Podcast: Deploying AI personalization at scale (with Christina Augustine)

Mobile game developers no longer need to pay up to 30% in major app store fees. With Exolo Web Shop, you can create a direct storefront, cut fees down to as low as 5% and keep players engaged with bundles, rewards, and analytics. Start today at exola.com. That's xsol.com. or use the link in the episode show notes.
The problem is that the distinction needs to be drawn between the competence of the economists and the correctness of their analysis.
Welcome to the mobile dev memo podcast. I'm your host Eric Seufort and I'm joined today by Christy Augustine. Christie, welcome to the podcast.
Thank you. Thanks for having me.
Well, thanks for being here. So today we are going to be talking all about AI enabled personalization and I'm happy to

have that conversation. Before we get there, can you introduce yourself to the audience and provide some background on on yourself?
Yeah, sure. I'm currently the chief operating officer at Bloom Reach. We are an agent commerce personalization platform. So really looking at combining customer data, product data across all the marketing and on-site channels to make sure that we can deliver personalization and uh my background is as a software engineer as well as a management consultant at BAM.
Just out of curiosity, how did you get into this particular space? Like what brought you into this element? Because Bloom has been operating for some time, right? So it's not necessarily seizing on the current kind of agentic, you know, sort of personalization trend. How how did you move into this area?
Yeah, I've been at Bloomra for about 15 years now. So, since we were an early stage uh startup company focused in SEO and we were leveraging AI for SEO optimization, make sure that people got discovered in uh in in Google and in

Bing at the time. And then the what we did was we were taking this the algorithms that we were building to identify relevance and bringing it on site. So our next product was site search and we were really optimizing the the site experience. And so with all of those predictive AI algorithms and and machine learning as generative AI came about, it was just a natural place for us to start building agents. So we have conversational agents for shopping, we have campaign agents for email marketing types of uh activities. And so it was kind of a natural evolution to go from really using as much data and signal as possible to personalize to moving into more of the agent to agent space.
So Bloom is an established company, right? So when did that transition start? like if you had to kind and just kind not not you know specific to you know the marketing use case but but as you've seen it like you just kind of the

agentic adaptation to my mind there's like kind of two distinct waves and and maybe correct me if I'm wrong or if you just saw a different timeline there's like the kind of very recent adoption with with things like open claw and you know people bringing agents into their computer and and having them operate across everything but then there's also been just kind of like this slower maybe more subtle adaptation of agents you know kind of just starting with the introduction of some of these tools from the bigger model providers, but but then even like just more discreet adaptations before that going back years and years. Talk to me about how you've seen that timeline evolve.
Yeah. Oh my god. Kind of happened overnight and then happened so quickly in a way I think for most people, but as you mentioned, AI has been around for a really long time. It's been around since the 1950s. And so from a development standpoint and what we've all wanted to do with software, I think this notion of personalization and AI has been around for a very long time, but there hasn't really been the interface or the

connectivity that we're seeing now. And so certainly generative AI was around for a while. Bloomri was playing with things like vector search before people were seeing AI really from chat GPT and this interface where now as a person I can interact with it. Once that kind of happened and the environment started opening up now you're seeing a rate of development that's really quick but you're also seeing all of these channels that couldn't talk to each other before now are able to talk to each other and people have even more signal. So, it's kind of like we went from, you know, predictive AI with a few signals and you kind of always wondered like, why are you recommending this product to me when it's a terrible idea? Um, this isn't what I wanted to much more accurate predictions because now I can see data from not just email and SMS, but I can start to see it from other channels that maybe were closed off before.

