Subscribe
Share
Share
Embed
Hosted by Scott Cohen and Garin Hobbs - Talking about the hot topics in and around email marketing.
Hello all. Welcome to that inbox army podcast. I'm your host, Scott Cohen. With me today, the Daryl Hall to my John Oates is my cohost, Garin Hobbs. Garin, how are you doing today?
Garin Hobbs:Not too bad, Scott. Pleasure to be here as always. Excited for today's presentation.
Scott Cohen:Yeah. Totally. Today, we're talking the topic du jour, AI. Oh, god. Yet another podcast on AI.
Scott Cohen:But we've got we've got the right person to talk about it because it seems like AI really is a panacea for marketers, isn't it? The cure all for our woes. And it could be. It could be. But not without some good thinking and planning behind it.
Scott Cohen:We're gonna dissect AI quite a bit today with our guests, specifically speaking to how data powers AI to ultimately power that elusive thing in email marketing we call personalization. To help us unpack this topic, joining us today is the founder of Campaign Genius, cofounder of Social Signal AI, and the host of the Future of Email podcast, doctor Matthew Dunn. Matthew, welcome to the podcast.
Matthew Dunn:Hey, guys. Great to join you. Thanks for the invite.
Scott Cohen:Absolutely. We you know, if
Garin Hobbs:we're gonna talk about nerd topics, we go out and
Scott Cohen:get the premier nerds. That's what I like to say. So, you know, it's it's what we do. Yeah. But before we start nerding out, I'd love to talk about love to learn about people's journeys, how they ended up where they are now.
Scott Cohen:Tell us about your journey to today.
Matthew Dunn:No. No. I'll keep it as I I'll I'll keep it short. My first job with email in the title was in 1990. Woah.
Matthew Dunn:Yes. Before SMTP email was getting used, before a whole bunch of other stuff back in the days of corporate corporate lands and Novo Networks and stuff like that.
Scott Cohen:We found an OG. We found an OG, Garin. This this this wow. Beat that date, man.
Garin Hobbs:Long live the listserv.
Matthew Dunn:Yeah. Beat that date. Long live the list serve. Yeah. And I used email on 300 baud modem, 3, 4 years earlier than that, but leave that one aside.
Matthew Dunn:Yeah. So jumped I jumped from the arts into technology a long time ago and have found that to be a great combination and haven't really looked back, And I've worked at big corpse and small corpse and started my own, and I've been a serial entrepreneur for something like 15 years and try to be a scholar of the space and understand where things come from and think about where they might go realizing we don't really know. We're just along for the ride.
Scott Cohen:Yeah. We really are. Yeah.
Garin Hobbs:Absolutely. It's a solid provenance. But let's let's dive into it, Matthew. You know, all of us, let's call and imagine the majority of our listeners have been around long enough to know that many things that are now being called AI today have been around for quite a number of years. Right?
Garin Hobbs:Same time optimization, dynamic content, audience selection, etcetera. Hell, even the windshield wipers on my Tundra are being called AI. I feel the raindrop. Let's start on. Right?
Garin Hobbs:But
Scott Cohen:Right. Right.
Garin Hobbs:Certainly, it seems to be a slight stretching of the truth. What's different about today's AI when it comes to email marketing?
Matthew Dunn:I think the threshold we passed, what, a year and a half ago or something like that when AI went from academic initials to cocktail conversation overnight with ChachiPT was that it's was that outputs suddenly felt credible, almost human. Gee, I could have written that. Gee, that was written by a machine. Your windshield wipers, your machine learning, you know, your machine learning results, like amazing stuff. And AI is over 50 years old as a field, over 50 years old.
Matthew Dunn:But it never hit that, never hit that threshold of how in the heck did you do that? And suddenly, they did. That's my that's my reaction. What do you think?
Scott Cohen:Yeah. I mean, it doesn't yeah. It's
Garin Hobbs:It it seems to be more of a label than a strict definition. Right? It's, and people are buying that, I think, more conveniently rather than via, you know, scientific qualification, so
Scott Cohen:to say.
Matthew Dunn:Abs absolutely. And and it's a big enough label where the fact that it could should be differentiated and that there are nuances. That doesn't matter. It doesn't matter. It's like it's like Internet.
Matthew Dunn:There's a whole lot of detail to that word, but it doesn't matter. We just go Internet, digital, and it it and it means something big culturally and economically even and even historically. And this is the second ride. You guys you guys both have a little gray, on the temples there. This is the 2nd Good.
Matthew Dunn:2nd time on the carousel. Right? We were all we all watched digital kinda bust loose and go bananas, and it was, like, it was fun. It was fascinating. It was challenging.
Matthew Dunn:AI feels like it's got the same cultural magnitude or cultural velocity. Maybe it doesn't, but it feels like it does. What do you
Scott Cohen:think? Yeah. I mean, I I was my snarky comment is we we we're we're past the days where people care about Hal and Skynet. Right? I mean, the the seventies, eighties, and nineties were filled with the robots are coming for us if they get too smart.
Matthew Dunn:Yeah.
Scott Cohen:And now we're like, no robots can do my job. Thank you very much. I'm still waiting for the robots to do my laundry, but, you know, it's but I but to your point, that that speed, the the ease of adoption
Matthew Dunn:Yeah.
Scott Cohen:Is is is really where where things things catch hold. Right? When the Internet was new, you still kinda had to be a nerd to know how to use it. Sure. Yeah.
Garin Hobbs:Yeah.
Matthew Dunn:Yeah.
Scott Cohen:Yeah. Once it once you got through the only nerds can use it I'm not saying nerds is a as a bad thing. Right? But I'm, like, we're in this you think about the last 10, 15 years, nerds are good now. Right?
Scott Cohen:Like, nerds used to be something people made fun of, and now nerds are a good thing.
Matthew Dunn:Well, you know, there's a typo in the bible. It's actually the geek shall inherit the earth.
Scott Cohen:I just
Matthew Dunn:didn't didn't believe that. Yeah. You know? And
Scott Cohen:what's the difference between a nerd and a geek? I really somebody explained it to me once, but I don't remember. But but that that ease of access, it's a d 20 and you know it. Or I can get I can do a 20 30 on you. I haven't played d and d too much.
