Assessing Hummingbird’s Impact On Search — 10 Months Later

On September 26, 2013, Google let slip that it had updated its overall search platform, which they referred to as “Hummingbird.” In this post, I am going to take a look back at the past year to see what impact Hummingbird may have had.

One of my goals with this post is to help dispel the myth that Hummingbird is an algorithm, and to provide some clarity on the nature of the impact it may have.

Hummingbird was revealed during Google’s 15th birthday party, but no mention of it was included in their own post about the event. However, Search Engine Land Founding Editor Danny Sullivan was in attendance, and this new platform was revealed to him and others who were there.

Very little was understood about what Hummingbird was all about, as supposedly it had been live for a month or so at the time and had impacted 90 percent of queries; yet, very few people had any idea it had happened. How could that be?

I wanted to know — so a few weeks later, on Halloween, Danny Sullivan joined me on the Digital Marketing Excellence Show, and he talked about the makeup of Hummingbird. It’s worth watching this short 4-minute summary video if you have not seen it:

[youtube]https://www.youtube.com/watch?v=iLqVuUqx57k[/youtube]

Bottom line, Hummingbird was a rewrite of the search platform designed to allow Google to process new types of signals in new types of ways. Panda, Penguin, and link signals are all examples of signals that “plug in” to the Hummingbird platform, just as they did Google’s prior search platform; and as Danny explains in the above video, these algos were not changed during the process.

Visualizing Hummingbird From A Developer’s Point Of View

Let me try to clarify what a search platform rewrite might look like. Imagine you created a very useful program. Then every week for 10 years you added a new extension or feature of some sort in a very tactical way. All the new code was probably good code, but all these new pieces kept getting bolted on the side until things got really unruly. In fact your code starts to look like this to you:

What’s happened is that you have lots of code using different naming conventions, common algorithms implemented six places in the code, nested calling structures that are extremely fragile, and much more. On top of that, you want to add new capabilities that allow you to understand entities and relationships far better, and you have this sense of “OMG, I have to modify 72 different places in this code to make this work.”

Then you think about how you might want to tweak the weighting of how links impact different types of scenarios in search (local vs. personalized, vs. videos, vs. basic web pages, and how to merge all those things into one ranking system), and you think, “OMG, I have to touch 43 places in the code.” It’s at that point that your head explodes.

So when you emerge from coder’s shock, you decide to take the bull by the horns and say, “Alright, I have to make this whole thing more flexible — I have to rewrite it. I have to remove the redundant blocks of code. I have to take concepts like ranking signals and modularize those, and concepts like ‘stuff I am trying to rank’ and modularize that, and allow each of those to apply ranking signals to scale the way I want.” (In the meantime, you don’t want to change the current basic ranking concepts at all. We will get to that later.)

Note: The above discussion is meant to explain the concept of how a rewrite of a piece of code that does not change basic functionality typically works, not an actual reflection of what happened with Hummingbird itself. The bottom line is that Hummingbird was designed to enable future changes in Google, and is not the direct source of those changes.

So What’s The Impact?

This is where it gets interesting. How could this have impacted 90 percent of queries? Why was there no buzz in the SEO community about this with that level of impact? Either that 90 percent figure is a misdirect by Google, or the changes were very subtle indeed. It’s this latter scenario that I believe to be the case.

The more important scenario to consider is the practical impact of a significantly more flexible piece of code where it is far easier to implement new types of algorithms. Short translation? If you see some new type of algo come out, or some major change to an existing algorithm, Hummingbird probably enabled it. Can we ever prove that? No, there will not be a clear way to do so. That said, let’s take a look at some significant developments that have taken place since the advent of H-bird.

Panda 4.0

Back on May 20, 2014, Google released Panda 4.0. The 4.0 designation was very interesting because it signifies a major change to the algo in Google’s eyes. In fact, it also suggests some restructuring of the algorithm, again, with the possible intent of making it more scalable and preparing for changes to come.

However, that is just speculation, and we don’t really know. I personally saw sites that previously had been impacted by Panda get a portion of their lost traffic back, and heard about many other sites getting hit by the update. Was this new release enabled in some way by Hummingbird? It’s possible, but my guess is that any impact here was quite small.

Authorship Changes Favoring Mobile

A little more than a month later, on June 25, 2014, Google made major changes to Authorship. Specifically, they stopped showing author photos in search, as well as some other changes. One big reason cited for this was to create a better mobile experience.

Combined with Matt Cutts’ pronouncement at SMX Advanced that Google expected queries from mobile to exceed queries from desktop before the end of 2014, this tells a pretty interesting story: The requirements of the mobile searcher are becoming more important to Google than the requirements of the desktop searcher. Woah. That’s pretty heady stuff to think about.

However, was Hummingbird involved in this? My guess is that it wasn’t in any major way. It’s a big change, but I am not sure how a platform rewrite would have driven it.

Step-by-Step Instructions In Search Results

On June 24, 2014, the day before the change in Authorship, Google started to show step-by-step instructions in the SERPs. Barry Schwartz wrote about it in this excellent post on Search Engine Land.

This is a significant change that I believe Hummingbird enabled. I believe this because it maps out something quite different: the concept of a procedure. To do this, Google has to parse the page and recognize that there is a set of step-by-step instructions. Let’s look a little closer at what they did in our french toast example, starting with a look at the opening of the page from which they extracted the information.

Interesting! Google extracted a little bit of the intro to show with their instructions. Now let’s go find the procedure in the article. Here is the first bit:

Notice how this is not a simple step-by-step list. The actual steps are laid out in heading tags, followed by an image, and then a detailed write-up for each step. After all of that, we then get step two, and, so forth. There is quite a bit of processing done on the content to find and extract this process.

You can review other examples for yourself to see the type of information that Google has extracted from them:

1. treat sunburn – This one is interesting because it does not show a complete process
2. reset iPhone
3. screen capture windows
4. train for a marathon – Also not a complete process
5. make pasta – Another incomplete procedure
6. shine shoes

As you go through these, you will see a variety of challenges that Google would face in extracting the data to build up its panel of results. The incomplete procedures are interesting because Google does not think that there are more steps, and those publishers must be pinching themselves due to the flood of traffic this must be bringing them. Consider the result for “make pasta.”

I have to believe that this is a phenomenal result for howstuffworks.com. Google identifying you as the authority on the topic and then only providing the first couple of steps? Yes, I’d take that.

So, is this Hummingbird? No, it’s not, but I think there is a decent change that Google was only able to implement a new algorithm like this because of Hummingbird.

Summary

That’s the real story of Hummingbird. We just won’t know what changes we see that were enabled by it. None of them will come with a Hummingbird label pasted on the side. What you are really looking for is new types of capabilities emerging. The step-by-step results may well be the first tangible result of that.

What we do know is that Google did it to set the stage for the coming decade, and one thing you can count on: It’s going to be a heck of a wild ride.