Chris Penn Explains How Dangerously Effective Marketers Can Be When They Understand Some Data Science
Traditionally, good marketers are natural detectives. If a campaign goes down in flames, we ask questions, make educated guesses about what went wrong, course-correct, and try again.
But marketers no longer need to play the guessing game. Data science can help marketers know with absolute certainty that it was the email blast, wielding the bad headline, that killed the campaign.
The question is: how can marketers hone their super-sleuthing skills to not only pinpoint past problems but to ensure the vitality of future campaigns?
To answer this and many other questions about data-driven marketing, Treasure Data recently sat down with Chris Penn, esteemed data detective, and co-founder and chief data scientist at Trust Insights.
Data-Driven Marketing Trends: A Conversation with Chris Penn
Treasure Data: Your education background is fascinating: government and political science in undergrad, information systems in grad school. How did your interests develop?
Chris: Well, undergraduate, to be perfectly honest, was a degree I pursued to get out of college as quickly as possible. I wanted to go to work in the real world.
My interests at the time were actually more IT and technology. I worked part-time all the way through college doing technical support. Technology has been an interest of mine since the second grade.
The graduate degree resulted from a love I developed in college of a particular Japanese martial art. My friend’s teacher, Mark Davis of the Boston Martial Arts Center, was based in Boston. So I wanted to go to Boston to study with them, and needed something to do during the daytime. So I took an Information Systems degree.
Treasure Data: How did your interest in data science lead you to martech and marketing?
Chris: It’s actually the other way around. So in the early 2000s, I started working at the financial services startup, Student Loan Network, and I was the IT director.
A lot of my work was: Update the web server, update the mail server, keep the software patched, and so on. And naturally, at a startup, you wear many, many hats. I used to joke: I’m the CIO, the CTO, and the guy who cleans the restrooms on Friday.
And what happened was marketing became marketing technology during that decade. “Update the web server” became “update the website.” “Update the mail server” became “send this week’s email newsletter.” That’s what brought me into marketing and marketing technology. It’s that logical progression.
And so at that point, around 2005 or 2006, Google Analytics became “a thing.” Google bought a company called Urchin and made the otherwise very expensive software available for free. A whole bunch of us started figuring out: There are actually better tools for measuring our websites’ performance. And, I focused on analytics and data from 2009 until 2012, when I started working in-house at a public relations firm, bringing in marketing technology there.
And that’s when I started making the pivot into data science, machine learning, and AI. Because we had all this data. What do we do with it? How do we use it? How do we get better with it? How do we take client data and turn it into something usable?
That’s essentially how I changed focus from pure marketing technology into data science and machine learning. And five years after that, I started my company, Trust Insights, to take it to the next level.
Treasure Data: What is it about data science that you find fascinating? What passion drives you to explore it?
Chris: Data science is the intersection of three different fields:
- Understanding business and business goals—strategic consulting.
- The academic discipline of data. I call it data detective work—trying to understand and solve mysteries with science. Hence, the name “data science.”
- On the technology side, it’s using statistics, data science, machine learning, and AI tools.
It’s that combination of three different disciplines that makes data science so interesting. As long as you’re minimally competent in each of the three, you can deliver outsized results. You can really find fascinating answers to questions—answers that you wouldn’t be able to get any other way. And, those answers turn into insights, which turn into action plans.
I was just doing some work for one of our marketing clients, looking at their brand terms and trying to understand what terms are most closely related, using a technique called vectorization.
So, if you imagine a page of words as a map with coordinates on it, what are the words that live closest to the words you care about?
There’s a way to do this mathematically, that tells you, “Yup, these are the words that live closest to your brand.” It’s very computationally intensive but it also tells you insights you could not get any other way. If you tried to read 100,000 pages on your brand, you’d go crazy.
So, in terms of passion, it really is that detective work of solving mysteries, getting answers. And the wonderful thing about the field is that it’s constantly growing, constantly changing. It is even faster-paced than marketing. If you think marketing is fast, data science is even faster. There are new techniques, new ideas, new models—every single day.
Like, just a few months ago, the folks at OpenAI, released the 774 million parameter GPT-2 language model. And with it, you can generate incredibly good text, and net new text based on stuff that you trained on.
I was testing the GPT-2 model for a session I did at Content Marketing World. I had loaded all these different tweet chats into this model to fine-tune it, and generated tweet chat questions for new tweet chats. Now I don’t have to write my own tweet chats, I can train this thing and have it build 100,000 questions—more than I’ll ever need.
So that’s the passion part and the fun part solving mysteries: Doing things that you didn’t know were possible until you learn them and deploy them.
Treasure Data: What future application for tech in marketing makes you most excited? Something we’re not fully exploiting yet, but has great potential?
Chris: From a data science and AI perspective, we have pivoted to a different way of doing things. The old way used to be: Create the model, train the model, and deploy the model. It was very complex and is still probably the best choice for certain situations.
But now with things like MelNet, XLNet, GPT, and all these major transformer-based networks, it has become: Get the pre-trained model from a big technology company, fine-tune it, and then deploy it. So you get to answers faster. And I believe that for marketing, that’s really where you’re going to see massive change.
When you look at the results MelNet got with speech generation, it is unbelievable. It’s almost human. It’s not quite right but it’s much more believable than like the monotone that we’ve been used to.
And so, having machines read audiobooks, for example, in the next six to 12 months will be a pleasant experience. We’ll listen and go, “Wow, that was really good, better than the author would have done.”