Scott Cohen:But but but but my point is, like, you know, that that it's accessible, that it's that's where things take hold is that accessibility piece of it. Not we talked about accessibility in email last last episode. But, you know, that when people can adopt it easily, that's when it takes hold. I mean, my kids are all younger both younger than the iPhone. They don't know what life was like before the iPhone.
Matthew Dunn:Before the iPhone. But let let me jump in on that because it's really it's really important for context. Right? What, who's quipped Newton's quip, if I've seen further, it's because they've stood on the shoulders of giants. That the same thing happens with innovation, and the previous innovations provide the platform for the new one and change the velocity of the new one.
Matthew Dunn:Right? Internet, what, 10 years to kinda roll out and become almost everyone's got it. Smartphones, what, 3, 4, 5 years, everyone's got one. Tad GPT went to iPhone like numbers in something like a month or 2 months because it literally took you typing 7 or 8 characters in the device in your pocket on the network already in the sky to get to it. Of course, it was fast.
Matthew Dunn:Yeah. Of course, it was fast.
Scott Cohen:Yeah. When you think about the speed of it, I can't remember where Roy read it, but it was something like arc my kids. And, like, by 2050, the rate of innovation that we see today in, like, 10 years will happen, like, every 11 minutes or something. It's it's something absurd. Yeah.
Scott Cohen:And I go, I can't even keep up with this crap. What's gonna happen by 2050? Right? So it's I'll be old and I can be the curmudgeon in the corner. It'll be fine.
Scott Cohen:But, you know, it's it's crazy.
Matthew Dunn:There's a point buried in that too, Scott, though, because it it's partially what it can do, and it's partially our you know, is it useful enough and are enough of us doing it to hit that sort of cultural velocity take up. Right? There were predictions that nanotechnology sorry. Helicopter. I'm dead.
Matthew Dunn:Interesting. There were predictions, what? 10 years ago, we're all in on nanotech. It's the big revolution. Didn't happen.
Matthew Dunn:Right? Doesn't mean there's not useful stuff. But when's the last time you talked about nanotechnology at a cocktail party? Somehow AI hit the cultural velocity thing and has enough utility, I would argue. There's enough utility in the various AI technologies to say, yeah.
Matthew Dunn:I might use it or I might try that or I might actually use it every day because it's making my life easier, things go faster, wash my clothes better, something like that. So this one this one hit.
Scott Cohen:This one
Matthew Dunn:hit, And still not making money.
Scott Cohen:Well, wait. Hold on. We're we're gonna we're gonna make money off of this?
Matthew Dunn:Yeah. But yeah.
Scott Cohen:Yeah. But you mentioned, like like, a lot of capability. Right? A lot of capabilities with it, a lot of general use. I mean, everybody can use chat g p.
Scott Cohen:That general use can really be but there's a lot that needs like, from an email perspective, from a marketer perspective, what do you need to power it correctly, effectively? Right? Like, what are the resources? What's the infrastructure look like? Because people think it's an easy button.
Scott Cohen:But to your point about standing on the shoulders of giants, like Yeah. There's so much work that goes into it. Walk us through that.
Matthew Dunn:I suspect if we could magically canvas 100 or a 1000 people working away in email marketing and say, are you making use of AI technologies, plural, what are you doing with it? We'd find a whole lot of experimentation, a whole lot of cheating, and a whole lot of adaptation of general use tools, your word, for what's actually a really specific, difficult, and targeted discipline. Right? Hey. I got chat GPT writing subject lines.
Matthew Dunn:That may not be the best idea you had. It may look like it's making your life easier, but is it are the, you know, are the results better? Are customers opening more? Do they understand what you said, etcetera? Or you just, I'm I'm great with experimentation.
Matthew Dunn:Like, we should be doing that. And any company that's not budgeting for go break stuff, go try stuff, is gonna get left behind eventually by the companies that are breaking stuff and trying stuff. But do we have tools specific to the discipline built on the that stack of technologies? Not a ton yet, I don't think. And a lot of the ones I've seen in the last year in the email space have been sorta text content centric because that's what the marquee platforms afford.
Matthew Dunn:That's what the marquee platforms make fairly readily available. You know? I've got code sitting right here that calls the chat GPT API and says, take this and do that with it and give me some text back. Why? That's relatively easy to do.
Matthew Dunn:Is that gonna solve my problem? Maybe. Wouldn't it need to be pretty specific for for the domain of of email and marketing, which is so emotional and touch centric. Right? So hard and delicate.
Matthew Dunn:My reaction. What do you guys think?
Garin Hobbs:I you know, I I've said in, some previous articles I've written that, you know, while I think AI is great for saving on some repetitive tasks, is good for sort of basic input, especially when it comes to marketing, the real danger is losing that EQ. Right? When it comes to making decisions, whether they're purchase decisions, service decisions, or any other decisions, you know, 90 plus percent of that is driven by emotion. Right? And, and and the context that feeds that emotion.
Garin Hobbs:And that's the big element that I see really missing from AI is it's under it's a bit current limited ability to understand and take into account context and then the emotional drivers that, are the summation of that context.
Scott Cohen:Because that requires training. The training of the AI, the training of the model to learn. And and I'm not saying it's gonna get there, but I'm saying, like, I think the the issue comes in of the easy button going, oh, it can make my decisions for me, and it already has my brain.
Matthew Dunn:Yeah. And it doesn't. Yeah.
Scott Cohen:And it and it never will. To your point, those those emotional decisions, those I mean, sometimes marketing is gut reaction. Right? I think this is gonna work. Experiment with it.
Scott Cohen:Let's test it. It's just it's just a numbers game. Yeah. At least right now. The way AI is right now is just a numbers game.
Scott Cohen:So but people think, oh, if I just plug it in, it'll learn everything and go, no. It just knows what the inputs are and is making some assumptions kinda like you would Yeah. But not as informed because you don't have your human experience to add on to it.
