Quick Overview

This is an introduction to a free, weekly, video+blog machine-learning series I’m publishing, tailored to SEOs and marketers like you. I’m gearing it specifically toward marketers to address the consistently bad and/or incomplete information published about machine learning as it relates to SEO and marketing…which, each time I see one of those articles, makes me cantankerous and cranky.

Here’s the schedule:

  1. Introduction to Machine Learning for Marketers:
    Cantankerous and Cranky.
    Importance of Machine Learning for Your Future in Marketing.
  2. Search Indexes and On-Page Factors:
    Inverted Indexes vs Relational Databases.
    TF/IDF, Cosine Similarity, and BM25.
  3. Topics and Categorizing Web Pages:
    Classification with Term Vectors.
    Bag of Words and Window Sizes, and Clustering.
  4. Link Value Graphs:
    Understanding PageRank.
    Applications to SEO.
  5. Link Relationship Graphs:
    Nearest Neighbor and Shortest Path.
    Applications to Social Media Marketing.
  6. Modeling the Human Brain:
    Artificial and Convolutional Neural Networks.
    Extra: Putting it All together.
  7. Not Quite AI: The Human Element:
    Normalization, Data Structures, and Training Models.
    Extra: How Penalties, includig “Manual” Penalties, Work.
  8. Understanding the “Big” in Big Data:
    Limitations in the Physical World.
    Data Storage, Parallel Computing (and GPUs), Distributed Computing.
  9. Conclusion:
    You Probably Don’t Need (And Shouldn’t Use) Machine Learning.
    Yet.

Hello, Again. Some Background.

So… I’m back. More on this at the end. But right now, let’s jump in.

Having been in SEO since August of 2001, I have to admit: I sometimes miss the months of Google Dance updates on the last Friday of the month. It was clear, clean, and straightforward: you got rankings or you didn’t. If you didn’t rank, you’d grind to gain. If you did rank, you’d grind to maintain. [nsfw: video: language]

It’s simply no longer the case – and it shouldn’t be – that search engines work this way. As Google moves more and more to a machine learning, moment-by-moment ranking process, things like keyword density, PageRank, link text, content length, link quality…all of it becomes more and more vague when valuing these metrics and how much those tactical efforts will impact your rankings.

Technology’s Absorption into Marketing

After spending an average of 100+ hours a week for over a year and a half going deep into lower-level programming, understanding the dark bowels of the internet and how it works, and dabbling in math and machine learning (and the previous eight years studying psychology and relationships), two things have become increasingly clear to me:

1 – The absolute vast majority of marketers (ie: probably you. Exception: Mike King), including SEOs, don’t understand the fundamentals of machine learning for document classification and ranking (let alone the entire process of writing a solid crawler, parser, normalizer, and indexer) well enough to make educated decisions on how to continue leveraging search engine optimization in an even more rapidly-evolving technical landscape. In other words, the humans aren’t ranking results anymore (that’s partially true, which I’ll explain later in the series); the machines are. Except the machines these days are far more intelligent than we could have ever imagined.

2 – Because of #1, ranking “tactics” must evolve to become de-emphasized (until the moment of execution), and “ranking philosophy” is what will drive the successful ranking campaigns of the future. I’ve preached this for a long time to those close to me. The ones who understand this aren’t cast about by Panda or Penguin, or whatever cute name-du-jour is given to an algorithm that creates mass destruction for those who focused on the tactics over the philosophy. The marketers with the best philosophies, instead of being cast about by the waves, they ride the landscape, staying upright, as the landscape churns and groans below.

You Still Won’t Do Anything Differently

The unfortunate consequence of “marketers marketing to marketers” is that we pump out content and spew exactly the right combination of words to steal as much attention as we can. We take each other’s attention and, like someone who’s slick at a club and horrible in bed, we provide a disproportionately-low amount of value for that thievish, attention-stealing headline. But because humans are animals with their own subset of biological and emotional behaviors, impulses, and motivations, articles about “tactics and strategies” work to get that attention by feeding our scarcity anxiety and desire for the moment of “now” to feel better, at the cost of the “future.”

So, given that I’d guess about 5% of you reading this will really take this idea to heart – that, in the SEO landscape, philosophy wins over tactics and strategies – I’m taking the next eight weeks to walk you through how futile it is for you to continue chasing SEO tactics and strategies.

These tutorials are clear and direct pieces designed to educate you more about how the current and next generation of marketing technology actually works. I am walking you through the technical concepts of machine learning, but from a marketing perspective…because 1, you probably don’t need to understand how to calculate an eigenvector in order to understand how PageRank is calculated, and 2, because continuing to see marketing articles about machine learning, filled with fluff and bullshit, gets me pretty cantankerous and cranky. But I digress.

To put your apathy (and my apathy) another way: we’re humans. And it turns out that most humans are twice as motivated by negative consequences as we are by positive rewards. (Tangentially related: ultimatum negotiations) So, for you to make a real change in how you approach your marketing, I can’t just tell you it’s going to be better and you go out and make the change. You’ll actually need to see from me that your current trajectory of “marketing tactics” will actually be harmful to you if you don’t make a change and make it quickly.