There’s another startup called Lyrebird where you read and it trains the machine how to speak in your voice. So being able to do that kind of thing is really interesting.
Language generation, speech generation, video generation, image generation—all these technologies are not futuristic, they’re here now. But they’re not being exploited yet because they are still incredibly computationally expensive and technologically difficult to use. But as they become easier to use, we will see their applications in marketing. And, marketers will change in their roles to be less of the doers, and more of the conductors of the orchestra.
So a marketer would work with a data scientist to train an AI on their specific use cases and, on the other side, would be the quality inspector saying things like, “Yeah, okay, out of this list of 100,000 tweet chat questions. That one’s a little weird. So let’s make sure we train it out of asking that particular question.”
And that’s a big career shift for a lot of people—to go from doing the work to being the conductor or manager.
Treasure Data: How can a marketer who wants to become more data-savvy start wrapping their head around all of this?
Chris: Well, becoming data-driven is a mindset. Data and the processing of data is fundamentally mathematics. And a lot of people have been deeply disturbed by the education system, especially in the United States of America, led into believing that they’re not good at math.
Or, even worse, cultural biases can imprison people in their own minds. You know: you’re African-American, you can’t do math. Or: you’re a woman, you can’t do math.
Those are completely wrong. Those are prison sentences for people’s minds. And so the biggest thing that a marketer has to do is free themselves from that mental prison.
Think about mathematics. It’s a language, right? If you can speak English or another language, you can also speak math. In fact, math is easier, because there are far fewer exceptions. So undoing the damage that the education system did is where to start from a mindset and emotional perspective.
All this stuff is accessible and, the heavy, heavy lifting is something that machines do. We have to learn the concepts and understand how the thing works. We don’t actually have to do the thing. We just have to know how it works so that we can quality-check it. Again, it’s that shift in my mindset from do work to manager that is really important.
Really, to become more data-savvy, the most important personal trait is to be curious, to want to know the answer to want to solve that mystery.
When there’s something wrong in the analytics, it’s a mindset change where you go from: Whatever. Just put it in PowerPoint, we’ve got other things to do.To: Why did that happen? Like what happened there? Why is there an anomaly? Why was our website traffic higher yesterday, for no discernible reason? Why didn’t this ad perform?
That curiosity will guide you into being more successful. If you are curious, if you are persistent, if you are driven, if you are motivated by that curiosity, then the sky’s the limit for what you can do.
Treasure Data: What’s the best advice you have for marketers at the start of this journey?
Chris: One of the things you should be studying, and it’s true for all of us, myself included, is yourself. Right? Who are you? What do you want? Why are you here? Where are you going? Those are all really important questions. And the only place those answers come from is, inside yourself.
So the best meta-advice is to figure out who you are. And be really honest about what you’re good at, what you’re bad at, what your weaknesses are, and then try to find ways to mitigate those weaknesses.
It doesn’t matter if you’re on the creative side, on the mathematical side, or on the quantitative side, there are thousands of resources that can help. Pick the top five or 10 and start digging in.
And once you hear a specific topic that interests you, dig into that one topic. Like I really want to learn more about active learning neural networks. I really want to learn more about regression algorithms. I want to learn more about Google Analytics goal conversions and how Google computes them. And that will take you down a rabbit hole of investigation.
There’s a technique from the martial arts I learned from one of my teachers. He said, “We’re going to take a basic tool and force you to use it unconventionally.”
So you’re only allowed to use a certain type of strike. So like you can only use your thumb—can’t use anything else. All the other techniques? All the cool stuff? Put that aside.
If you can only hit somebody with your thumb, you get real clever, real fast. Like how do I adapt my technique and my knowledge in this unusual situation? The same thing is true in marketing.
How do you increase your skill with one tool? Maybe it’s Audacity, the audio editor. Maybe it is Photoshop. Maybe it is a certain type of writing, like poetry. The question is: how do I use this in as many different ways as possible to truly understand it inside and out?
It’s the same in cooking, right? So, you have a frying pan. How many possible different ways can you use a frying pan? Yes, some are going to be sub-optimal uses but to learn the limits of a technique or a tool, you have to study it.
In his book Outliers, Malcolm Gladwell highlighted the idea of 10,000 hours of practice based on a study by Anders Ericsson. As much as people have protested or complained or attempted to debunk the 10,000-hour theory, if you want to master something, it takes practice. There are nuances to it but it’s not a bad idea to say, “Yeah, it’s going to take me two, three, four, five, or 10 years to master a technique.”
So I would encourage you to pick a tool or a topic you want to study deeply and dig into that. You can’t learn all of data science at once. You can’t learn Google Analytics all at once. But you can learn a thing and learn it so well that you become competent, and then masterful. And I think that is a great aspiration for marketers.
Crack Open the Case on Customer Understanding
Marketing is filled with mysteries. How can I increase coupon redemption? Why does Sally Shopper review products online before buying them in the store? How can I improve campaign results? While the answers to these questions may be unknowable with old-school guesswork, data science and martech can provide better insights into customer behavior, along with better marketing outcomes.
Learn how Treasure Data enterprise Customer Data Platform (CDP) can help you crack the case on customer understanding. Request a demo to get started.
Chris Penn is a bestselling author, keynote speaker, and Co-Founder and Chief Data Scientist at Trust Insights. Chris also co-hosts the Marketing Over Coffee podcast. Visit his website for more information or follow him on LinkedIn and Twitter.