

We wanted the data to tell us.

To show you exactly how the Hummingbird search algorithm works, let’s look at a popular niche — personal finance.

We chose some of the most authoritative sites dealing with that topic, so there would be some subject overlap.

image09

To get the most power from the data, MarketMuse built a semantic search analysis algorithm, similar to the one that Hummingbird uses.

Why go to the trouble of using semantic search analysis? Because Hummingbird is all about semantic relationships. The only way to get the right data was to semantically relate each piece of content to all of the other topics in the niche.

Each section below shows you four things:

  1. The site that we crawled.
  2. The Hummingbird grade it received. MarketMuse assigns a letter grade to the site as a whole. This is helpful to understand, at a glance, how a site is performing.
  3. The topic of the site. This is important, because a site’s declared topic is how Hummingbird will rank it. Hummingbird figures out the topic of the site by analyzing the most important pages (home page and About page), and by crawling the main navigation.
  4. Our findings. In each section, we’ll show the various topics and how the content performs. One of the important data points here is the “keyword rankings per page.” This is the average number of keywords, ranking in Google’s search results, that each page has. Higher is better, because it means that a given page could potentially rank for dozens of search queries. (We measured all US search queries greater than five per month.)

Download this quick step by step guide to analyzing your web site’s hummingbird performance. This is more important than ever with your target audience using search engines on their mobile device many times every day.

We scored each site based on three things:

  1. Satisfactory Coverage: This number reports how often the topic/search keyword appears on the site. In order for a topic to be considered satisfactorily covered, it should appear on a minimum of 2.5% of the pages of the site. All of the keywords/topics that we detail below have satisfactory coverage.
  2. Relative Gap Topics: A high relative gap number is bad. It shows that the page mentioned few keywords that are related to the topic under discussion.
  3. Absolute Gap Topics: Absolute gaps take place when a site has no site-wide mention of a topic or keyword that is considered semantically relevant to the stated topic of the website. This is especially bad news. The higher the absolute gap topic score, the worse the site performs in Hummingbird search results for relevant search queries.

Now, let’s check out the data.

Magnify Money

image04

Website:  MagnifyMoney.com

Grade:  A-

Topic:  Here’s how MagnifyMoney describes their goals and user intent:

Reduce credit card interest (and therefore debt), eliminate all unnecessary fees, maximize cash back on everyday spending and earn interest on all savings accounts. These four simple things can make a big difference and ensure that more money stays in your wallet.

Findings

MagnifyMoney.com was the number two most successful website in the search engine rankings, by words ranked per page.

This was surprising to us. Why? Because their link profile is tiny. They rank lower in trust and authority, yet receive high results.

Let me pause here for a second, because this is insanely significant.

All throughout search engine optimization history, backlinks have been the single biggest reason for a site’s rankings. A stronger backlink profile means better ranking on search engines, right?

This isn’t true anymore. Not with Hummingbird.

You’re looking at the evidence.

Here is the backlink profile for MagnifyMoney. It’s tiny.

image03

Domain and trust are low.

Now, look at the backlink profile for another of the sites we analyzed, GetRichSlowly.com.

image00

It has twenty times the number of backlinks, plus a higher authority and trust score.

But, the little guy, MagnifyMoney, outranks it. Why?

Because of the Hummingbird search technology.

MagnifyMoney has content that appeals directly to the search engine ranking factors that are important in Hummingbird.

All this, in spite of the fact that its link profile is really slim.

Here’s the bottom line: A Hummingbird-optimized site will outrank a huge link profile, making your site more visible to the target audience who enter a relevant search query or key phrase into the search box.

If you don’t have the budget or time to build up your backlink profile, don’t worry. As long as you are creating kick-ass content that appeals to Hummingbird’s natural language processing, you can outrank the big boys.

Let’s take a look at just how effective their content is and how it appeals to Hummingbird.

This is their overall topical depth coverage. Note that they are in the green for most topics.

What this means is that they have great coverage and depth, as it relates to their selected topics and the niche as a whole.

Personal loan rates

image07

Checking account fees

image32

High interest saving accounts

image19

Cash back credit cards

image27
screen-shot-2016-11-10-at-6-55-50-am

MyMoneyBlog.com

image02

Website:  mymoneyblog.com

Grade:  C+

Topic:  After taking a glance at the homepage, one would assume the site is about early retirement, investing, saving money and credit card rewards.

We pulled this directly from their top navigation. So, how well do they actually cover these topics?

