Guest Appearance on Digging Deeper With Jason Falls

All right. Enough of me babbling. Christopher Penn is here. He might be uh one of the more recognizable voices in the digital marketing world because he and his pal John Wall are the two you hear on the Marketing Over Coffee Podcast. Uh I think that's in its 14th year.

Chris was also one of the co-founders of Podcamp, way back before podcasting's new wave, which by the way is actually in its second wave, uh major wave anyway. He's also uh known far and wide for being an analytics and data science guru. I've had the pleasure of knowing and working with Chris a number of times over the years, and it's always fun to chat because I come out feeling both overwhelmed with uh with how much more he knows than me, but also a lot smarter for the experience. Chris, good morning. How are you, sir?

Hi, you know, I'm I'm fine. No one I know is uh currently in a hospital or a morgue, so uh it's all good. That's great. So uh I I I want to bring people up to speed on how you got to be uh the analytics ninja you are. We can save the real ninja thing for another uh time.

Because for those of you who don't know, he is an actual ninja. It's not just something I throw out there like he's trained or something. I don't know. But that's not what we're talking about. So you got your start though in the the digital marketing world, I think, in the education space, right?

Give us that backstory. Yeah, very briefly. Education Financial Services. I joined a startup in 2003 where I was the CIO, the CTO, and the guy who cleaned the restroom on Fridays. Uh it was a student loan company, and uh my first foray into digital marketing.

I was I came in as a technologist to run the web server and email server and update the website, update the web server became update the website, you know, fix the email server became send the email newsletter, and over the span of seven years, I basically made the transition into what we now call marketing technology. It was it had no name back then. And part of that was obviously reporting on what we did. You know, if the those who have uh a lot more gray in their hair uh were in the space at the time remember uh a tool called AW Stats where you had to you had to manually pull logs from the server and and render them into terrible looking charts. Uh but all that changed uh in 2005 when a company called Google bought a company called Urchin uh and then rebranded and gave it away as a tool called Google Analytics.

And uh that was the beginning of uh my analytics journey and has been uh pretty much doing that ever since in in some form or fashion because everybody likes to know uh the results of their work. So take me a little bit further back than that though. You you entered this startup in 2003 as you know, technologist, but take me back to like where did you get your love for analytics, data, computers? Because you and I grew up at roughly about the same time, and I didn't really have access to a lot of computer technology until I was at least probably junior high. So there had to have been some moment in your childhood that where you were like, ooh, I like doing that.

What where did it come from? Uh that would be uh when I was seven years old. Uh our family got one of the uh Apple II Plus computers, that horrendous beige and like chocolate brown computer, you know, the super clicky keyboard and the screen screen, uh two colors, black and green. Um, and it was at that point when I realized I really like these things, and more importantly, I could make them do what I wanted them to do. Uh so it's all about control, right?

It really is. Uh, you know, I I was a small kid in school. I got picked on a lot, but I found that information gave me control over myself and more importantly, gave me control over other people. When I was in seventh grade, our school got an Apple IIGS in the computer lab, one of many. And the school's database was actually stored on one of those little three and a half inch floppies.

So I at recess one day I went into the lab, made a copy of it, took it home because I had the same computer at home. And uh that was a complete record of 300 students their grades, their social security number, their medical history, everything because nobody thought of cybersecurity back then. They're like who in the hell would want this information to begin with? Well it turns out a curious seventh grader and just be able to to understand that this is what a database is this is what it does. These are all the the the threads I call them that that make up the tapestry of your life.

You see them very early on they just keep showing up over and over again. You know whenever I talk to younger folks these days and say I don't know what I want to do for my career. Look back at your past there are threads that are common throughout your history if you find them if you look for them you'll probably get a sense of what it is that you are meant to be doing. So cybersecurity is your fault that we we've learned that um and here so I I I take it you would probably credit maybe your parents for keeping you from taking that data and like stealing everyone's identity and you know being being a a criminal or not right well so again back then it was so new that nobody thought oh how can you misuse this data there really wasn't an application for it right back then there was no internet that was publicly accessible. So it's not like you could contact you know Vladimir, the Russian identity broker and sell them off for seven bucks a piece.