Matthew Dunn:It it does, but but but there's there's a bit of the promise as well. And I and I'm sure I won't say this well, but I'll fumble my way through it. In some of the work that I've been hands on literally for the last month on with AI, it's really kind of jarringly cool to use it as a mirror that doesn't look like me, that doesn't have my bio. You know, like, I write up something that I think is emotional, compelling, engaging, and I say, okay, Skippy. Reword that for me or reword that for a different audience or rewrite that.
Matthew Dunn:And, like, wow. I wouldn't have thought to put it that way or I wouldn't use that phrase. And it's like we're all captive of our own bio. Right? What we think is compelling, emotional, engaging, etcetera, yes, For some of the audience and no for others.
Matthew Dunn:And it's a little hard to get out of your own frame. And because the, because the outputs look so credible, because, you know, the language is well structured because models can do that kind of stuff, you're like nothing else, it's pretty refreshing to go, oh, that's a good rewrite. I may not use it, but I'm gonna learn something. Mhmm. I'm gonna learn something by taking a look at it, even even just to pick on subject lines.
Matthew Dunn:Right? Even if you say, okay. I think this is gonna be the winner subject line. You know, give me 6 others. Even if you don't use them, it's really it's jarringly cool to have them come back and be worth looking at.
Matthew Dunn:That it's jarringly cool that you can learn something from them even if you don't use them. Yeah.
Scott Cohen:I didn't think about that. Yeah.
Matthew Dunn:Yeah. And and and of all the things I see value in the AI tool says du jour. It's it's the it's the, learning the pedagogical the learning assistance. Right? If you're writing code, if you're writing English, I wouldn't I wouldn't say for for graphic design.
Matthew Dunn:But being able to go, crap. I need a quick request loop to throw this at that API, and I don't feel like writing it for the 50th time. Would you take a crack at there for me, Skippy? And Skippy comes back with really well written, well commented code. I would be too lazy to comment that well.
Matthew Dunn:Cool. Thanks. I'll use that loop. I'll ignore that other one, but that's pretty that's pretty cool stuff, and I can focus on something that's more interesting, more challenging, whatever else. Yeah.
Matthew Dunn:And and there's gotta be thousands of places where this stuff is starting to creep in, and we didn't realize that's how you did that or that's how he did that or she did that. Yeah.
Scott Cohen:Yeah. My wish was when I was working on Salesforce Marketing Cloud, and I wanted them to write me a SQL query, and ChatGPT couldn't write Salesforce's SQL query. Okay. Right. Right.
Scott Cohen:Generic SQL query. So there's that you talk about nuance. Right? Like, that's the part where the the model has to learn and go, oh, you're in Salesforce. Bloop bloop bloop bloop.
Scott Cohen:Here's what you gotta do. Right. Right.
Matthew Dunn:Is it there yet, by the way, on on Salesforce specific SQL syntax? I Or is it not?
Scott Cohen:I it it's been a year and a half since I've tried it, so I don't know.
Matthew Dunn:Because I I do a bunch of stuff with, BigQuery, Google's, data warehouse in the the the it uses, like, old English. Right? It's got it's really, rigid SQL syntax. And I'm like, it's helped me learn it being the various AIs that run-in various windows here. It's helped me learn a ton.
Matthew Dunn:I'm like, how the heck do you do blah blah blah? And sometimes the approach is like, one approach was a complete meltdown. I actually made BigQuery lock up, which takes you through from, like, yeah. Nice try, Skip. But I understand I understand why you tried the coalesce function.
Matthew Dunn:Let's just not do that again, shall we? But it knows the syntax for that particular flavor of of SQL, which is freaking remarkable to
Scott Cohen:me.
Matthew Dunn:Yeah. Like,
Scott Cohen:wow. I imagine just learned quite a bit since I tried it, but I just went I I don't wanna bother my developer. Can I just do a SQL query to chat GPT and do this really fast? And I went, nope. Nope.
Scott Cohen:Not even close.
Garin Hobbs:And then I
Scott Cohen:sent it to sent it to my developer, and he just shake this head. Like Right. Where did you get this? I'm like, I tried chat p t GPT. I'm sorry.
Scott Cohen:Yeah. I'm sorry.
Matthew Dunn:But how long how long till your developer starts using a similar tool to help with something that's re repetitive but still has to be done?
Scott Cohen:Oh, I'm sure. You know? Anything that makes our lives easier. Right? Anything that makes our lives easier that's actually legit.
Scott Cohen:You know, I always I always question the, if you can get you 90% of the way there, does it is that last 10%, is that really where your time is spent anyway? Yeah. If you're talking, like, any sort of code, like, oh, it got me 90% of the way there, but, actually, that stuff's easy. It's the 10% that I gotta do anyway that once it solves that 10%, then you're a business.
Matthew Dunn:Yeah. Yeah. Or
Garin Hobbs:more than you're in trouble.
Scott Cohen:Or you're in trouble. Well, yeah. Yeah. Trouble. It's true.
Matthew Dunn:Yeah. Yeah.
Garin Hobbs:Matthew, when I when I think about the application of AI to marketing, rightly or wrongly, the image that is evoked in my in my in my mind here is that scene from Minority Report. Right? Tom Cruise's character walks in the gap and the, you know, the robotic AI assistant says, mister Anderson, here are the white t shirts we know that you wanted to buy today. Right? Pretty impressive.
Garin Hobbs:And hope maybe we'll get there one day. But, you know, it strikes me that the ability to get there and to deliver that level of personalized experience is very much data driven. And we know that AI, just like many other things that are data driven, is very much a garbage in, garbage out consideration. Right? When it comes to data, it always seems to boil down to collection and usage.
Garin Hobbs:Right? How can AI help facilitate the collection and later analysis of data? And part 2, what are the inherent risks in that given the state of AI today?
Matthew Dunn:Well, the collection side, let's tackle that first. You get a collection use analysis, and then we talk down something. Right? The collection side's either intriguing or scary. Would I be accurate and fair in saying average email program now is built on first party data.
Matthew Dunn:Yeah?