Introducing: Machine Learning for Marketers

For the next eight weeks, I’ll be putting together articles and videos that explain how machine learning works. But, instead of going deep into the math, I’ll instead be spending that time framing it for SEOs and other marketers. I say “other marketers” because some of this will dive into graph algorithms. You’re likely familiar with these sets of algorithms via PageRank, but many of them also apply to Facebook’s algorithms (and almost any other social or link network) and will, therefore, also apply to social media marketing. You’ll quickly see that machine learning is influencing a lot more of the marketing you use, and to a much deeper degree, than you’ve previously realized.

After those eight weeks, I’ll continue to put together weekly videos and posts showing you where tactics and strategies are legitimately useful and applicable. (Hint: it’s after-and-only to support the philosophy, or when creating an initial foundation.)

And, by the end of the year, it’s my hope that you’ll have a much clearer understanding of machine learning so that you can execute far more effective marketing campaigns for 2017, and even for the rest of your career. (Or at least until quantum computing?)

Here’s a rough draft of the weekly schedule of Machine Learning for Marketers:

  1. Introduction to Machine Learning for Marketers:
    Cantankerous and Cranky.
    Importance of Machine Learning for Your Future in Marketing.
  2. Search Indexes and On-Page Factors:
    Inverted Indexes vs Relational Databases.
    TF/IDF, Cosine Similarity, and BM25.
  3. Topics and Categorizing Web Pages:
    Classification with Term Vectors.
    Bag of Words and Window Sizes, and Clustering.
  4. Link Value Graphs:
    Understanding PageRank.
    Applications to SEO.
  5. Link Relationship Graphs:
    Nearest Neighbor and Shortest Path.
    Applications to Social Media Marketing.
  6. Modeling the Human Brain:
    Artificial and Convolutional Neural Networks.
    Extra: Putting it All together.
  7. Not Quite AI: The Human Element:
    Normalization, Data Structures, and Training Models.
    Extra: How Penalties, includig “Manual” Penalties, Work.
  8. Understanding the “Big” in Big Data:
    Limitations in the Physical World.
    Data Storage, Parallel Computing (and GPUs), Distributed Computing.
  9. Conclusion:
    You Probably Don’t Need (And Shouldn’t Use) Machine Learning.
    Yet.

Re: I’m Back

I mentioned above having been in SEO since August of 2001. It was then that I started working at KeywordRanking/WebSourced, and worked with many wonderful folks like Andy Beal, Jenny Halasz, and Garrett French. A few years later, I began consulting in November of 2006. In 2008, I started ontolo and it grew well for some time. But in January of 2007, I had become deeply interested in the inner constellations of the human experience: psychology, philosophy, spirituality, and relationships. And I had to go deeper into that rabbit hole and eventually set ontolo aside for a bit.

In other words, I had to mostly disappear for a few years to figure out the right way to solve the complex and unique problems that ontolo solves today. As a result, some time around 2011 or 2012, I put ontolo on the back burner. It might seem counterintuitive, but by the end of 2013, at the urging of several of my closest female friends, I had spent the entire year leading workshops and coaching, working with women to help them in becoming more rested in their femininity. Approximately 90%+ of you will find this confusing and strange. And that’s understandable.

By September of 2014, I knew it was time to come back to ontolo. My original vision for what I felt it should become, hadn’t yet been realized and wasn’t even close. The problem was that “I’m not a programmer” and the things I wanted to do were highly specialized, challenging, time-consuming, and somewhat neurotic. They’re not the kind of thing you can just go hire someone to do. At least not at a cost that made sense. There were ways of doing it that were locked in my mind that I couldn’t explain, but I knew had to be done. And if it was going to be done, I needed to figure it out myself.

So I set out to learn and do everything I needed to in order to make ontolo what it is today: the fastest, most detailed prospecting tool on the market when it comes to finding very specific prospect types, content, and contact information, applicable to a wide range of marketing strategies beyond SEO and link building. (Also: consulting to solve the hard problems with effectively collecting and organizing your large sets of data.)

Now that this version of ontolo is out and refined a bit (and now in the cycle of minor improvements until the next major release targeted for January, 2017), it’s time to help other people understand better how it works. It’s not difficult to learn, it’s just different from any other tool because of how much it gives to you.

But only teaching folks how to use ontolo…that gets a bit boring from repetition. And I get antsy. And when I get antsy, I get ornery. So I’ve decided I’d also dedicate a fair amount of time to helping you more deeply understand the principles interwoven into the technology that gently, but powerfully, moves the subterranean undercurrents beneath your marketing campaigns.

And that begins now, starting with this series: Machine Learning for Marketers.

Getting Started with Machine Learning for Marketers

Friend, follow, and subscribe to each of the channels below to make sure you get each batch of content as quickly as possible.

See you next week.

Ben Wills