You couldn't do that back then. So it was more just a curiosity. Now, you know, kids growing up today are like in a much different world than we were, where that information is even more readily available, but it also has much greater consequences. All right. I'm going to jump over to the comments uh already because uh our our friend Hoseline Main has jumped in with a good one uh right off the bat.

Uh what are but what uh is is analytic, what's what are people's biggest analytics mistakes, Google Analytics or other? What should everyone do uh to set up at a minimum analytics-wise is Google Analytics where you start, or how would you advise someone who doesn't know anything about analytics to set up and what a mistake do people most often make with analytics? The one they most often make is they start data puking. That's something that Avanash Kashik's has a lot, but I love the expression. And it is you get in Google Analytics, there are an accounted 510 different dimensions and metrics you have access to.

For the average business, you're probably going to need five of them, you know. Uh, there's like three to five you should really pay attention to, and they're different per business. So the number one thing that people do wrong, and that is the starting point. And I was talking with uh my partner and co-founder Katie Robert about this yesterday. Take a sheet of paper, right?

You don't need anything fancy. What are the business goals and and measures you care about? And you start writing them from the bottom of the operations funnel to the top, and then you ask yourself, well, checkbox, can I measure this in Google Analytics? Yes or no? So, like for a B2B company, sales, can I measure that in analytics?

No, you can't. Um, can I measure uh opportunities and deals? Probably not. No, can I measure leads? Yes.

Okay, great. That's where your Google Analytics journey starts because the first thing you can measure is what goes in Google Analytics, and then you you know, you fill in the the blanks for the for the rest. If you do that, then it brings incredible clarity to this is what is actually important. This is what you should be measuring, as opposed to here's just a bunch of data. When you look at the average dashboard that like you know, every marketing and and PR and ad agency puts together, they throw a bunch of crap on there.

It's like, oh, here's all these things and impressions and hits and engagements. Like, yeah, but what does that have to do with like something that I can take to the bank or get close to taking to the bank? If you focus on the the your operations funnel and figure out where do I map this to, then your dashboards have a lot more meaning. And by the way, it's a heck of a lot easier to explain it to a stakeholder when you say you generated 40% more leads this month rather than you had 500 new impressions and 48 new followers on Twitter and 15% engagement. And they're like, uh what does that mean?

But they they they go, I know what leads are. Yep, that's true. Uh uh, and just to clarify, folks, to translate here, uh, probably the smartest man in the world just gave you uh advice that I always give people, which is keep it simple, stupid. Like, yeah, drill it down. And I say keep it simple, stupid, so that I understand it.

That's that's my goal in saying that phrase. Uh, but if you boil it down to the three or four things that matter, well, that's what matters. Yep. Now, if you want to get fancy, um, here we go. Exactly.

If you want to get fancy, you don't have to necessarily do that. Uh, there are tools and software that will take all the data you have, assuming that it's in an orderly format, and run that analysis for you. Because sometimes you'll get um I hate the word because it's so overused, but you will that it does actually there are synergies within your data. There are things that combined together have a greater impact than individually apart. The the example I usually give is like if you take uh your email open rates and your social media engagement rates, you may find that those things together may generate a better lead generation rate than either one alone.

You can't see that you and I cannot see that in the data. It's just you know that much data and that much math it's not that something our brands can do. But software can do that particularly uh there's one package I love using called um IBM uh Watson Studio and in there there's a tool called auto AI and you give it your data and it does its best to basically build you a model saying this is what we think are the things that go together best it's kind of like you know cooking ingredients like it automatically figures out what combination of ingredients go best together. And when you have that then suddenly your dashboards start to really change because you're like okay we need to keep an eye on these even though this may not be an intuitive number because it's a major participant in the number you do care about. Very nice.

