050: Beyond Data: How ChatGPT Uncovers Emotions to Fuel Effective Ad Concepts with Joe Putnam

Speaker A: Welcome to the Intended Consequences podcast. I'm Brian Massey, your host, and I'm pleased to be joined by Joe Putnam. Um, he's the founder of Conversion Engine, an E commerce performance marketing agency. He's been doing this for 15 years and he's got a proven track record of helping e commerce brands quickly scale and grow. As it turns out, he went to school right here in our hometown of Austin, Texas at the University of Texas, and he just seems to love innovative strategies that help brands grow. Uh, Joe, welcome to Intended Consequences.

Speaker B: Thank you. Appreciate it. Great to be here.

Speaker A: So I've been looking forward to this conversation, uh, especially because, uh, one of the strategies you've been using is AI and um, in particular sentiment analysis. So those are some kind of big words. Um, most of us know what they mean, but why don't you tell us a little bit? Just give us the overview of what you're doing and how that works, uh, in your area where you like to grow e commerce companies.

Speaker B: Sure, yeah. So first thing I would say is, um, I guess personally and as an agency, we have a very big interest in the kind of the emotional side of marketing and uh, the emotional experience when it comes to brands and their ads and, and things of that nature. Um, just specifically because we're very big believers in. People buy based on emotions and then they justify with their logic. Um, and my favorite example of that is, you know, I always think of a BMW car and a, um, lot of people want a BMW, um, but most people don't need one. But, um, what happens is their, their emotions get activated, they desire it, they want it, um, and then eventually they kind of justify the reasons that they do need it. Even though, you know, a Toyota Camry will get them, uh, from point A to point B just as well and more reliably. Um, but you know, people start telling themselves this story of, you know, why they need a BMW. And then as we know, they go out and they purchase it and they drive it and things like that. And so, yeah, just big believers that the buying process, the marketing process, um, you know, brands are going to do better, have better results, build stronger brands when they're able to connect emotionally with customers, kind of create an emotional experience with their ads and things like that. And so I guess that's kind of the starting point and the genesis of this. And then, you know, kind of, as you mentioned, we're using AI to, uh, um, you know, kind of help with that process and make sure that we're able to do that for brands.

Speaker A: Yeah. Are Most of the brands you work with in consumables, you know, I'm thinking apparel, uh, accessories, things like that. Or will this work for people who are selling bathroom scales or storage for their closets, Things like that. I'm kind of distinguishing because you, uh, know BMW, um, has this emotional content. And I used to be a BMW guy myself, made all of those justifications. But versus things that are more utility, um, do you understand my distinction?

Speaker B: I do. And I would, I would say the answer is yes. You know, no matter what you're selling, you're looking to, um, tap into the emotions of buyers. So, you know, even if that emotion is fear or if that emotion is anxiety or whatever it might be, um, you know, when it comes to marketing, when it comes to ads, um, you know you're going to be more successful when you do find a way to touch people at an emotional level, kind of connect to them in that type of a way. And so, yeah, it might be more fun to do it for a brand like BMW. And we do work with a lot of apparel brands. You know, we've worked with, with baby brands, we've worked with sports brands, um, lacrosse brands, golf brands, things like that. But I would say, you know, I would, I would uh, I would start by saying that, you know, no matter what you're doing, um, this kind of analysis would be helpful for your marketing.

Speaker A: Fantastic. You'd mentioned billing anxiety. One of my favorite strategies in terms of revising copy on websites is to agitate the problem. So, uh, talk more about what would happen if you didn't take advantage of the offer, which is anxiety building. Um, how does this work in your area, which is primarily ads, I assume. And are we talking about social media ads? Uh, paid search ads do. Since you're in with E Commerce, are you working with paid shopping ads? Um, which feel to me like there's less space to create sentiment or to create emotion, but I may be wrong about that. Talk a little bit about what that looks like in implementation.