Yeah, I mean, I think that's that's that's really helpful context. Maybe just kind of as a as a broad kind of kickoff question, I usually start the podcast with like a big big picture high concept question. What is AI empowered personalization mean in the marketing context? What are we talking about when we talk about AI empowered personalization?
Yeah, it's such a good question because I feel like 10 years ago, we would have said something very different than what we would say today. 10 years ago, we would have said, "Oh, you know, we're sending AI powered personalization means I'm sending an email that says,"Hello, Eric." like, "Okay, cool." Like, "I know your name." And it's supposed to be awesome and really cool that I could put your name in the email and then everyone's getting the same products, the same email subject line, the same everything, but I said but I said your name. So, that was that was personalized to now. I think AI empowered personalization in the marketing context recognizes that marketing is a whole host of channels. It's not just that I'm sending you email, but maybe I'm sending

you an SMS, maybe you're on Instagram, maybe you're you're looking on Google, but now I'm really maybe you're on Open AI and you're looking in the AEO answer engine space. And so now I have a proliferation of channels and I have really complicated customers and so many more signals than I had before. So for me, AI empowered personalization now in marketing means that I can focus on the customer and how they want to behave and how they want to be treated. No matter what channel they're coming from, whether they're coming from email or SMS or they're already on my site, I can start making really smart decisions about what message they need, what products they're looking for, what their intent is, maybe where they are in the journey, and I can trigger based on that. Like one of my favorite examples of a customer I'm working with leverages data across the customer behavior across many different ways. So they'll know

you actually as a customer. If you spend more than 2 minutes on my website, I should probably pop up an AI agent. I can tell that for you, this means you're starting to struggle and you're not sure what you're looking for. Whereas for me, it might be 10 minutes. I mean, I love to shop and scroll. Spending time on a website is not an indicator that I'm not going to buy. Um maybe for someone who's more impatient, you need to know like, oh, at 2 minutes I need to start helping them versus someone who maybe responds better over text with a discount offer. And knowing what kind of shopper you are, what kind of persona you are as well as those signals across all the channels, I think is where AI personalization is just taking it to another level. And I guess like as a kind of correlary to that, how much needs to be known about that individual and how much can be inferred from the similarity of their kind of like very early behavioral patterns relative to

historical user? Like do you need to know very specific things about that person or is it really just kind of pattern matching across groups of people that have behaved the same?
Oh, that's a great question because to to date we've really been pattern matching, right? creating these customer segments. And so it's not true onetoone personalization. You're not really identified identifying me as like a cohort of someone who acts in a certain way. Maybe I'm the cohort that's in your loyal customer base and I'm a high spender. Or maybe I'm in your cohort that you've never seen before, new shopper, and I like a certain category or I like to be communicated a certain way on email or SMS. That's really where we were before. Now I would say AI has opened up this complexity that you can see where not only do I need to know a bit about you and I can know you on a customer level and I can start to do

this very specific onetoone personalization, but I can also treat you like the complex person you are. I can treat you like a human and not this fortuneelling pixel gathering kind of um pattern matching, but I can tell that you in this visit you're actually buying a gift for someone. So all those signals maybe, you know, maybe I had a I used to have a quiz on the site and I'd ask you your size and your color preferences and I'd ask you that you're really only interested in shoes instead of tops. But people are pretty complex. We buy gifts. We purchase on behalf of someone else. And then we've got different types of purchases. We've got those impulse purchases because I was in on Instagram last night and you know what? They just got me with the feed and I clicked through and I made an impulse purchase. And that's very different than what to me is a high consideration purchase,

which is very different for everyone. It's based on their budget. It's based on what they do. for someone shoes who's who's a runner, their consideration, high consideration for those shoes. They're going to be running in them day and night. For me, tennis shoes get me to work. Um, not a super high consideration purchase. I did don't need to know the wear on the heel and what kind of stride I have. I don't have a stride, but I need to know. And so, in that, I would say, you know, how much do I need to know? Ideally, I need to know about you, but I need to know what you're doing in the moment. What do you want in the moment? So, no amount of quizzing is going to get you into that. It's going to be really on these realtime behavioral signals that you're picking up on. And so, I think what we're able to do with machine learning now is really interesting to treat people like people and not clicks. And I