One of the uh the many awesome things about uh that the marketing world not just me but the marketing world loves about you is how willing you are to answer people's questions. In fact that's basically your blog now your whole series of you ask I answer is almost all of what you post these days but it's really simple uh to do that when you have an area of expertise right people ask you questions your answers are great blog content. Has anyone ever stumped you oh yeah people have stumped me all the time um people stump me because the they have questions there where there isn't nearly enough data to answer the question well or there's a problem that is challenging enough they're like you know what I don't actually know how to solve that particular problem or it's an area where there's so much specialization that I don't know enough. So uh one area that uh for example I know not nearly enough about is the intricacies of Facebook advertising, right? There are so many tips and tricks.

Uh I was chatting with my friend uh Anne Popolizio, who runs uh Social Squib, which is a Facebook ads agency, and I was saying to her, like, uh you know I'm running this campaign, I'm just not seeing the results. I'm like, Can you take a look at it? Because we we barter back and forth every now and again. You know, I'll help her with like tag management analytics, and she'll help me with Facebook. As she look opens the campaign, looks at it, goes, Well, that's wrong, that's wrong, and that's wrong.

Fix these things, turn this up, turn that off. And I'm like, two minutes later, the campaign is running. The next day later, it it has a uh some of the best results I've ever gotten on Facebook. I did not know that. I was completely stumped by the software itself.

But the really smart people in business and in the world have a uh a guild, uh, uh uh an advisory council, uh a close knit uh group of friends, something with different expertise, so that every time you need like I need somebody who is creative, I'll go to this person. I need somebody who knows Facebook as I'll go to this person. If you don't have that, make that one of the things you do this year, particularly now, this time of year where you're sitting at home in a pandemic, uh hopefully you're if you're wearing a mask when you're not, um, and you have the opportunity to network with and reach out to people that you might not have access to otherwise, right? Because everyone used to be like in conference rooms and you know, meetings all day long, and now we're all just kind of hanging out on video chat going, I don't know what I'm gonna do with my day. And that's a great opportunity to network and and get to know people in a way that is much lower pressure, especially for people who are you know are crunched on time, they can fit 15 minutes in for a Zoom call.

You might be able to build a relationship that really is mutually beneficial. The biggest takeaway from this show today, folks, will be Chris Penn gets stumped. Okay, I don't feel so bad. So that's that's that's good. If you're not stumped, you're doing it wrong.

That's a good point. If you're not stumped, you're not learning. Uh I am stumped every single day. I was working at a piece of client uh code just before we signed on here. And I'm going, I don't know what the hell is wrong with this thing, but there's something erroring out, you know, like in line 700 of the code.

I gotta go fix that later. But it's good. It's good because it tells me that I am not bored and that I'm not stagnated. If you are not stumped, you are stagnated and your career is in trouble. There you go.

So uh you are the person that I typically uh to turn to to ask measurement and analytics questions. So you're you're my guild council member of that. Um, and uh so I want to turn uh around a scenario, something that I would probably lob at you uh for other people as a hypothetical here, just so that they can sort of apply here's here's you know what Chris Penn thinks about this, or this is a way that he would approach this problem. And I don't know that you've ever solved this problem, but I'm gonna throw it out there anyway, and and try to stump you maybe a little bit here on the show. So um uh uh on on this show we try to zero in on creativity, but advertising creative, whether campaigns or individual elements are kind of vague or at least speculative in terms of judging which creative is let's say more impactful or more successful.

And the reason I say that is you have images, you have videos, you have graphics, you have copy, uh a lot of different factors go into it, but you also have distribution, placement, targeting, all these other factors that are outside of the creative itself that affect performance. So so much goes into a campaign campaign being successful, I think it's hard to judge the creative itself. So if I were to challenge you to help Cornet or uh any other agency or any other marketer out there that has creative content, uh images, videos, graphics, copy, whatever. Um so put some analytics or data in place to maybe compare and contrast creative, not execution, just the creative. Where would you start with that?