Speaker B: Yeah, so we're primarily focused on um, meta ads for this, this type of analysis. Um, but I would say there is, you know, kind of some spillover to something like your Google search ads or whatever that might be. Google display ads, um, shopping, like you said, probably is a little bit more of a, uh, you know, kind of what you see is what you get type of experience, but you um, know you're reaching people at a point where that's what they need. But um, but yeah, so I guess how, you know, what are we doing how are we using AI? What does it look like this look like for us? You know, so first of all, I would just say without AI, um, it means that we would have to be kind of doing this kind of thing manually. Um, so, you know, let's say, let's think about conversion rate optimization for a second or let's think about copywriting for a second. One of the best things you can do for both of those, um, is, you know, study, uh, the reviews, so study the words that customers are actually using. Um, you know, so for our copywriting process, you know, that kind of a thing, um, you know, part of that process historically has been, you know, we would go to Amazon and you read through hundreds of reviews and then you start to, you start to categorize, okay, what language are they using? Um, um, you know, what problems is being solved, kind of what is the experience? Um, that sort of a thing. And then eventually you use that to craft your copy, your angles, your, your, Your ads and things like that. Where even, I mean, to me, some of the best copy we've run is we find, you know, a review for a baby bassinet on Amazon that is just, you know, it's in the mother's own words. It's their language, it's powerful, and you just kind of copy that, put it into Facebook, and that ends up being, you know, amazing, uh, Facebook ad copy versus, you know, trying to have an ad buyer put it into words for you. You know, you can kind of mind that by studying reviews and things like that. And so, so again, that's kind of the manual side and what people have been doing for a long time, and we've been doing for a long time. Um, but what we're able to do with, with AI is we're able to, you can either copy and paste your reviews into chat GPT, um, or you can, um, uh, you know, whatever, kind of find a way to import those. So, you know, you might have a way to download them and import with them, whatever that might be. But, you know, a lot of times what we're doing just now because just to make it faster, um, is we'll, you know, copy and paste hundreds of reviews into chat, uh, GPT, and then we'll basically have it ask it to conduct emotional sentiment analysis, um, on. On those reviews. So have it kind of pull out, you know, what is the emotional experience? What problem is being solved? What are the pain points? Um, and, you know, AI just does a very good job of if you ask it for those things, um, you know, what are the unique selling points? What are the pain points? What, um, you know, what problem does this product solve? But also, again, you know, um, do an emotional sentiment analysis on this and kind of pull out what emotions are people expressing, you know, kind of what is their experience with this product. Um, so we're able to, to make that process, you know, go a lot faster with AI than if we had to do it by ourselves.

Speaker A: How complex are those prompts? Are you, uh, are you, like, dropping in information you have about Personas? And do you have some pretty complex prompts that give you exactly what you need? Or could someone in our audience, uh, try this, drop a number of reviews in and start asking questions about it, like, what's. What are the. What are the sentiments of this? Um, what are the primary care abouts? Um, you know, basic language. Is this. Is there some secret sauce in your prompting?

Speaker B: Yeah, so I would say, um, I don't have secret sauce to the prompting. I personally have found that actually keeping it simple and kind of letting the AI figure it out has worked well for us. Uh, and I don't know if somebody is better at AI than I am and could kind of, you know, maybe put in a better prompt. But we basically keep our prompts very simple, simple. So we'll say something like, you know, conduct emotional sentiment analysis on these reviews, um, and then it will pump it out. And then based on what is kind of, you know, pushed out from. From, um, gbt, then that's where we'll kind of start to drill in. We might ask more specific questions. But yeah, if you're getting started and you're not sure, um, I would start with kind of some more simple prompts and then just continue prompting it if you need something more specific or if you need more information or anything like that.

Speaker A: So I assume that ChatGPT knows what the words sentiment analysis mean, and so you can just tell it to do that, or do you need to say, like, what emotions and concerns, like, do you have to be more specific?

Speaker B: So, yeah, so we just, we just tell it to kind of, uh, analyze based on. Based on that.

Speaker A: What sorts of sentiments does it spit out? Give me some examples of, of what, what do you expect to see?

Speaker B: Yeah, so, great question, Brian. Um, here. Here's a couple examples of, uh, you know, something we've done for, for a client. Um, so this is a boot brand. And so, you know, like we talked about, we put the prompt in and said, conduct emotional, um, sentiment analysis on this particular brand. And, um, so one of the things that came out was the customers talk about a deep emotional connection to heritage, pride, belonging, and gratitude. And then, uh, just some of the themes that kind of. It pulls out. So again, kind of what's their emotional experience with the brand? Um, is, you know, first of all, heritage and pride. Um, so people kind of talk about the heritage of the brand and how they feel, experience a sense of pride with the brand. With the brand.