Take a look at the semantic search results:
screen-shot-2016-11-10-at-6-46-07-amMyMoneyBlog.com is a great place to settle in for a while and unpack the data. If we can understand why it has so many gaps (and what to do about it), we’ll be able to discover exactly how a Hummingbird-friendly site should look.

Findings

The “early retirement” topic — the first one in the list above — has very little coverage and a high number of gaps.

The site should treat this topic, but the content does not reflect it. As a result, the keyword rankings per page are low.

screen-shot-2016-11-10-at-6-47-13-am

Let’s look at each topic individually.

Early Retirement

image30

screen-shot-2016-11-10-at-6-47-43-am

Here is where they missed out.The topical depth coverage of the term early retirement is 45%, which is a considerably poor score, especially considering the fact that the site itself seems to promise content on the topic.

  • There are 25 relative gaps, meaning 25 related topics have thin topic coverage.
  • There are 15 absolute gaps, meaning 15 related topics have no coverage at all.
  • A quick scan of the top 5 relevant topics also reveals two major title gaps. The related topic “life expectancy” only appears in one title across the entire site. And, the related topic “retirement age” appears in zero titles.

An absence of related content (or even thin related content) sends a negative signal to the Hummingbird search algorithm that you might not be an authority on the focus topic.

If you were a competitor, you’d want to write on long-tail keyword topics like worker compensation, early retirement age and disability retirement. They totally missed these topics.

We can also examine a specific related search term to better understand how Hummingbird might factor that into the equation of measuring the breadth and depth of coverage on a given subject. For example, let’s look at the related term social security.

In our analysis, this topic — social security — is the most relevant search phrase topic connected to the focus term, early retirement.

294 pages on MyMoneyBlog.com mention social security, which is pretty good. But, how are those articles actually performing in the semantic search results?

Are they comprehensive? Or, is it just a bunch of fluff? Our analysis revealed the exact depth of coverage:

image05

Tip: A low score shown in red means bad news. You can use MarketMuse to check your site and get a similar content score like this.

We can see here that the articles containing the related topic social security are actually low performing pieces.

None of the articles are over 1,000 words and they only mention social security a few times. All of the articles shown above received a low content score from our content audit.

It’s not enough to use a given keyword. Sure, they sprinkled it across the website and wrote some decent-sized articles (200-900+ words) that included the topic.

But, as far as depth goes, they missed the goal entirely.

What should MyMoneyBlog do to remedy this problem?

They should write a lengthy, solid, comprehensive, deep, home-run article on social security. Make the article 4,000 words, add images, include authoritative discussion and create the best piece of content on the internet on that topic, followed by effective social media marketing.

This would strengthen the overall performance of that highly related term, which would then strengthen signals to Hummingbird for the breadth of coverage for the core focus topic, early retirement.

Hummingbird is smart enough to look beyond the initial topics. It examines the full set of related topics to see if there’s a bridge between them. Ignoring those related topics or covering them superficially can set you back.

Investing

Take a look at their coverage of the topic “investing.”

Depth is approaching decent, but there are a high number of relative gaps.

screen-shot-2016-11-10-at-6-50-25-am

This tells us that the site has considerable breadth on the topic, but not much depth.image24

Quick tip: Hummingbird wants breadth and depth.

For example, the word investing appears in 154 titles across MyMoneyBlog. That’s healthy coverage on the core topic, but it’s not enough.

For Hummingbird to categorize your website as the true expert in a field, the related topics need to be addressed thoroughly, as well.

In this case, we can see that the highly related topic long-term care doesn’t appear in any titles.

image26

Actionable Insight

If you were trying to rank higher on search engines for the topic investing and related search queries, you’d be smart to write an article focused on long-term care and include that keyword in the title and anchor text of your article.

You’d want to look at the other 35 relative gaps and build content around those search terms, too.

Saving Money

How did they do on the topic “saving money?”

screen-shot-2016-11-10-at-6-50-44-am

image14

The topic saving money has the strongest topical coverage of the four focus topics that we chose to examine. It clocks in at 67%.

Still, the topic has 23 relative gaps and 5 absolute gaps. Not good.

Actionable Insight

Topical gaps are easy to miss. For the term saving money, MyMoneyBlog has two relative gaps near the top of our list of related topics.