You can't just decouple it because it literally all is all the same thing. If you think back to Bob Stone's 1968 direct marketing framework, right? List offer creative in that order, the things that matter. Do you have the right list? Is or in our modern terms, the right audience?

Do you have the right offer that appeals to that audience, right? If we have a bourbon bourbon crowd, right, a bourbon audience, and my offer is for light beer, that's not going to go real well. Um well, depending on the light beer, I guess. But if it's you know it's something that uh we allowed a swear in this show or no? Sure.

The 1976 Monty Python joke, American beer is like sex in a canoe, it's fucking close to water. Um you have that compared to the list, and you you know that it's gonna be a mismatch, right? So those two things are important. And then the creative, the question is what are the attributes that you have? Is it what's the type?

What is what's in it when it comes to imagery, the things like colors and shapes and stuff. And you're gonna build out essentially a really big table of all this information, flight dates, day of week, hour of day. And then you have at the right-hand side the response column, which is like the performance. Again, the same process you use with Google Analytics, you would use with this, assuming you can get all the data, you stick it into a machine like you know, IBM Watson Studio, and say, You tell me what combination of variables leads to the response, the highest level of response. And you're gonna need a lot of data to do this.

The machines will do that, and then we'll spit back the answer. And then you have to look at it and proof it and make sure that it it came out with something intelligible. But once you do, you'll see which attributes from the creative side actually matter. Was it an animation? Did it feature a person?

What color scheme was it? Again, there's all this metadata that goes with every creative that you have to essentially tease out and put into this analysis. But that's how you would start to pick away at that. And then once you have that, essentially it's a regression analysis. So you have a correlation, it is then time to test it because you cannot say for sure that is the thing until you then say, okay, if it says ads that are that are read in tone and feature two people drinking, uh seem to have the highest combination of variables.

So now you create a test group of just you know ads of two people drinking, and you see, does that outperform you know an ads of a picture of a plant and you know two dogs and a cat and a chicken, um, and see is that actually the case. And if you do that and you prove, you know, with its statistical significance, yep, two an ad with two people drinking is the thing. Now you have evidence that you've done this. It's the scientific method. It's the same thing that we've been doing as for humans as humans for millennia.

It's just now we have machines to assist us with a lot of the data crunching. Okay, so um when you're you're narrowing in on statistical significance to say, okay, this type of ad works better. Um, and this is a mistake I think a lot of people make is they'll do, you know, some light testing, some A B split testing, if you will, and then they'll say, okay, this one performs better. Let's put all of our eggs in that basket. I wonder where your breaking point is for statistical significance, because if I've got, let's say, five different types of creative, and I do as many A B tests as I need to do to figure out which one performs better.

I've always been of the opinion you don't necessarily put all your eggs in one basket because just because this performs better than this doesn't mean that this is irrelevant and doesn't mean that this is ineffective. It just means this one performs better, and maybe this one performs better with other subgroups or whatever. So, what's your stigness statistical significance tipping point to say all eggs go in one basket versus not? Well, you raise a good a good point. That's something that our friend and colleague Tom Webster over at Edison Research says, which is if you do an A B split test and you get a 6040 test, right?

You run into what he calls the optimization trap, where you continually optimize for smaller and smaller audiences until you make one person deliriously happy and everyone else hates you. Um when in reality, version A goes to 60% of your audience and version B goes to 40% of your audience. And if you throw away version B, you're essentially pissing off 40% of your audience, right? You're saying that group of people doesn't matter. And you know, one of the things Tom says is, would you willingly throw away 40% of your revenue?