Scott Cohen:For the best
Garin Hobbs:part of first first. Yeah. Exactly.
Matthew Dunn:Okay. So if you wanna do anything other than send everybody exactly the same stuff, you got some degree of how do we how do we tell them apart. Right? How do we group them or what's the term, segment them? Well, this guy told us these things and that guy told us these other things.
Matthew Dunn:Okay. Cool. We can put a dividing line in our dataset and group part. None of us have the tolerance for providing very much of that stuff, either for privacy concerns or for the I am not gonna fill in that many freaking boxes concern. At least, like, sort of voluntary disgorging.
Matthew Dunn:Like, you wanna know my household income? Yeah. I don't think so. Right? Now we may we may sort of tack it on secondarily if we're running the show really, really well and go, oh, well, we'll purchase data and move that over from our platform where they buy things and go, oh, wow.
Matthew Dunn:Scott buys a lot of fill in the blanks. Right? D and d stuff maybe. Cool. And one more about Scott.
Scott Cohen:I talk a big game. That's not true. Yeah.
Matthew Dunn:Is that get is that part gonna get easier? I'm sure I'm sure there'll be algorithmic approaches to say we can infer a lot more a lot more accurately with less effort. Right? Scott buys got bought one D and D thing. Ergo, he probably has this fashion sense.
Matthew Dunn:I don't know. Just make I'm just making up connections. Right? Well, we'll probably get better at at the at the statistical guesswork because of the at least one family of technologies we're calling AI. Creates another problem set that we should talk about downstream, but we're I'm not gonna spend more time filling in boxes.
Matthew Dunn:Neither are you. Right? And if you look at if you look at law as an ex law and laws as an expression of sort of cultural thinking, cultural evolution, We sure have a hell of a lot of privacy laws we didn't have 2 years ago to state by state level in the US. Why? Because we're all going, little time of bummer of a birthmark.
Matthew Dunn:I'm a little tired of that target on my chest and of of feeling like I've been sliced and diced and stuck in a whole bunch of databases that I didn't actually ask to be in in the first place. So when someone says, oh, AIs gonna make that easier, we all have the same reaction like, here we go. Hey. I want it to be easier. I want it to be harder.
Matthew Dunn:I want a big red button that says, leave me the heck alone somewhere on my desk, which creates an interesting risk. Like, I suspect companies are gonna do a lot more making stuff up thanks to AI. Interesting.
Scott Cohen:Making I mean, they're already making stuff up now.
Matthew Dunn:Of course. Why? Because nobody has I mean,
Scott Cohen:just I mean, just look at LinkedIn and then here's the story of me meeting this CMO and they talked about this and you go, no. No. Or I mean, in the garbage in, garbage out, you know, I get cold email. Yeah. And I I mean, as CEO, I've gotten even more.
Scott Cohen:Right? Okay. You know, moving to moving to the CEO position, it's but then you get the whole, hey, as the wrong title. And I go, I don't hide who I am. Right.
Scott Cohen:So it on the one hand, it should make things easier. It also makes mistakes that much worse. Right? Because if people are out there with certain things about them and you get that part wrong, how does AI make that? Does that does it close the gap?
Scott Cohen:Or to your point, are they just going, whatever?
Matthew Dunn:Well, here's a different riff on making stuff up, and this is one this is the one I see as perhaps more dangerous. It's it's a lot easier to make, excuse me, credible enough stuff up at low to no cost that the volume is gonna go like this. Right? The cliche, not even 5 years ago in the email space was email from a Nigerian prince wanting to share wealth with you. And we would all laugh because we could all spot it.
Matthew Dunn:Right? Now I gotta bet there's some very well tailored, very PII driven emails that are scary good Mhmm. At looking like something that I should act on or respond to. Like, I got a string of spam mails the other day with the Wells Fargo logo and pseudo bank language, and, like, they weren't that well done. Like, oh, give me a break.
Matthew Dunn:No freaking way. Right? But they were a lot better than they were even a couple years ago.
Garin Hobbs:Oh, yeah.
Matthew Dunn:They keep getting closer. Yeah. And okeydokey. Let's say, hypothetically, they hit perfect. Right?
Matthew Dunn:Someone someone builds Spambot, and Spambot is just brilliant at taking everything known about Garin Hobbs and putting together the most compelling email from someone he knows and loves that gets him to click a link. How the heck are we gonna stop that? Yeah.
Garin Hobbs:Yeah. Just, and I don't know that we as humans are getting any smarter, so we're gonna have to rely yet again on technology to protect us from technology. Right?
Matthew Dunn:Yeah. Yeah. Yeah. Yeah.
Garin Hobbs:What was it?
Scott Cohen:They the AIs are gonna talk to each other? Like, no. No. No. No.
Garin Hobbs:Yeah. Yeah.
Matthew Dunn:It's a bit of a
Garin Hobbs:wolf and head head house type of situation for sure.
Matthew Dunn:But we'll and we'll always have the we'll always have the root reaction, I think. This is historically grounded. Root reaction for completely new set of possibilities is always a healthy degree of fear. We're all we always do the speculation about, like, the the bad side of it. Right?
Matthew Dunn:But when they when we didn't know what lay beyond a certain, you know, number of miles out to sea, we stuck monsters on the map and said, here be monsters.
Garin Hobbs:Yep.
Matthew Dunn:Why? Because we didn't know. Right? Yeah. And it was deep, and we had didn't have submarines.
Matthew Dunn:It it might be scary down there, and it might not. We don't really know. So one of our reactions is certainly gonna mix in a healthy, you know, as an animal species, healthy sense of of, of fear and flight. There's always gonna be people who exploit that gap too. Always, always, always.
Garin Hobbs:Mhmm.
Matthew Dunn:You gotta bet, you gotta bet there are, lots of bad guys doing whole lot of work with AI right now, like that stupid spam email that I got. And the notion of bad guy with some intent and a little bit of taste armed with an AI is a lot scarier than just the AI itself. Right? Well, I
Scott Cohen:mean, it's it's, you know, what's bad, the tool or the person using it. Right. Right? Yeah. It's it's like, going back to the seventies eighties nineties, it was when the tool becomes self aware and doesn't want the user to exist anymore.
