Stay marketing-savvy and tech-savvy. Get the latest in martech by subscribing to MarTech Today.
What the heck is machine learning, and why should I care?
Understanding the impact of machine learning will be crucial to adjusting our search marketing strategies -- but probably not in the way you think. Columnist Dave Davies explains.
There are many uses for machine learning and AI in the world around us, but today I’m going to talk about search. So, assuming you’re a business owner with a website or an SEO, the big question you’re probably asking is: what is machine learning and how will it impact my rankings?
The problem with this question is that it relies on a couple of assumptions that may or may not be correct: First, that machine learning is something you can optimize for, and second, that there will be rankings in any traditional sense.
So before we get to work trying to understand machine learning and its impact on search, let’s stop and ask ourselves the real question that needs to be answered:
What is Google trying to accomplish?
It is by answering this one seemingly simple question that we gain our greatest insights into what the future holds and why machine learning is part of it. And the answer to this question is also quite simple. It’s the same as what you and I both do every day: try to earn more money.
This, and this alone, is the objective — and with shareholders, it is a responsibility. So, while it may not be the feel-good answer you were hoping for, it is accurate. With this in mind, we can now quickly ponder what Google needs to accomplish this.
There are a variety of ways Google can increase its revenue. Here are some of the more obvious:
- Increase their users
- Increase the number of times each user returns
- Increase the revenue generated per user
- Reduce the need for users to leave their sites to complete an action
- Increase the number of ways a user can be reached
So, Google needs to be present in as many places as possible; they need users to rely on them consistently and frequently; they need to hold their users in their sphere of influence so as to increase their ability to advertise to them; and they need to find ways to increase their revenue from the users they have performing the tasks they’re already doing.
So, what about machine learning?
Thanks for sticking with me through that, but it’s important to fully understand what Google’s trying to accomplish to get the full scope of how they might apply machine learning toward achieving this goal (both currently and in the future). Knowing that, let’s now consider what machine learning is, restricting our discussion to how it impacts Google search.
By definition, machine learning is simply a form of AI that gives computers the ability to learn and adapt based on incoming data or signals. That is, the ability to develop knowledge, information and skills it was not specifically taught by its programmer.
In the past, the engineers at Google would sit down and hammer out ways of calculating a website ranking algorithm manually and push it live, making adjustments to that formula as strengths and weaknesses were discovered (or as pesky SEOs figured out ways to game it).
Machine learning (big picture here) significantly changes this. Imagine a scenario where we have a formula:
(Links * weighting) + (keyword relevancy * weighting) + (time on site * weighting) = rank
In this example, we have a ranking based on:
- the number of links multiplied by a weighting factor
- the keyword relevancy multiplied by a weighting factor
- the time on site multiplied by a weighting factor
Added together, these produce a ranking value which, in the search engine results pages (SERPs), would be a simple result set listed in descending order. Typically, engineers would manually adjust the weighting factors of each of the signals (plus many hundreds more), and an update was pushed out. One day, links would weigh heavier; other days, it would be keyword relevancy in the content. And so on.
Making changes and testing in this environment is extremely taxing, but it’s the way it’s mainly done, even today.
But let’s imagine…
Let’s imagine for a moment that we gave a computer control with our starting point and a goal. Let’s imagine we gave it a simple goal: for any search query, a result will be considered good when fewer users click on additional results for the same query. Seems like a logical and fairly straightforward goal, and one a computer could work with easily.
From there, you would hand over the formula and allow the computer to adjust the weighting until it finds the perfect values to maximize the goal.
This example is the most basic form of machine learning, so let’s take that one step further. What if there were no clicks measured for a given search query? That leaves us with one of two possibilities:
- None of the results appealed to the user, or
- The answer was given on the results page itself
The first option indicates either the query was inaccurate or the results poor; the second option indicates that the results successfully helped the user with their query. Obviously, there’s a rabbit hole I’m about to take us down if we want to consider every permutation in just this very simple scenario with a brutally rudimentary formula. In the end, however, we cannot possibly program every success metric for every query, which is where the allure of machine learning really kicks in.
Imagine now if we program in the various success metrics across all types of queries and send the machine off to determine which metrics match which scenarios.
Now, let’s take that another step. What if the machine were given that ability to add a new ranking factor entirely and adjust the weight for it? How about if it could create its own success metrics with only the big picture goal of “Google needs to make more money” as the driving criterion? Now it gets more interesting — and this is where the impact on you, your website and your ability to market really comes in.
So, where are we with machine learning?
You may be asking where we are with machine learning and why it matters. Let’s put this into the context I do. I remember riding the bus to school one day in 1991. On the radio was a story about the development of the 486 computer processor.
When I heard the announcement, I thought to myself, “Why would anyone need a 486? The 386 with its whopping 64k (that’s right … K) of RAM does everything you could want!”
Of course, I can now see that computing was still in its infancy back then. And that’s where we are now with machine learning. We’re just beginning to imagine the things that may one day be possible, like me at my first computer thinking toward the future of PCs. To put that into context, my first computer looked like this:
So… what do you do?
The answer here is easy: you do what makes Google money. That’s not to say you have to invest in AdWords, but you have to adapt to the new world order.
The top 10 is shrinking, and with voice search, Google Home and mobile surpassing desktop, I daresay the days of organic search as we know it are coming to an end. That’s not to say organic SEO is done — simply that the idea of striving to rank in the top 10 organic search results will become outdated.
We are heading to a world where machine learning will facilitate such rapid adjustments to algorithms and such customization of individual results and understanding of context that the only requirement of a result is that it meet the user’s needs. Not that it’s organic, not that it has links — just that it meets a need.
And that is what we need to do. We cannot keep up with the algorithms at this point — the only thing we can keep up with is the technology and understanding how users interact with it (with the understanding that Google needs to make money and that there are only so many ways that can be accomplished).
You can’t win by trying to game an algorithm that’s increasingly based on machine learning, but what you can do is understand the goal and build your internet presence toward that.
Some opinions expressed in this article may be those of a guest author and not necessarily Marketing Land. Staff authors are listed here.