Speaker A: Um, it's a work boot brand or a cowboy, uh, boot brand. Can you give us a little bit more context?

Speaker B: Cowboy boot brand.

Speaker A: All right.

Speaker B: That boy. Perfect protection. That's right. That's right. Yeah. Great example. Um, so a sense of belonging and family. Um, so, you know, people, um, talk about being part of a family, being part of. Part of a community. Um, so that kind of came out, um, trust and security. So they feel a lot of confidence with the brand. Um, and they say things like, I know these will last. I, um, can rely on them. They're worth every penny. You know, as you might imagine that, you know, kind of a line like worth every penny is a. Is a great, um, thing to be pulling out. Um, you know, kind of joy and satisfaction is another one. So it talks about exceeding expectations, um, you know, pure joy. You know, these are kind of phrases that people are using. So when they buy it, there's a lot of satisfaction, kind of a lot of joy. Um, and the last one, this is probably my favorite is, um, you know, personal achievement and confidence. So, um, you know, wearing the boots gives them a sense of confidence. Um, they also, you know, feel like they're. They're kind of, in a sense, rewarding themselves. It's a personal milestone. There's a feeling of pride, a feeling of self worth, a feeling of, hey, I'm showing up, you know, uh, in the workplace with more confidence. And so, as you might imagine, to me, out of all of these, that's the one as an advertiser, that kind of gets me the most excited to kind of be able to pull that out and start working with that. And so I guess one thing I would say is we use this and, and we have kind of a step by step process after we do something like this, to drill into each of these further and then come up with different angles for the brand. So, you know, this one ad will target this angle. Um, this other set of ads will kind of target another angle. And, you know, sometimes you want to have, or a lot of times you want to have multiple angles going at once. Um, or you might find that angle A and B works really well. And C and D is not as effective, so let's focus on A and B, but it at least gives you kind of more things you can test. Um, and what I, what I've personally found is a lot of people, when they run their ad campaigns, they might have kind of one angle and all of their ads. So let's just take a boot company as an example. All of their ads might. Their Facebook ad copy targets. Um, are you tired of boots that hurt? And so they're just kind of hitting that over and over again. Tired of boots that hurt, tired of boots that hurt, etc. And that's fine. I mean, that's. That's a good angle. Let's test it, let's run it, let's try it. But, um, you know, if people are talking about, hey, we're buying these boots, and, and I feel more confident and I feel, you know, like I'm, I'm, uh, you know, more successful when I show up in these boots, then I would say, hey, let's make sure we have ads that target that angle as well. And kind of the, you know, kind of the outcome angle. And, and this is what, you know, you're going to experience how we're going to make your life better and not just, you know, camp out on one angle. Like, tired of boots that hurt. So that's. Those are kind of examples of how we would use this.

Speaker A: Yeah, let's tease that apart a little bit, since I'm in the target market and I can have some commentary on that. So I grew up in Texas and absolutely relate to the messages that are sent. So if I'm, if I wear my cowboy boots to a business meeting, especially if I'm, you know, in the Northeast or something like that, I'm making a very particular kind of statement. Sure. Sometimes I'm worried that it clashes with the, uh, you know, the lab coat Persona, the scientist Persona. But, yes, boots that I'm gonna wear for years because they get more and more beautiful as they wear and get sure the right kind of scuffs and, and creases in them. Um, my favorite pair of boots was given to me by a friend who's, who were his grandfathers. They're just gorgeous. They're 40, 50 years old. I was able to get them restored. And so there is a, uh, there is this definite sense of that. Now if the sentiment analysis kicks out something like, talk about joy. Uh, it's not something that I really go in thinking about. And even, even if we get a little bit closer to like, am I, am I shopping for boots that make me feel more confident work? If you said those words out loud, it may not land with me, but I am looking for that, that feeling. Uh, I am picturing what they're going to look like in a few years. How does the marketer. Well, there's two things here. Number one, the AI might be right. So intuitively I say, well, I don't know, Joy is the right way to go. So you, you edit it out and that's our brains using their shortcuts and their, my own personal preferences rather than trusting the, the prompt. Do you recommend testing some of these corner cases to see if there's something there? Or do you want the, the marketer to, to do some of that editing?

