Hummingbird In The Trenches: A Canary In The Coal Mine

I’ve been gathering and analyzing reports and in-the-trenches data from in-house marketers about Google’s Hummingbird, and I believe what I am seeing is on par with finding a dead canary in a coal mine. Pardon the morbid analogy, but the writing is on the wall, folks, and it has been for a while. Change is here, now. 

The following observations from other SEO experts and in-house marketing professionals all lead me to conclude that Hummingbird marks a new paradigm of SEO that revolves around concepts and not keywords.

Keep reading for compelling evidence and, more importantly, what you should be doing about it as marketers.

New SEO Paradigm

With the advent of Hummingbird and recent changes to the lack of availability of keyword data, it is clear that the SEO landscape is one that is often wrought with change and ambiguity. With the advent of these recent changes, a new paradigm of SEO is evolving that is less about keywords and more about concepts and topics.

This isn’t a new idea — Bill Slawski predicted this in an article titled “Should you be doing concept research instead of keyword research?” over a year ago — but in recent months, the shift in focus on concepts and not just keywords has become more noticeable.

We’ve heard directly from Amit Singhal about Google’s direction:

We’ve always believed that the perfect search engine should understand exactly what you mean and give you back exactly what you want. And we can now sometimes help answer your next question before you’ve asked it, because the facts we show are informed by what other people have searched for. […] We hope this added intelligence will give you a more complete picture of your interest, provide smarter search results, and pique your curiosity on new topics.

This is telling of the transition that has taken place on the Google search engine results page (SERP).

Today, signals such as keyword co-occurrence, user behavior, and previous searches do in fact inform context around search queries, which impact the SERP landscape. Note I didn’t say the signals “impact rankings,” even though rank changes can, in some cases, be involved. That’s because there’s a difference. Google can make a change to the SERP landscape to impact 90 percent of queries and not actually cause any noticeable impact on rankings.

Here’s how: given that the Google SERP is also now aiming to “pique your curiosity on new topics,” it appears that refined topic modeling, more knowledge graph references, and semantic signals are coalescing. As a result, they are driving a shift away from a traditional focus on keywords to a focus on concepts and topics. Still scratching your head? Read on.

Three In-The-Trenches Observations From SEO Experts

1. Markus Renstrum, Head of SEO, Yahoo

“What we have seen at Yahoo is that the strategy of the ‘topic page’ is no longer working. We have large databases of movies, musicians, news related topics that were interlinked from the articles and it worked great. Now, in general the search engine prefers to surface the latest article instead of a topic page, and again, this is in my mind a way of relying fresh, trusted content instead of static topic pages. They know Wikipedia will represent the topic page on every SERP, and the rest are the latest news from trusted sources.

Moving forward, for us it is all about being seen as a quality source; content, platform and authorship, and be surfaced on specific or generic terms with the latest story using the article page instead of a topic page.”

2. Michael Ngyuen, Director of SEO, Shopzilla.com

“Hummingbird is a continuation of Google’s effort to provide answers directly in search results. Within the retail query space, we’re seeing a clear trend of Google shifting traffic toward product pages at the expense of category and topical pages. As Google is able to improve its ability to understand the context of a page, I expect Google to do a better job of answering the intent of user queries. My recommendation for site builders is to focus on delivering great experiences that are direct answers to a user’s search. Google is going to figure out how to get users to those answers.”

3. AJ Kohn, Owner, Blind Five Year Old

In his article “The Hummingbird update,” AJ Kohn observed:

“Based on client data I think that the May 2013 Phantom Update was the first application of a combined topic model (aka Hummingbird). […] Hummingbird refined the topic modeling of sites and pages that are essential to delivering relevant results. […] But now they’ve been able to extract entities, understand the topics to which they refer and then feed that back into the topic model. So in some ways I think Hummingbird allows for a type of recursive topic modeling effort to take place.”

And, in further correspondence:

“Hummingbird blends natural language topic modeling with entity based topic modeling for a more precise measure of topics overall. Publishers simply need to use those subjects appropriately in their content (i.e., limit pronouns), link out to entities to establish graph relationships and implement schema.org markup when possible.”

Concept (AKA Topical Research) Vs. Keyword Research

To fully understand the difference between concept research and keyword research, it’s important to understand how search engines are evolving to use search entities to improve the relevancy of search results for a given query. “Search entities” (as opposed to entities that relate to people, places, and things) are primarily about relationships between keywords. Relationships between keywords or entities help provide a layer of semantic relevancy. Let’s simplify this further and look at a type of query that benefits from co-occurrence and one that clearly doesn’t.