Probably not. Um in terms of like A B statistical testing, I mean there's any number of ways you can do that. And the most common is like P values, you know, testing P values to see like is the P value below a 0.05 or or below. But it's no longer a choice you necessarily need to make depending on how sophisticated your marketing technology is. If you have the ability to segment your audience to two, three, four, five pieces and deliver content that's appropriate for each of those audiences, then why throw them away?

Give the audience in each segment what it is they want, and you'll make them much happier. Malcolm Gladwell had a great piece on this uh back in, I think it was the tipping point that when he was talking about uh coffee. Like, like you know, he he and this is in his TED talk too, which you can watch on YouTube. Is he said if you know if you ask people uh what they want for coffee, everyone says dark, rich, hardy roast. But he said about 30% of people want milky weak coffee.

And if you make a coffee for them, the satisfaction scores go through the roof, and people are deliriously happy, even though they're saying the opposite of what they actually want. Um, so in uh in this testing scenario, why make them drink coffee that they actually wouldn't want? Why not give them the option if it's a large enough as the audience? And that is a constraint on manpower and resources. Now, you talked uh about Tom Webster who is at Edison Research and and does a lot of polling and surveying as a part of what he does.

I know you have a tendency to deal more with the ones and the zeros versus the you know the human being element of whatnot. But I want to get your perspective on this. I got in a really heated argument one time with a CEO, which I know not smart on my part, but um uh but about the efficiency in sample sizes, especially for human surveys and focus groups. He was throwing research at me that was done with like less than 50 people, like a survey of less than 50 people. I've never been comfortable with anything less than probably 200 or so to account for any number of factors, including diversity of all sorts, randomness, and so on.

If you're looking at a data set of survey data, uh, which I know you typically look at you know millions and millions of wines of data at a time, so we're not talking about that kind of volume. But if you were designing a survey or a data set for someone, what's too little of a sample size for you to think, okay, this is this is gonna be relevant? It depends. It depends on the population size you're serving. So if you're serving if you got a survey of 50 people, right, and you're serving the top 50 CMOs, guess what?

You need only 50 people, right? Um, you don't really need a whole lot more than the day because you've literally got 100% of the data, the top 50 CMOs. Um there are actual calculators online, you'll find them all over the place called sample size calculators. And it is always dependent on the population size and how well the population is mixed. Again, referring to our friend Tom, he likes to he talks about soup, right?

If you have a uh a tomato soup and it's stirred well, you only need a spoonful to test the entire pot of soup, right? On the other hand, if you have a gumbo, there's a lot of lumpy stuff in there. Um, and one spoonful may not tell you everything you need to know about that gumbo, right? Like, oh look, there's a shrimp. This whole thing made of shrimp.

Oh, nope, though, you know. Um, a lot goes into the data analysis of how much of a sample do you need to reach the population size in a representative way where you're likely to hit on uh all the different factors. That's why when you see national surveys like of the United States, you can get away with like 1,500 people or 2,000 people to represent 330 million as long as they're randomized and sampled properly. When you're talking about you know 400 people or 500 people, you're gonna need like close to 50% of the audience because there are there's enough chance that this is gonna be that one crazy person that's gonna throw the whole thing over that one crazy person is the CEO of a Fortune 50 company, right? And you want to know that.

Um the worst mistakes, though, are the ones where you're sampling something that is biased and you make a claim that it's not biased. So uh there are any number of companies, HubSpot used to be especially guilty of this back in the day. Um they would just run a survey to their email list and say, This represents the view of all marketists. Nope, that represents the people who like you. Um, and there's a whole bunch of people who don't uh like you and don't aren't on your mailing list and won't respond to your survey.

And even in cases like that, if you send out a survey to your mailing list, the people who really like you are probably going to be the ones to respond. So that's even a subset of your own audience that is not representative, even of your audience because there's a self-selection bias. Market research and surveying, and this is something that Tom says all the time, is a different discipline. It is different than data analytics because it uses numbers and math, but in a very different way. It's kind of like the difference between you know prose and poetry.