Matthew Dunn:Right.
Scott Cohen:That that was what I that you know, that was the whole thing with Hal and with Skynet and all that was all good. That Yeah. They're gonna get rid of us because they don't need us anymore except as batteries if we're going to the matrix.
Matthew Dunn:You know, it's That's the here be monster scenario. Right? The we be monsters is the scarier scenario.
Garin Hobbs:Right. It is.
Scott Cohen:Yeah.
Matthew Dunn:Yeah. And what happens? Well, you get pirates. You get privateers. You get pilgrims, you get people rush into the new space, and then I'm fast forwarding a whole bunch.
Matthew Dunn:The successful ones ask for laws and barbed wire later on. Always. The guy who went out to the frontier and built a big old ranch is the 1st guy to own a law office and and titles on file and barbed wire defining his ranch. Why? Because he worked his butt off to build it, and he wants to keep it.
Matthew Dunn:Yep. So Yeah. And we're in the early stages of that right now. We're in the really early stages of the AI stuff. It's it's messy.
Matthew Dunn:It's goofy. There's gonna be bad guys. There's gonna be, you know, entrepreneurs AKA pirates. Like, we're gonna have we're gonna have the whole mix of it. But because it's not like a clear spatial division, you could say, I choose not to go to the frontier.
Matthew Dunn:We're living in the frontier. It does feel a little more imminent.
Scott Cohen:It's wild west right now. Yeah. Yeah. It is really what it is. Yeah.
Garin Hobbs:Yeah. I I liken it to the explosion of the oil boom in the early part of the last century. Right? Just wildcatters everywhere. Everyone trying to stake their own ground.
Garin Hobbs:Mhmm. Get what they can out of it. Take a different approach and just take what they can without really thinking about the responsible way forward. It's just forward, come hell or let it.
Matthew Dunn:Yeah. Yeah. And and look, we we you know, in the in the in the oil boom, you know, the rail oil industrial boom, we threw a whole lot of stuff over the side of the train onto the dirt and then think about it. And now we're going like, oh, that was that was awful. Right?
Matthew Dunn:What what we thought was an externality turns out to be very much not an externality. I wonder if trust and attention are at risk of being the things we pollute with what we're doing now.
Scott Cohen:Well, let's go back to that Wells Fargo example as a scary good example of what personalization can do to break through that. But let's talk about personalization on the positive side. Because a lot of what we talk about with that data and personalization is the output of that data.
Matthew Dunn:Yeah.
Scott Cohen:And but, I mean, if you ask 5 different people what personalization means, you get 5 different answers. Right? So what does personalization mean to you, doctor Matthew Dunn? I'm glad you asked because I actually wrote a,
Matthew Dunn:I don't know, came for a rant blog or just harassed people. I don't know why I could there is no actual definition of personalization. There is no definition. It's a BS term because we all use it, and we don't actually have any sort of rigorous structured definition of what the heck personalization is.
Scott Cohen:It's the email space. Do we define anything?
Matthew Dunn:No. We
Scott Cohen:just have to say for the same same thing. As I say, the same word Yeah. For the same thing. Yeah.
Matthew Dunn:Yeah. Yeah. But I mean, it looks to find the SMPP hello in that that way. It's like no. It it's it's it's remarkably squishy, remarkably, nonrigorous thing.
Matthew Dunn:When you look at the number of dollars attached to personalization, either spent or in effort, like, every company in the world is doing any marketing. You're spending tons on personalization. What is it? We don't really know.
Scott Cohen:I'm sorry. It's your it's your first name?
Matthew Dunn:Right. Right. Right. Now what's the motive? Because you always see the survey.
Matthew Dunn:Well, consumer surveyed want want personalization. Well, you asked them that. So, of course, they said yes. What do we think they mean when they say they want personalization? I mean, if someone asked me, I'd say, well, yeah.
Matthew Dunn:Of course. What do I mean by that? Well, more efficient, more relevant, less irrelevant, more not targeted. It's not what I want, but closer to my, interest predilection, something like that. I usually give the example.
Matthew Dunn:There's a big retailer that I bought a ton of outdoor stuff from over a lot of years, and they still send me ads for the other stuff. I'm like, frick, guys. Laser in. You know I fly fish. You know I this.
Matthew Dunn:You know I that. If you would put more of that under my nose, I would buy more of it. And I would hate your emails less because I'm not gonna buy the other stuff. Seems like personalization to me because there's a marriage of my interest and their interest, their business interest. Mhmm.
Matthew Dunn:That's probably better for both of us. So I don't know if that's a definition, but it's kind of a working heuristic or something like that. Like, okay. Let's let's find an overlap point where I don't want you to say, oh, yeah. We noticed you run Scott's podcast an hour ago, so we think we're gonna sell you a fly rod.
Matthew Dunn:That's creepy. Yeah. Right? That's too much. Right?
Matthew Dunn:I didn't I didn't give you a yes for that. But if I bought enough of this from you for you to go, oh, you seem to like this stuff, seems like a reasonable trade off. Right? Reasonable trade off of of interest in time efficiency and things like that. Well, especially with repetition.
Matthew Dunn:Right? I mean, I think there's because the repetition means learning.
Scott Cohen:Yeah. You know, if you only bought
Matthew Dunn:well, and but to the
Scott Cohen:point, it's like, if you only bought one set of fly fishing, like, that was your only thing you did with that retailer Yeah. They could go, okay. You did that, but you may also like these things. But if you've been buying the same stuff from them for 15 years and only that stuff Yeah. Yeah.
Scott Cohen:Yeah. Seems like a signal. It seems like a signal. But I I I'd like to say, you know, what we want in personalization is, like, the old 19 fifties, 19 forties, like, general store guy who knew what everybody wanted in town. Right?
Scott Cohen:It's like, oh, you have this? I'm gonna keep that in stock for you. Oh, you want this? I'm gonna keep that in stock for you. But nobody expects the big companies to be the general store owner.