What are they? Gas mileage and fuel economy.

image37

You see why it’s easy to miss? You might not automatically think that “gas mileage” is a topic that you should be discussing in a personal finance blog. Yet, apparently, according to the MarketMuse algorithm and semantic search analysis of close to 80-100 authoritative pages on the topic, this is how real people enter a search query and think. It’s something that is relevant.

A content director might have a hard time spotting this gap, but Google Hummingbird can quickly make the connection.

So, overall, the site’s coverage of money-saving tips is good. But, it’s quite thin on tips specifically related to driving.

In our research, we see a strong correlation between related topic coverage and topical authority. Identifying the gaps is the hard part.

Credit Card Rewards

How about “credit card rewards?” We’re seeing a trend. Depth coverage is okay, but the high number of relative gaps is astonishing.

screen-shot-2016-11-10-at-6-51-09-am

image31

The term credit card rewards shows similar performance to the other terms that we examined. The coverage is slightly above average, but we still see plenty of relative topical gaps.

Actionable Insight

Big sites with average or sub par performance tend to have an abundance of ranking keywords and not enough long-tail keywords.

The problem? Those words don’t rank very high.

In the knowledge graph below, mymoneyblog.com is so low that it’s almost off the screen, when compared to its closest competitors for high volume keywords. (MyMoneyBlog is located in the very bottom right hand corner.)

image01

Why does it rank so low? Because there isn’t enough focus. The site seems to take on too many subjects, but doesn’t cover them fully. Remember all the relative topic gaps?

If you care about high search engine rankings, don’t try to cover everything. Stick to a niche and dominate it.

If MyMoneyBlog wants to rank for the long-tail keyword “credit card rewards,” for example, they need more than a smattering of articles that reference the search term. They need several high-octane, long form, deep-dive articles on the subject, plus articles dealing with the topics related to credit card rewards.

It’s better to leave off addressing a topic entirely, than it is to go at it with half-baked effort.

Get Rich Slowly

image28

Website:  http://www.getrichslowly.org

Grade:  D

Topic:  Get Rich Slowly is a personal finance site that helps individuals and families gain financial security. They go in-depth on issues such as debt elimination, saving money and practical investing.

Here’s how the site is doing in keyword rankings per page.

screen-shot-2016-11-10-at-6-51-50-am

There are some Hummingbird-related concerns, when it comes to topic coverage.

The site tries to cover every imaginable subject in the “personal finance” sector. The more trust and authority your website has, based on fully covering narrower topics, your link profile and other well-executed search engine optimization strategies, the more that you can expand topic coverage.

It’s important to not be all over the map in your subject matter. Personal finance is a huge space and it requires that you further niche yourself when you begin creating content on the subject. Focus on the most relevant long-tail keywords across your content, anchor text and social media marketing, to achieve high ranking on search engines.

There is a disconnect with their “most popular categories,” the main topics on their navigation bar and what they say in their “About” section.

We’re not sure they know what they’re about. Perhaps the best indication of the disconnect is on their “About” page.

  • It is linked only in a footer after over 7,000 words of scrolling text (these are recent blog headings and the initial text from each piece).
  • It sits under four paragraphs of “disclosures,” which are easier to find.
  • It is about four font sizes smaller than the body text.
  • The font size is the same as Helvetica 12 point, which is barely readable on today’s computers.

Do you tell your readers and, by extension, Google’s semantic search algorithms, precisely what your website is about?

Do you link to that clearly from navigation on the top of your page and/or the side, with clear anchor text? Better yet, do you make it clear from simple content or navigation on your homepage what readers and Googlebots should expect to find?

The text on your homepage can be the most important on your website. If your homepage has half the external and internal links to your pages, as in most cases, every phrase on it is far, far more important than what’s on the internal pages.

Findings

Here is how the topics of Get Rich Slowly perform.

Savings accounts

image06

CD rates

image35

Credit cards

image10

Insurance rates

image29

Bank reviews

image21

They mention certain banks a LOT, yet other large banks, like TD Bank, almost not at all.

Is there a reason Ally Bank needs so much coverage? They claim to provide “bank reviews,” yet don’t say they are focused on particular banks.

Hummingbird is all about answering the intent of user search queries, not specific keywords themselves. You are not going to rank for “bank reviews” related keywords, if you are only covering a select number of banks and leaving out other important banks.

The Penny Hoarder

image18

Website:  http://www.thepennyhoarder.com

Grade:  A-

Topic: The Penny Hoarder is one of the world’s largest personal finance websites. The site proclaims that they have over 16 million monthly readers and more than 5 million active subscribers. That’s bragging rights, as shown in large font above their social media channel links.