Yes, they both use words and letters, but they use them in a very different way. And you know, one is not a substitute for the other. Right. Wow. I love the analogy.

And uh Chad Holzinger says he loves the soup analogy, which gives me the opportunity to tell people my definition of soup, uh, which I think is important for everybody to understand. I I've never liked any kind of soup because soup to me is hot water with chunks of shit in it. So there you go. Um checking in a couple other people, Chip Griffin uh back at the beginning said this is gonna be good. Hello, Chip, good to see you.

Chip had a really great look for Chip on the Facebooks. He had a really great uh live stream yesterday that I caught just a few seconds of, and I still want to go back and watch. Uh for all of you folks in the agency world about how to price your services. Uh, and and so I was like, Oh man, I really need to watch this, but I gotta go to this call, so I'm gonna go back and watch that chip. Thanks for chiming in here.

Uh Anya Rosina is here today. She's with three stream, re-stream yo. There you go. So Jason.online slash restream for that. Uh Kathy Colliver's here again.

Hello, Kathy. Good to see you again. Uh Peter Cook is here as well. Peter Cook is uh our director of interactive at uh Cornet. So good to see him uh chiming in and supporting the franchise.

Um, okay, uh Chris, back to my hypotheticals. Uh similar select scenario, but not as complicated. I don't think you've got a friend who owns a business. Size is kind of irrelevant here, uh, because I think this applies no matter what. They want to invest in influencer marketing, which as you know is one of my favorite topics because I get the book I'm working on.

What advice would you give your friend to make sure they design a program to know what they're getting out of their influencers so they can understand which influencers are effective, efficient, which ones aren't, and or is influencer marketing good for them or not. So it's a really there's a bunch of questions to unpack in there. First of all, what's the goal of the program, right? Is is if you look at the customer journey, where is this program going to fit? And it may fit in multiple places, but you'll need different types of influencers for different parts of the customer journey.

There's three very broad categories of influencers. I wrote about this in a book back in 2016 uh which is out of print now and I have to rewrite it at some point. But there's there's essentially the uh again this has uh there's the sort of the expert there's the the mayor and then there's the the loudmouth right uh most of the time when people talk about influencers they're thinking of the loudmouth the Kardashians of the world like how can I get you know eight million views on on my you know perfume or unlicensed pharmaceutical um but there's this whole group in the middle called these mayors these are the folks that B2B folks really care about. These are the folks that like hey Jason do you know somebody at HP that I could talk to to introduce my brand right I don't need an audience of eight million. I need you to connect me with the VP at of marketing at HP so that I can hopefully win a contract.

That's a really important influencer and it's one you don't see a lot because there's not a lot of very big splash there's no sexiness to it's just so yeah, let me send an email and and it'll connect you and they owe eight an eight million dollar deal later, like holy crap, you know. Um I I owe Jason a case of bourbon. And then it's gonna be three or four cases of bourray. And then there's then there's the the expert, right? Which is kind of what you're doing here, which is there are some people uh again for those folks who have a lot of gray hair, they remember back in the in the 70s and 80s, there's this whole ad series, you know, that when EF Hutton talks, everyone listens, right?

Uh the bank the the advisory firm. And it's kind of the same thing. There are folks who don't necessarily have huge audiences, but they have the right audience. You know, I I I hold up like my friend Tom Webster is one of those. Like when he says something, when he writes something, I'm gonna go read it.

I don't need I don't even need to think like, do I have time to read this? Nope, I just gotta go and read what he has to say. Um and so depending on the con the goal of your campaign, you need to figure out which of those three influencers types you want and what your expected outcome is. Second, after that is how you're gonna measure it, what is the the measurement system? If you're doing awareness, you should be benchmarking, certainly giving your influencers you know coded links to track direct traffic, but also you're gonna want to look at branded search and and co uh co-branded search.