Speaker B: Yeah, so, I mean, I think that is a, a fantastic question. Um, and so, and I also think it comes down to, you know, people are worried about AI, um, you know, taking, you know, my job or someone like me, um, taking, taking kind of our work away. And you think as a brand owner, hey, I can just do this all by myself and I don't need a marketing expert anymore. Um, I would say you hit the nail on the head why you still need a marketing expert. So, um, you know, anyway, even for us personally as an agency, when we're, when we're working on something like this, not everything AI out is going to be, you know, kind of a home run or it's going to be something that you're going to try to do. And so, you know, since it's kind of mechanical and so I would say really what you're doing is you're trying to use it to generate more ideas and then you're trying to use it to find the ideas that are going to make a bigger difference when put in an expert's hands. When put into somebody who knows a lot about marketing, you know, has been doing marketing for 15 plus years like I have. And so, you know, again, um, my thought is how do we create some ads that kind of are focused on the outcome? You know, it's focused on how do we make someone's life better? How do we make them feel, you know, emotionally feel like, hey, I buy these boots, but my life's going to be better in these ways. Um, you know, I'm going to be, uh, people are going to be more impressed when I show up. These booths are going to last a long time. Uh, I'm going to feel comfortable and confident in them. So those are kind of marketing type words. But you have to Figure out as a marketer how to um, customize that in a way where it actually makes sense for the demographic, for the actual customer, things like that. So with this one specifically as an example, I agree with you completely. Um, you know, um, for, for you might have to kind of spend more time on the joy and satisfaction, um, and see, hey, is there something we can pull out here or not? Um, but for sure, like I said, out of all of these, when I got to the last one, personal achievement and confidence as a marketer, I tell myself, okay, this is where we need to camp out. This is where we need to spend some time. We need ad creative that shows this to people. We need copy that kind of conveys this to people. Um, so anyway, I guess the shorter answer to that is I would say it makes sense to test more angles, but you do want to have a marketer and expertise and not just kind of throw spaghetti at the wall. It's like, how do we find the 3, 4, 5 angles that are most likely to succeed and not just do everything that A.I. uh, tells us to do?

Speaker A: Yeah. And you and I have an interesting additional challenge in that we have to sell this to the customer. So if we come in and we say we think we want to test one of these corner cases because there might, we might be really surprised if, you know, the joy of a comfortable pair of boots and cowboy boots is a fairly macho esque, uh, marketplace, um, just uh, the way it's been marketed to the west and everything like that. And uh, you can see the, the, the customer going like joy. Do we really, do we want to talk about joy here? But the, the right answer is it showed up in sentiment analysis. It significantly. And so it's, it's a, we should test it to see if we would be surprised by this or if we can use data to eliminate it. And that is the choice when we're working on product. Um, page copy is probably the closest analog to what you're doing. Um, and we have not been doing sentiment analysis on those things. Um, but the beauty of AI is that it allows us to test multiple product pages with very consistent rewrites of the descriptions.

Speaker B: Sure, sure.

Speaker A: That is a, ah, beautiful ability that was just very difficult to do with human beings, um, prior to this. What are the mistakes that we're going to make? Um, things that come to mind is how do you know that AI is not hallucinating in its sentiment analysis and giving you ideas, uh, that really aren't supported in the uh, reviews? How do you guard Rail that. And is there a way to kind of backstop it and see if uh, you find some other data that points to that direction and talk a little bit about how you safeguard things?