Scott Cohen:If it, like like, that that's what one to one really means. Right? Like, if you run a general store and I come in and I I always buy us, an ice cream float and, you know, candy. And then the next day I walk in, you go, hey. Would you like a gun?
Scott Cohen:It might be a little weird. But if I came in and I said, hey. I need a gun. Yeah. You're gonna go out and get it for me.
Scott Cohen:Right? So it's just, like, you learn these things, and that's you can't you cannot replicate that at scale.
Matthew Dunn:You well, you can you can't. Right? And if you did, you'd probably scare the hell out of me, and I'd run screaming away and
Garin Hobbs:get
Matthew Dunn:bye from you again. Right? At the same time, and you figured we'd get here sooner or later, the number of things that Amazon knows about me. Like, oh, crap. This is just this is bad or it's good.
Matthew Dunn:I think it's good because I keep buying stuff there, but it's unbelievable the the number of purchases and the pattern to be derived from those purchases. And interestingly enough, one, Amazon never chases me an email to sell me more stuff. They've they opted a lot not to quite some time ago because to. They don't need to. Right?
Matthew Dunn:I'm gonna go on the page or more likely the app, and it's the pixels that are gonna show up in front of my eyes are as personalized as you could freaking imagine, and that's one of the reasons why I keep using it.
Scott Cohen:Like I'm gonna say and and Garin knows this story. But if they were to do that now, I'd probably be getting, like, items for coyote urine because I've been trying to get squirrels out of my backyard. And I bought that on Amazon. Talk about what you can buy on Amazon. Yes.
Scott Cohen:You can buy legitimately farmed coyote urine in a squirt bottle to spray around your house if you need
Matthew Dunn:to get rid of vermin. Because those will run away from that stuff.
Garin Hobbs:Apparently. God bless that honor, but I wouldn't wanna meet him.
Scott Cohen:And and back. Thanks. Sorry.
Garin Hobbs:Go ahead.
Matthew Dunn:And and the the crude digital no AI of not that many years ago, you'd had coyote urine ads chasing you all over the Internet for for weeks. And you'd be like, a, creepy, b, I bought some. Leave me alone. I don't need
Scott Cohen:2 of them. Right?
Matthew Dunn:So if if Amazon or a merchant like that could go bought 1, our pattern mining says he's probably not gonna need a second bottle, so let's lay off the coyote urine and find another topic that's more of interest. Right? Does not tend to be repeat searches. Check. Or would you like to subscribe, Scott?
Garin Hobbs:Let's talk about the repeat purchase versus the sort of anomalous, sort of purchase that seems completely out of cycle. Right? So going back to your early example, Matthew, of the, the outdoor retailer, we are continually buying fly fishing gear and and Scott's claim about, hey, the repetition is what makes that easy. Here's an example. I, I live in a rural area, so my, my local options for shopping are incredibly, incredibly limited.
Garin Hobbs:So I tend to buy a lot of my clothes online. Now I have one retailer that I favor over the others, and I'm continually buying shirts, pants, belts, shoes, things of that nature. Right? And in a consistent size despite my somewhat fluctuating waistline. In any case, occasionally, I'll buy a dress, or an item for my wife.
Garin Hobbs:What I receive next from this, from this retailer is other dresses you will love. Right? It's not in my size. It's completely out of category, from from the things that I buy, and yet they look at me as though this is something that I would love to continue buying without seeing that this is something that is out of cycle for me. What I'd love to see is that they recognize this is not my size.
Garin Hobbs:This is out of my category. He's buying for somebody else. Instead, it should be, yep. Here are other dresses she love. Not only is that more relevant, but plenty of studies show that when people are buying for others, they tend to spend more than when buying for themselves.
Garin Hobbs:Mhmm. Now when I received that email that's, well, I don't love dresses, delete. But if it were other dresses she'll love, it evokes that sort of warm feeling I get when I give something to my wife that she really enjoys. I'd be far more compelled to open the email clip and likely affect a purchase.
Matthew Dunn:Yeah. Yeah. No. It's it's a good example. What but on on let's unpack it a little bit and try and figure out, still somewhat, why it's not working like that now and what we'd have to change to get there.
Matthew Dunn:I mean, 1, the the repetition and pattern detection stuff. I mean, you'd think that's possible, but it's really easy to take the, you know, the convenient to hand narrative example of a story like that and go, should work. And then you're in the back end running the SQL queries or building the database, you go, oh my god. Right? We have so much stuff.
Matthew Dunn:How do we find those signals out of all of this noise? Right? Feels feels impossible now. We will get better at it, but it feels impossible now. One of my quips about AI is, like, computing stats had a baby.
Matthew Dunn:And and truthfully, one of the reasons I think we'll start sorting out stuff like finding that signal of this is Garin's size and gender, etcetera, is because it's really a stats and compute problem, a a a weird goopy pattern recognition problem. You're not gonna solve with human beings, but it's probably eventually algorithmically more solvable than it is now. Okay. Let's say we do solve it. If you got that email with the follow-up saying, shoes she might like as well.
Matthew Dunn:Is that cool by you? Is that creepy? Is that how did they know? What do you think?
Garin Hobbs:And now and now we're back to the subjectivity of what is creepy, with regards to product recommendations, especially those driven by AI. Personally Yeah.
Scott Cohen:Yeah.
Garin Hobbs:I would love it.
Scott Cohen:There I
Garin Hobbs:would love that. There's nothing more terrifying than the prospect of just sort of finally picking out something that I hope my wife will really enjoy. Right? I mean, so much is motion tied to it. There's, you know, certain tastes.
Garin Hobbs:There's
Matthew Dunn:Yeah.
Garin Hobbs:Mood. There's there's a lot of temporal factors, not just preference factors. Right?
Matthew Dunn:So We land fashion Scott's general store, don't we?
Garin Hobbs:No. We're Oh, shit. Yeah. The gun versus the kidney. For sure.