So if um you know if I search for you know Jason Falls and Chris Penn, uh, how many times is that searched for in the last month after I do the show? If it's zero, then you know, we didn't really generate any interest in the show. Um, if on the other hand I see that spike up even for a little while, okay, people watched and uh or have heard about this and want to go look for it. So branded organic search sort of at the at the top. If you're not using affiliate links and affiliate type style tracking with your influencers, and your goal is lead generation, you've missed the boat.

You've completely missed the boat. Um and you know, for those uh for those like you know, mayor influencers, that's where you're gonna track that directly in your CRM. Like, hey, Jason referred you know, Patricia to me over HP, and you just track that code it in your CRM, and then later on you can see did that did that deal close? Um or e do we even was she even receptive to like because you can have a terrible sales guy who just sucks. Um it's not your fault as the influencer for referring somebody who then the sales guy completely hosed the deal, but at least you got the at bat.

So for influencer marketing, it's it's knowing the types, having clear measures up front and baking that into the contract. Again, this is something that I've seen brands do horrendously bad at. They'll the influencers very push for activity-based metrics. I'm gonna put up eight Facebook posts and four photos on Instagram. I remember I was doing work for an automotive client uh a couple of years ago, and they engaged this one fashion influencer who said, I'm gonna be uh gonna do uh four of Instagram photos um and and eight tweets, and it's gonna cost you 140,000 for the day, and that was it.

Um and the brand's like, sure, sign us up. I'm like, are you insane? Um, and sure enough, even just doing a complicated regression analysis after the fact, we did an analysis on you know, even just the brand social metrics, and it didn't move the needle at all. The person got great engagement on their account, uh, but you saw absolutely no crossover. And the last part is the deliverables.

What is it you're getting? So the measurements are part of the deliverables, but you have to get the influencer just to put in writing. Here's what I am delivering to you. And it's more than just activity, it's like you're gonna get, for example, in a brand takeover when in an influencer takes over a brand account, you should see you know a minimum of like 200 people cross over because they should have that experience from previous engagements. They they probably know they can get like 500 or a thousand people to cross over, but if they sign the line for 200, they know they'll they'll nail it.

Again, these are all things that you have to negotiate with the influencer and probably their agent, and it's gonna be a tough battle. But if they're asking for money and they're asking for a lot of money, you have every right to say, What am I getting for my money? And if they are not comfortable giving answers, you probably have somebody who's not worth worth the fight. Great advice. So I know uh you do a lot of a lot of the work you're doing now with uh Trust Insights is focused on artificial intelligence, and you've got a great uh e-book, by the way, on AI for marketers, which I'll drop a link to in the show notes uh so people can find that.

How is AI affecting brands and businesses now that maybe we don't even realize? And what are the possibilities for businesses to leverage AI for their marketing success? So AI is there's kind of a joke. Uh AI is only found in PowerPoints. The people who actually practice it call it machine learning, uh, which is is somewhat accurate.

Artificial intelligence is just a way of doing things faster, better, and cheaper, right? That's at the end of the day, it's like spreadsheets. Uh, I often think uh when I hear people talking about AI in these mystical terms, would did you talk about spreadsheets the same way 20 years ago? Like this is gonna be this mystical thing that will fix our business? Probably not.

At the end of the day, AI really is just a bunch of math, right? It's stats, probabilities, and calculus, and linear algebra. Um, and it's all on either classifying or predicting something. That's really all it does at the end of the day, whether it's an image, whether it is uh video, what no matter what? Brands are already using it, even if they don't know they're using it, they're already using it.

Like if you use Google Analytics on a regular basis, you are using artificial intelligence because there's a lot of it built into the back end. If you're using Salesforce or HubSpot or any of these tools, there's already some level of machine learning built in because that's how these companies can can scale their products. Where it gets different is are you gonna try to use the technology above and beyond what the vendor gives you? Um are you gonna do some of these more complicated analyses? Are you gonna try and take the examples we talked about earlier uh from Google Analytics and stuff that into IBM Watson Studio and see if it m its model comes up with something better?