Speaker B: Sure, yeah. I mean, so I think I would say, um, the main thing we do is that's where human intelligence comes in to be the safeguard and the, you know, the backdrop. And, and again, I think people have this concept of AI is going to be able to replace people and things like that. Um, but really, um, you know, so if you put this information into um, a 22 year old college graduate's hands, they may not know how to create ads that are going to scale from this information. So you do need somebody who has 5, 10, 15, 20 years of experience, is an expert marketer, um, uh, expert in Facebook ads, to you know, kind of, um, use the information to turn it into something that actually resonates with, with people, which is the goal. And so again, I think the biggest is if you expect AI to be able to do all of that, um, you know, they can't kind of get it all the way from, from all the way to the goal line. But a couple things that we do is we have kind of an extended process for this. So we talked about kind of the emotional sentiment analysis of the reviews, but that's really actually a starting point. So what we do next is we again we have this process I'm going to use in house, where we start there, but then we do pain point analysis. So we say from these same reviews, you know, what are the actual pain points people are experiencing that our products solve? Um, what are some of the, what are the unique value propositions? So based on the reviews, what people actually talk about, what they're saying, what are some unique value propositions for this brand? Um, we use all of that to then get to the point of, hey, what should the core messaging pillars for this brand be? You know, maybe it's three, maybe it's four, maybe it's five. But like these are, you know, whether it's a social media post, whether it's an email, whether it's an ad, whether it's landing page copy, these are the core messaging pillars that this brand can really stand on and get behind. Um, anyway, then, then finally we use that to come up with sample ad angles. So based on those core messaging pillars, what are some, some ad angles we could use, um, as a launching point for testing and seeing what's going to work. Um, but again, throughout that entire process, but especially when you get to that ad Angle level. That's where you need somebody who can say, hey, we've got five ideas here. These three are great. These two are not on brand or these two are not something that's going to be a good fit for the target market. Um, so again that's where you know, human intelligence comes in. And at least for the time being, we are still, um, you know, smarter than the AI when it comes to understanding people.

Speaker A: Yeah, yeah. And we, we tend to shy away from what sometimes AI will generate. Game show host style copy and it.

Speaker B: Correct. Yeah.

Speaker A: There's only certain kinds of brands that it fits on. Let's talk a little bit about the landing experience. The thing that kind of sticks in my craw is that when an ad doesn't work, you know, it might generate a lot of clicks, but it doesn't generate conversions. The answer is to try a different ad rather than go and see if maybe there's something happening. The landing experience that is not converting those clicks. How I know it's not something that you play in a lot, but certainly you've been frustrated by the, the kind of pages that you're having to send ads to. What's your advice for clients in terms of like, you know, we need, we need something better, uh, to send these folks to.

Speaker B: Yeah. So, uh, I love the question. Um, yeah, I would just say, you know, I mean as an advertiser you have to figure out, hey, where is the best place to send people? You know, what is it that we can do? And I would also say that there's only so much you can do on the ad side. So you can have an ad where, you know, kind of all the metrics are super high. So uh, the stickiness rate of the video, kind of the watch time, how long people watch it, um, you know, the click through rate, the click through rate is high, all that kind of stuff. So that, you know, the ad numbers are saying, hey, these ads are resonating with people, um, but then you're kind of dropping them on the site, um, and you know, kind of it's not working or whatever. So I would say first thing is anytime you can have more alignment between the ad and the landing page, that's going to be um, a great opportunity for a win. So thinking about these different ad angles and core messaging pillars as an example, you know, if you're emphasizing one particular ad angle or pillar, you know, you want to drop people on a landing page that kind of matches that experience, matches that sentiment. Um, so I would say, you know, for something like this if you really want to get the best results, you know, you probably would have different landing pages for each of those, you know, potentially messaging pillars. Um, or maybe you combine all of it onto one page. But alignment is obviously a big thing where whatever they experience in the ad that got their attention, if they have that same kind of sentiment, experience messaging on the landing page, that's going to be more effective. Uh, but the other thing I would say when it comes to CRO, um, and one reason why I think it is, is underrated, is again, there's only so much you can do on the ad side. But let's say, you know, your conversion rate is 1.5% and you actually design a landing page and you, you, um, um, you know, whatever, you know, you improve conversion rates to 3%. Um, all of a sudden your ads are twice as effective, your ad budget is twice as effective. So now you just unlock this opportunity to scale and to grow and things like that. And so I do think people have a tendency to make mistake of kind of everybody gets very focused on the ad account, very focused on roas, very focused on the numbers there. And I like to tell people, especially brand owners, um, you know, our job is to improve the ad account. Uh, and we're experts at it and we do it all the time. And I just tell them, you know, what we need from you is for you to think of everything outside of the ad account you can do to improve results in the ad account. And again, if you double conversion rates, all of a sudden, you might be able to, you know, 4x your ad spend because your, your ad spend is twice as effective. And so, yeah, uh, yeah, that's what I would. How I combine those two, it probably

Speaker A: bears a little bit of refining the terms that we're using. So, um, in our experience, we're social ads. They are, uh, either going to send someone to a particular product page. If they're using on a particular product, they're going to, uh, go to a category page.