When you search on the keyword [jaguar], it’s difficult for Google to determine whether you’re looking for information on a giant cat or a sports car. But add in the keyword [XK], and it becomes clear you’re interested in info about the car. The more keywords you add, the better the engine understands what it is you want. Refine your query to [Jaguar XK parts], and you get accessories. Change that to [jaguar xk for sale], and presto, more co-occurrence equals a better search result. Or does it?

Most of the time, it does equal a better search result; however, co-occurrence can occasionally yield strange results. Below is a good example.

On a Google search for [Where is the best albacore fishing in the world?], you will also find that there are results matched to pages about the game, World of Warcraft. While this is clearly not what the question was about, for the most part, Hummingbird appears to have been an improvement toward helping users answer questions.

In fact, we’ve noticed that it even appears to have impacted paid search. Impressions of question-related keywords have spiked by 50 percent in broad match paid search campaigns compared with several months prior to the Hummingbird announcement.

For Google to properly answer a question such as the query above, it would also have to know that there is a relationship between albacore and fishing and that because albacore are migratory, the query is also temporal. Thus, the most relevant result for the best fishing spot in the world will be different based on the freshest content and not an aged article with strong link signals or co-occurrence signals.

That’s a lot to ask of a search engine, which is why search engines will continue to invest in further exploration of search entities as well as public and commercial knowledge graphs. As they continue to do so, the SERP landscape will continue to evolve into more of a blend of topical and specific responses to queries, giving more cause for marketers and/or publishers to consider different strategic approaches to provide relevant, optimized content for both types of responses.

This means that keyword research should go beyond search volume analysis, toward uncovering different concepts and aspects related to the keyword as well as uncovering related queries that people might use when they search for background information related to that keyword.

The Evolution Of Keyword Research

The biggest challenge of concept- or topic-based research is that it can represent a paradigm shift for how SEO influences change in an organization, content strategy, or even content hierarchy and categorization decisions.

We have traditionally focused on keywords and search volume data surrounding those keywords to drive our SEO strategies. Going forward, we will also expand our focus to include this idea of concept research. Not only because it is what search engines are moving toward, but because it helps us better organize and add more depth to our content to the benefit of our audience. Our focus on creating valuable content to answer searcher’s questions is not going away, but we need to adjust it to include a more thoughtful and comprehensive approach that might also include more background information and related topics.

In the past, we used the Google Keyword Tool to conduct keyword research. Today, Google Keyword Planner has replaced the old Keyword Tool. I prefer not to use the planner tool as an indicator of actual search demand — the numbers just vary too greatly from prior to the change to warrant any credibility beyond a relative measure.

Google Trends has also been a decent relative measure of insight into keyword demand and includes a “related searches” and “rising searches” section. The related and rising searches may be more useful to monitor than ever before, especially given recent changes related to Hummingbird.

Today, as a result of the Hummingbird change, keyword research that considers auto-suggest results from tools like ubersuggest.com could be more valuable in your overall content strategy. Also, keyword research that includes semantically relevant keywords from free tools like LSIKeywords.com can potentially help search engines better apply more relevancy to your content, especially for content that is optimized for long-tail queries.

Beyond Keyword Research

Beyond changes to content strategy, the in-the-trenches observations also suggest that in some cases SEO website architecture and site design adjustments (such as a shift away from excessive deeply nested subcategories and deep hierarchical navigation in favor of smaller taxonomies and shorter paths to deep content) could be another strategic consideration relating to internal site architectural optimization considerations. Here’s why.

Although there are some exceptions to the rule, Google typically does not like to land users on search results pages from its own search result page because it causes the searcher to have to do two searches. The same idea would be true if Google moves to turn the SERP landscape into more of a topic page for each keyword.

If you have important (revenue generating) content on topic-level pages that you do not have on deeper detail-type pages, anecdotal evidence and data suggest that Google similarly does not want to send searchers directly from one topic page (the Google SERP) to your topic page, which is kind of like sending a searcher back to a search results page. As a result, this might jeopardize any value that you gain from your topic pages.

Do you believe Hummingbird marks a significant change in the world of SEO? Do you have any data or in-the-trenches observations you would like to share? If so, please let me know in the comments or on Twitter.

Image courtesy of Warren Lee.