Garin Hobbs:What the hell's going on temporarily there? But, no, I mean, we don't tend to feel that it's particularly creepy when we see those types of, recommendations coming from folks like Amazon. They may seem a little unbluff, but I don't know too many people who took test you know, really get creeped out by that level. Yes. If it's helpful, and if it's relevant, and if it isn't just too completely wild, then I'd imagine it probably passes the creepy factor test.
Matthew Dunn:And it also passes a bit of a common sense test. Right? You go, I can I can see how you came up with that? Like, that's actually kinda reasonable. Like, oh oh, okay.
Matthew Dunn:But we bother ourselves, and I think probably should bother ourselves with the the the notion of the less obvious, less common sense, but nonetheless compelling offer. Like, when I when I talk with companies in the marketing space, particularly the ones who are in the sort of the data business that I just don't like. Oh, woah. Append blah blah blah to your first party records. I'm like, I really hate you guys because what you're what you're doing is arming someone to chase me with a rifle scope that I don't want them to have.
Matthew Dunn:I don't want I don't want another party to know enough about me to be able to compel me to do things based on what they know about me. And I at the same time, I don't know how to stop it.
Scott Cohen:Or let's Yeah. I mean, it's go ahead.
Garin Hobbs:I was gonna say, let's stick with this topic of creepy. Right? When we think about creepy, let's think about the root cause of creepy. And that's typically we feel as though we've been violated in some way. Typically, our privacy has been violated in some way.
Garin Hobbs:I'm I'm I'm, you exposed. Right? Which compels the Debbie Downer question here. How do we reconcile? Thinking everything that we've discussed, the good scenarios, the bad scenarios, how do you, Matthew, reconcile privacy amidst all of this?
Matthew Dunn:Partially a power thing, isn't it? Right? Power and control thing. If I looked up and I saw my next door neighbor with his nose pressed against the front window, I'd be pretty cheesed off. Right?
Matthew Dunn:It's my window. It's my lawn. Yes. I was sitting reading a book, but, you know, it's my business. If I wanted you to look, I freaking ask you to look.
Matthew Dunn:Right? There they there's a boundary of control that I don't think is unnatural that says, that's not okay. Jerry, go, like, go away. Clean the window before you do, by the way. Right?
Matthew Dunn:And and and companies are doing the equivalent of equivalent of it right now, and and I think that sorta creeping over same root word. Creeping over that boundary of what we thought was locus of control for ourselves is what or having power over us is what bothers us. Right? Oh, man. I got that super compelling, really highly targeted offer.
Matthew Dunn:Didn't ask for it, and yet it's right on the mark. And somehow it's really bothers him to think there's enough of me floating around for that to have worked or been economical to produce and stuff like that. And, I mean, look. The the boundary conditions that we all live with now would probably have flipped our respective fathers or grandfathers the hell out in space. Right?
Matthew Dunn:Like, you what? You let them what? You tell them what? You actually sign on and buy things on a computer? Are you crazy?
Matthew Dunn:Right? That that kind of stuff. And we've we I mentioned laws earlier. Right? We're we like, oh, we as a whole.
Matthew Dunn:We're going like, yeah, maybe they're maybe that's got a little too porous, a little too much. EU, completely different. Right? Barbed wire, no, You can't, you won't, you shouldn't, you don't, and the level of innovation they got to go with it. But leave that aside.
Matthew Dunn:Right? And one guy's tolerance and another guy's tolerance don't necessarily ever match up, so we we wait to make enough of a joint concern to make a joint decision to say, that's not okay, and that's now illegal. You can't do that anymore. And operators marketers operate to some extent right on the boundary condition of those things. If if you're if you're running a campaign for a client, Scott, and they're like, we really wanna go whole hog personalization.
Matthew Dunn:You're like, no. You don't have the data to do it. But leaving that aside, how far do you wanna go with this?
Scott Cohen:Right.
Matthew Dunn:Like, what like, in there's like, if you go too far, very destructive, isn't it? Like Yeah. Right? You can
Garin Hobbs:go to your feet. We noticed yeah. We we your daughter, Susie, looked really lovely stepping off the bus today in her blue dress. Here's a belt. Here's a lovely red patent leather belt that would look excellent against this blue dress.
Scott Cohen:It's like Yeah.
Matthew Dunn:Woah. Woah.
Scott Cohen:Well, I think there's also nat there's also the natural limitations of your product too. I mean, if you are an overstocker at Amazon, you have a ridiculous amount of SKUs. You could recommend anything under the sun.
Matthew Dunn:Yeah. Yeah.
Scott Cohen:We we have these. They are listening to us. Yeah. Regardless of what they tell you, we're not listening. How is it that Instagram gets ads that I go, I didn't even think I needed that, but holy crap.
Scott Cohen:That's cool. Right? It's probably because I mentioned something around it. Just around the phone. Right.
Matthew Dunn:It's The pattern says, why did the the the the the statistically highly likely to need one of these. Right. Right? Coyote urine, maybe it maybe squirrel gun next. Who knows?
Matthew Dunn:Right? Kinda relevant.
Garin Hobbs:It
Scott Cohen:seems that I've I'm I'm I'm almost there. I'm almost there. I'm not a gun I'm not a gun guy, but I'm almost there. Yeah.
Garin Hobbs:But They are But but
Matthew Dunn:They are really hard to
Scott Cohen:get rid of. But the only way the only way that somebody will truly know is, you know, the the must going, hey, we're gonna have Neuralink, and then we're gonna have that. I'm like, there's like, where like, people are willing to trade privacy for value just in general. I mean, you guys in degrees, especially in the US, but in degrees, of course. Right?
Scott Cohen:And and and those degrees are very dependent on brand or person or whoever it might be, like, to literally an audience of 1. Like you said earlier, like, our thresholds for what's creepy and what's not are different.
Matthew Dunn:Yeah.
Scott Cohen:And what's what would not be creepy for me for 1 brand, that same level would be creepy for another brand is what I'm trying to say. So it's you're talking about something extremely subjective Yeah. That but you still have to reconcile it. Right? Because once technology exists, who puts the barriers in?
Scott Cohen:Are we are we expecting the operators to put it in? Are we expecting other people to put it in? It's a it's a big question.
