G5 BLOG

We'll keep you in the loop. We'll keep you ahead.

 

Return To Blog

G5 Chalk Talk - Machine Learning

Tackle the mysterious subject of machine learning with our VP of Digital Performance, Dave Beltramini.

Including AdWords, GCN, RankBrain, Advertising, Social and Websites. 

 
Learn more about RankBrain.

 

 

Video Transcription

Good morning everybody. I’m Dave Beltramini the Vice President of Digital Performance at G5 and welcome to another session of G5 Chalk Talk.

Today we're going to focus on machine learning; where it's been, where it’s going and one of the biggest proponents of machine learning in the world. So machine learning, as a quick definition, is a type of artificial intelligence that is about computers learning, not actually being programmed. They learn from each other. They learn from these gigantic datasets and that feeds in and improves these algorithms.

As soon as you talk about search engine marketing and algorithms, the first company that pops to mind is Google, of course, right because really everyone goes “Oh, the algorithm change” and they have all sorts of algorithms. In fact, the organic search algorithm is an algorithm of algorithms. So, they have been using machine learning for years.

What are the 6 examples of machine learning?

1. Adwords

One of the first examples is AdWords, ok? So that's not their organic but that is their paid advertising product. They have been using it in what is called their quality score and their ad rank score. It takes hundreds of signals and processes them in milliseconds, and this is what gives the definition or the differences in the ads that you see.

2. Google Display Network

The next one you see is another form of advertising that’s called the Google Display Network. Recently, Google has been using these responsive ads. Ads that adapt to the size of your mobile device, the size of your desktop, or your tablet, and over the course of running campaigns, if you use responsive ads, Google selects the ad that best fits what people are looking for in the winning ads and it does this with machine learning over time.

3. Rank Brain

Another example, and maybe the most famous example, one that came out right away, is Rank Brain. So, what is Rank Brain? Rank Brain is Google's machine learning in their organic search. This has been something that's been going on for almost a year half now. They announced it about a year ago. It's an important set of signals where Google used to use these genius phd's and stats and algorithms to figure out what their search ranking algorithm would look like. Now, they augment those phd’s. They still use it, but they have this set of algorithms that they've unleashed that learns from all the things.

And mostly Rank Brain is about intent, right? They’re trying to figure out the intent of all these searches. If you were going to say something like polish furniture, right? P-o-l-i-s-h— do you want to polish the furniture or do some weird fetish with polish furniture? So Rank Brain can decipher really what your intent is on that. And to be clear, it doesn't run the whole algorithm. It is actually a factor. Which is kind of unusual that it would be a factor so, it is a factor in the ranking algorithms what the machine learning can do for those algorithms.

It's been in place, like I said, publicly for a year but they've been doing it for a while. You'll start to see more fluctuations and search engine result pages that aren't the cause of some major update. It's just part of the everyday changes that Google does. So, Rank Brain is driving a lot of those minute changes because that's really what machine learning output is. Right? It's just constant change based on the data set, and if you think about it, no company in the world has the computing power that Google does. Nobody has the data that Google has. So, their machine learning is going to be the most robust and may be able to make changes everyday.

Now, for you our customers you know because you don't have as many visits machine learning is a direction that G5 is headed but won't be anything for a while because there's not as much data for each individual customer. But we use all these machine learning as a Google Partner. I was just at their Partner Summit and they have plans to share their machine learning with their Google Partners. So, this is something that we're looking forward to implementing across all three product lines that Google drives machine learning.

4. Advertising

When I say that, right, over here in Advertising, how are we using that? What's a good example of that? Well, you know, there was an example of a billboard actually, that changes over in England that they're using machine learning based on viewing how people interact with a billboard of the train station and it changes the actual ad based on what they've learned. I talked about responsive display ads and those are things that we can use to change our advertising based on machine learning.

Google touts that they have seven products that have over a billion users. Think about that—that's YouTube, it’s AdWords. That's their Chrome. That's Gmail. And these are all places where they make signals or show stuff to their customers and they have to process these hundreds of signals in milliseconds. That's where machine learning comes in. That's where they're using it. In their Display Network, in their AdWords, and in their organic results.

5. Social Media

Social media. People aren't using social media to write posts with machine learning just yet. That's not where it’s going. Artificial intelligence has not evolved to that point yet. However, where machine learning has evolved is sentiment analysis. There's an experiment now where machine learning is trying to teach companies actually what TL;DR means. We all know, right, “too long didn't read,” but machines don't know that until they learn that. So really what they're using is trying to get sentiment analysis out of the social media and what people are writing, but the more popular usage of machine learning and social media right now is to analyze your audience. Instead of in the past where advertising agencies would come up with these personas based on qualitative or subjective research, machine learning can take all of those people, all of that data, mash it together in seconds and spit out to you what your audience is really looking like—extra interests, all the different segments. We can do that in a way that gives you not opinion-based marketing but data-based marketing. So that's where social media is going in right now and understanding your audience.

6. Websites

Websites. So, Google and MIT just published a study that they did with Progressive Insurance and what one of the aims for websites are, one of the things that we're driving towards at G5, is a more custom experience for people using websites. So that the website I get based on what they know about me—or people that look like me in terms of demographics, or most importantly where I came from—is different than the website that someone sees if they come from an advertising campaign.

So this is one of the things that we're looking at—is a website, or set of websites, that take this data that provide the best customer experience. In the case of Progressive they logged 15 billion miles in this tool they have for cars and it changed the way they present their websites, and that is something that is probably the future of websites. It's not static. It's not the same. It delivers a different customer experience, it delivers different messaging, different targeting. All in the hopes that it will drive a better conversion.

So, that's where machine learning is going and all those things. So, this subject we could talk about for hours—it's continuously changing, as you would imagine, but that's where we are now, where Google is, and where we are hoping to go with machine learning.

Thank you, see you next time.

Tags