That's the starting point, I think, for a lot of companies is to figure out is there a use case for something that is very repetitive or something that we frankly just don't have the ability to figure out, but a tool does. Can we start there? The caution and the warning is there's a whole bunch. Number one, this is all math. It's not magic.

AI is math, not magic. If you can't do it with regular math, you're not going to be able to do it with AI. Um AI only knows what you give it, right? It's called machine learning for a reason because the machines are learning from the data we give it, which means the same rules that have applied for the last 70 years in computing apply here. Garbage in, garbage out.

And there is a very, very real risk in AI, particularly around any kind of decision-making system, uh, that you are reinforcing existing problems because if you're feeding the existing data in that already has problems, you're going to create more of those same problems because that's what the machine learned how to do. Amazon saw this uh two years ago when they trained an HR screening system to look at resumes and it stopped hiring women immediately. Why? Because you fed it a database of 95% men. Of course, it's going to stop hiring women.

You didn't think about the training data you're sending it. Given what's happening in the world right now and with things like police brutality and with uh systemic racism, everybody has to be asking themselves Am I feeding our systems data that's going to reinforce problems? I was at a conference, the Martech conference, uh, last year, and I saw this vendor that had this predictive customer matching system for, and they were using Dunkin' Donuts as an example. And it brought up this map of the city of Boston, and you know, there are dots all over it. Red dots for ideal customers, black dots for not ideal customers.

And again, for those of you who are older, you probably have heard the term redlining. This is where banks in the 30s would draw lines on a map, red line saying, We're not going to lend anybody in these predominantly black parts of the city. This software brought up Boston and said, here's where all your ideal customers were. And you look at Roxbury, Dorchester, Madapan, Ashmont, all black dots. I'm like, are you fucking kidding me?

You're telling me there's not a single person in these areas that doesn't drink that no one drinks Dunkin' Donuts coffee. You're full of shit. You're completely and totally full of shit. What you have done is you have redlined these these predominantly black areas of the city for marketing purposes. I was at another event um two years ago in in Minneapolis, and I was listening to an insurance company say we are not permitted to discriminate on policy pricing and things like that.

We're not permitted to do that by law. So what we do to get around that is we only market to white sections of the city is effectively what they said. I'm like, I don't believe you just said that out loud. Um I'm never doing business with you. Uh but the danger with all these systems, with AI in particular, is it make helps us, it's like coffee.

It helps us make our mistakes faster. And then bigger. Um and we got to be real, real careful to make sure that we're not reinforcing existing problems as we apply this technologies. Now, when you start small, like, can I figure out you know what gets me better leads in Google Analytics? That's relatively safe.

But the moment you start touching in on uh any kind of data at the individual level, you run some real risks of of reinforcing existing biases, and you don't want to be doing that for any number of reasons. Easy the easiest one is it's illegal. Yeah, that's good. Well, if if people watching or listening uh didn't know why I love Chris Penn before, uh they do now, because holy crap. Uh we could uh it's a master's thesis every time I uh talk to you, and I always learn something great.

Uh thank you so much for spending some time with us this morning. Uh tell people, I've got links to copy and paste, but tell people where they can find you on the interwebs. Uh two places to the easiest to go. Trustinsights.ai is my company and our blog and on all the good stuff there. Uh, we have a weekly podcast there, too, called In Ear Insights.

And then my personal website, ChristopherSPen.com. Easiest uh you find all the stuff there. And you can find your way to all the other channels from those places. But those are the two places to go. Trust insights.ai and Christopherspen.com.

That's great. Chris, thank you so much for taking some time and sharing some knowledge with us today. Uh always great to talk to you, man. You too. Take care, sir.

All right. Christopher Penn. Want help solving your company's data analytics and digital marketing problems? Visit TrustInsights.ai today and let us know how we can help you.