Matthew Dunn:It it we'll get that slow lag historical thing that says, eventually, you know, barbed wire and fences and lines and boundaries, eventually. We we we don't have it yet for some of these new spaces, conceptual spaces, but, eventually, we will. Just like we're now having conversations about Internet privacy that maybe we wish we'd had 10 years ago, but we, you know, we didn't we didn't. There wasn't enough for human cry to do it. And for what it's worth, not a product plug.
Matthew Dunn:With social signal, like, that's a the issues you raised have been front and center conversation for us because what we're about is telling you what broad group Garin's probably a member of and what that group is like and not telling you anything about Garin, thinking that this is our thesis. Right? Knowing and by broad group, I mean, like, half a 1000000 people, 1000000 people, 2,000,000 people, surprisingly surprisingly tight segment, thinking that that's actually closer to what, Garin or Scott wants in a way of personalized treatment, then let me spend a year building up a file on you and then talk about Susie stepping off the bus. Yeah. Yeah.
Matthew Dunn:And it's, it's an interesting challenge to sell because the the sort of first party collect first party to like, a squirrel collects stuff in Scott's yard to eventually know enough to personalize. I think I don't think that method works. I think it's talked about was it, Gartner. Gartner said 18% of companies, 18% have the 360 degree view of the customer, meaning almost everybody fails at this. Yeah.
Matthew Dunn:And I'm not surprised.
Scott Cohen:And I would say 50% of 18% are probably lying. But but I
Matthew Dunn:think you're right.
Scott Cohen:We're all we're
Matthew Dunn:all facing this weird chimera because of dear first name, I think, that that somehow collecting enough nuts means you're gonna get a cake. Yeah. It doesn't. It was talking about, hey. Listen to me.
Matthew Dunn:How did you guys do this? That's annoying.
Scott Cohen:You know, to your point, AI has been around for 50 years. Right? So it's we've been talking about the same stuff for 50 years. I mean, people ask me all the time, what's different in email? I'm like, the how is different.
Scott Cohen:The speed is different. The why, not so much. We're still trying to communicate with people, still trying to reach them emotionally, still trying to get them to buy things that they may or may not actually need. You think it's gonna get harder too?
Matthew Dunn:Yeah. I mean, you guys are both practitioners. Just, like, I I look at my blast screens for what hits my inbox and the exhaustion of opening that inbox, and I'm like, shit.
Scott Cohen:But direct mail it's why direct mail is back. Interesting.
Garin Hobbs:I I think I think it will get harder. We're seeing both the evolution as well as the democratization of these technologies, meaning they're available to more and more people, which has really leveled the playing field with regards to, you know, retail, ecommerce, even just, you know, provision of services, things of that nature. Yeah. Yeah. Choice is only gonna continue to grow, and, you know, competition for our dollar, for our attention Choice.
Garin Hobbs:Has has exactly. It has to grow accordingly. So Yeah. It it is gonna become difficult. I don't think it's gonna become as difficult for those companies.
Garin Hobbs:I think it's gonna become difficult for us as consumers, as decision makers to really understand what is, what is best, and why, and and and the impact of, you know, the impact of me and my decisions on that. So, I I I think it's gonna get easier for those who have for the purveyors. I think it's gonna get a little bit more difficult, challenging, and certainly confusing for those of us on the receiving side.
Matthew Dunn:Yeah. Yeah. Which means that I think. Which means that the companies, brands, organizations, or whatever, who figure out how to hit it right will do well, and the ones that hit it wrong will probably have an even harder time staying alive. And I don't I don't know the right and wrong.
Matthew Dunn:I mean, that's that's what you told that's what you guys advise clients on, but, you know, I this this super digital monopolies that are already in place are probably gonna keep doing just fine. Right? Amazon's not going away. And the companies and brands that that might earn my particular allegiance because of how they treat me. Maybe different maybe different for somewhere else, but I think our tolerance for being handled poorly is probably going to get even lower.
Scott Cohen:Oh, yeah.
Matthew Dunn:Yeah. Yeah.
Garin Hobbs:Yeah. You yeah.
Scott Cohen:You look
Garin Hobbs:at the younger generations and, you know, all studies indicators the majority of studies indicate there is no brand loyalty there. It's literally driven by price of value. Right? So I think those become the sort of tips of the sphere.
Matthew Dunn:Price, value, convenience. Right?
Garin Hobbs:Exact well, I mean yeah. Correct. And then convenience and value both being subjective. You know, price is really Yeah. Yeah.
Garin Hobbs:The primary driver of that.
Matthew Dunn:So Yeah. Yeah. So are we gonna commoditize ourselves to death with that?
Garin Hobbs:I think we kind of already are. We are the product in most cases. Right? It's not, it it's it's not who we're buying from. It's it's what we're providing back.
Garin Hobbs:It's Yeah. They
Scott Cohen:They also don't have money. So, you know, you you buy what you're able to buy too. I mean, you get to a certain level, and you can say, I'll invest in things I use every day and go cheap elsewhere. But, I mean
Matthew Dunn:But I'll bet you didn't I'll bet you bought the a grade coyote urine. I'm thinking you did not cut corners. That's important.
Scott Cohen:I did read reviews. I will tell you that much. It had to be at least 4 stars. Had to be at least 4 stars. Well, I think that's a great place to stop since we started and ended with Coyote Urine.
Scott Cohen:Matthew, where can people find out more about you and Campaign Genius and Social Signal?
Matthew Dunn:Oh, let's do a backwards social signal dot ai, website, obviously, and then campaign genius is campaign genius dot I o. Haven't figured out those 2 yet. And then, for professionally, I'm easy to find on LinkedIn. Just look for Matthew Dunn or doctor Matt.
Scott Cohen:Sounds great. Well, thank you so much for joining us, and thanks to our listeners. Oh, yeah. This is great. We could have gone on for a lot longer, but, you know, let's Always.
Scott Cohen:Let's let's be fair to our listeners. If you'd like to learn more about Inbox Army, go to inboxarmy.com. Until next time, be safe and be well.