Speaker B: Correct.

Speaker A: Or a product listing page. Um, the disadvantage of the product listing page is it's great for people who want to shop, you know, just for the phone shopping. But it doesn't do a lot to talk. There's not space to talk about, um, the, the key overall benefits, like the reason that this product set is better. So developing a landing page which shows the product, ideally that was in the ad, but then talks about the line, the other options, why it's the best, uh, the cheapest, the biggest, the highest quality, whatever your value proposition. Is and then start listing a few of the options. There's with opportunities to go to a category page, um, sometimes that's going to perform better. And in our experience m, those landing pages can really perform better on the, what we call interrupt advertising which we find in social media as opposed to the intent driven, uh, advertising that we see on search engines. So yeah, yeah, um, I think I've got a pretty good idea of how this could work and I feel like I could go and do this. Any final thoughts that you want to leave our intrepid listeners to? I think there's a lot going on right now where people are jumping into ChatGPT and um, not getting the results they had hoped. Like they're not happy with the results. They don't seem to be much better than what they would have thought of themselves. My theory is that they're not using AI Right. Anything you can tell them to help, uh, help them get more uh, or get better results, more surprising results for from AI.

Speaker B: Yeah. And so I would, I mean, yeah, my thought is, and we're all still learning and it's all very early and so, um, you know, still learning on our side as well. Personally, I feel like I have found that um, uh, not worrying so much about making the prompt perfect and actually keeping it simple and kind of letting AI um, figure out the prompt has actually worked well for us. So I know sometimes people feel like they have to give it a ton of parameters and that's the only way it's going to work. Personally, um, I found that keeping it simple actually tends to work really well. And then you can kind of refine it on the next prompt if you don't get what it is that you're looking for. Um, but then the other thing I would say is for us, AI has been great as a, as an idea generator and as a starting point. And so again, you know, um, if we have to do it ourselves and we have to study reviews on our own, you know, we're talking about hours and hours of work. Um, and then, you know, we might come up with three different ideas or whatever that might be. But it's, you know, it's, it's tied to a lot of man hours to do that process. Um, so obviously AI provides a shortcut in that way. But you know, let's say it shoots out seven ideas. Um, two of those might be fantastic and five of them you don't like. And so don't write it off because five ideas don't work. Um, you know, realize, hey, this Helped us to get to seven ideas faster. And these two potentially can be winners. And so, you know, Instead of putting 10 hours into this, um, you know, it took us an hour to find these two new angles. Um, and then obviously you kind of run with that and go with that and so. And then I guess the last one would be just to realize there is a need for human editing and there is still is a need for expertise. And so, you know, you're going to get better results from this, this type of information in our hands since we have 15 years of marketing experience than you are if you just kind of try to do it by yourself, although anybody can kind of try it and start it. But you know, at the, at the, for the last mile, you do need someone like yourself or myself or someone on my team to um, kind of put the finishing touches on it and do it in a way that really kind of connects and resonates with customers. So, yeah, I think it opens up a lot of potential, a lot of opportunities. It's a shortcut, um, but still needs, you know, kind of some human editing and, and human taste to get it to the, to the final product.

Speaker A: Yeah. Do, uh, do you have any, um, information on your blog about this sort of thing? Anything you can send our listeners to?

Speaker B: I don't yet. Um, I need to, to work on that. But, um, anyway, people can find us@ ConversionEngine.com so, um, great place to get more information. Um, and yeah, I haven't published a ton about this yet. This is kind of one of the first places we've released and talked about it. So, um, happy to share it about it here. But, um, yeah, anyway, people can find us@ ConversionEngine.com and then I'm on Twitter at Joseph Putnam. So if anybody wants to go there and uh, check me out there, they can.

Speaker A: And Joseph Putnam, if you've got questions, I'm sure that you could sign up for a consultation with Joe and ask him some questions. And um, Twitter is where you hang out the most. Twitter?

Speaker B: Yeah, I guess. LinkedIn. So if people want to find me on LinkedIn, they can find me there as well. So I'm the Most active on LinkedIn and Twitter.

Speaker A: Okay. Okay. Well, I'm glad you were able to join us. Um, this conversation was everything I'd hoped it would be. Um, thank you and uh, to our listeners, thank you for listening.