Their objective is very clearly stated.

Our goal is to improve the lives of everyday folks by helping them spend less time worrying about their finances and more time enjoying their lives.

Let’s take a look at some of their scores.

Findings

screen-shot-2016-11-10-at-6-52-20-am

Personal saving tips

image08

Earn extra money

image15

Saving money

image25

The data that stands out for The Penny Hoarder is the low number of absolute gap topics. That’s a good thing.

The Penny Hoarder is steering clear of generic content and they are trying to cover topics that are super relevant.

For these reasons, the content of Penny Hoarder directly appeals to Hummingbird.

A Step-by-Step Guide to Analyzing Your Site’s Hummingbird Performance

If you are a MarketMuse user, here is how you can replicate our study and analyze your own site’s performance in Hummingbird.

In this example, I’ll use http://www.thepennyhoarder.com.

First, open MarketMuse.com and log in.

Second, click on “Advanced Tools” in the top header and then “Content Audit.”

image36

Enter your URL.

image22

Add your “focus topic.”

image16

Click the “In Content” option.

image33

Click “Analyze.”

You will see a list of all your site’s page titles, along with the number of focus topic mentions, word count and content score.

The important number here is “content score.”

The software uses a color-coded scoring system.

  • Green is good
  • Yellow is caution.
  • Red is bad.

The fastest way to improve your site is to first work on the pages that have a content score below nine, indicated in red.

image13

Find the first red scored page and click “Improve.”

image12

You’ll see a complete content crawl of the page, along with a list of related topics in the right pane.

image11

Be sure to analyze the list of related mentions. If you see a term that you don’t think is relevant, you can remove it. When you mouse over the list, you’ll see a small blue “X” appear next to a topic. Click the “X” to remove it from the list.

Now, download the list of related topics in an Excel spreadsheet.

image34

Scroll down to the listings of “top ranking pages for ‘personal finance.’”

image38

Scroll through the list and find all of the articles with a green content score (above 20).

image20

Click “Compare” on the first green scored page.

You’ll see your page compared with the top-ranked competitor, based on “related mentions.”

image17

Create a list of these phrases or download the data.

Repeat for any additional high-scoring articles in the list.

Now comes the fun part. Rewrite the article using as many of the “related topics” phrases as possible.

This will require some time. You don’t want to carelessly throw the terms in there. Instead, you want to maintain natural language structure, flow and authority, but use these relevant terms where appropriate.

Once you rewrite and publish the content, it will begin to rank higher on search engines. In 1-2 weeks, analyze the content again to see how your score has changed.

Repeat this process for each of your low-scoring pages.

Once you have improved all of your website pages with a red score, move on to the pages with a yellow score.

Over time, you can completely optimize every low-performing page on your site, creating a site that dominates the SERPs.

This is the data-driven method for guaranteed success with the Hummingbird search algorithm and Google’s Knowledge Graph.

It’s not about links anymore. It’s about an entire approach to creating comprehensive, deep and authoritative content.

8 Practical Things That You Can Do to Improve Your Rank in Hummingbird

To rank in Hummingbird, you want to create a website with comprehensive and thorough content.

How do you do that? Here’s a handy list of eight things that will help you crush it in the Hummingbird era.

  1. Select, refine and state your site’s topic using a clear purpose statement, above-the-fold content and specific navigation elements. (Don’t be content with fuzzy or broad statements.)
  2. Create long form content. (Avoid short content.)
  3. Create in-depth content. (Avoid generic content.)
  4. Summarize the purpose and intent of the site with specificity and directness. (Don’t hide your purpose or make it vague.)
  5. Create content that appeals to readers (Don’t create content for search engines.)
  6. Create focused content. (Don’t try to provide comprehensive content on every sub niche in your niche.)
  7. Create a lot of content. (Don’t be happy with a few blog posts or evergreen pages.)
  8. Create content that is entirely relevant to your area of expertise. (Don’t write about off-topic subjects.)

By now, you know that SEO isn’t about cramming content with keywords.

But, it’s not about links, either.

Our data conclusively proves that top-ranked content doesn’t need a heavy backlink profile, let alone a specific keyword density.

What top-ranked content has in common is that it’s deep, comprehensive and authoritative.

If you can produce content like that, then your Hummingbird game is complete.

What have you learned about Hummingbird? What changes have you made as a result?

Share