think it just unlocks a whole new level of types of of purchase and conversations you can have with customers.
So I think like a kind of a natural kind of followup there is like where do you get the most leverage? So there there are these different customer interaction points that you could apply this to, but my sense is and and like you know a lot of my experience comes from like the mobile gaming space where like the retention curve is just very very severe, right? It's it's just very very punishing.
Yeah. And so all the leverage comes from day one, hour one, minute up to 10, right? Like I mean that's it. That's if you can bend the curve a little bit there. Like that's where all the leverage is versus like in 30 days well that person is going to be retained for forever, right? But like that just the vast majority of people have turned. But but I mean kind of like more in the ecom sphere, right? there are more like win back opportunities and um there are more like upsell opportunities and you're thinking more about like different AOV opportunities over time like what are

the different interaction points where personalization can be applied and like which are the ones that produce the most leverage?
Yeah. Well, I love the way you framed it at the beginning because honestly, I feel like every industry should treat customers that way really kind of it's that f first impression that you're going to give someone. You that should always be your best moment is that first impression. And then definitely you want to treat people great along the way. But if you're not starting off on the right foot, you're not in any industry, it's not going to it's not going to be a great experience. But when I'm talking to customers in in e-commerce, usually the first thing I try to understand is, you know, your leverage point, what leverage are you trying to get today? Are you trying to get topline or bottom line leverage? So, for example, on bottom line leverage cost piece, SMS is really expensive to send compared to like email. And so if I am trying to send out more and more SMS and I've

identified this customer pool that really um does respond to SMS, I want to leverage that AI where I know I'm only sending SMS to the people who matter for that and I'm not just sending it out and wasting a lot of money since since it's really not a a great margin thing for me to be doing. Whereas email is maybe a topline impact or on-site is a topline impact where the more personalized my email can be or my landing page can be the more benefit I'm going to get from a revenue perspective. It's not that expensive for me to be doing relatively speaking. Those people are showing higher intent. So the more I can get them to that purchase intent or that activity that we're looking for for them to take, then the better off I'm going to be. So for me, it's it's kind of like looking at each of those channels and saying, how much personalization benefit

can drive from a revenue or activity perspective versus what's super costly and can I leverage AI to make that more efficient? And then you've got AEO blowing the whole thing up right now as an emerging channel with traffic increasing but conversion iffy at best. No one's really seeing the dollars come from it. And I think that's kind of throwing an experiment wrench into everyone's plans right now.
Yeah. I think with AEO it's interesting because maybe just for people who aren't familiar with the acronym AO I think you're using that is synonymously with GEO. I've heard it be called
Yeah. go a we can't decide as an industry what we want to call it, right? Generative or answer engine optimization.
The idea here is that it's publishing content to be discovered by these chat bots to then be picked up and served in an answer. It's very like essentially similar essentially equivalent to SEO for chatbots.

Exactly. But for perplexity and chat GBT, how do I get my content discovered? How do I get it to convert in the conversation or on my website later? How do I get all that directed over to me? And that's just a big open unknown. I mean, a couple weeks ago, there was instant checkout on Chat GPT. Now there's not.
Now there's not.
Now there's maybe there will be again in the future. We don't know. So, it's this big concern, I think, is people are seeing the SEO drop off and they're assuming that it's kind of cannibalizing for AEO. I think in some cases there's just traffic drop as people spend less right now. But if you assume the one, you're seeing this drop in SEO, this increase in AEO, but I don't know how to get discovered in AEO, and I don't know how to get it to convert yet. And so there'll be a big place to play there I

think with driving driving driving more revenue and traffic once uh once people have it figured out.
Yeah, it's kind of reolent to me of the ASO approach that emerged in the kind of earlier days of the app stores, right? So this app search optimization. It was just just kind of being discoverable.
Yeah. from search and using whatever tactics with kind of manipulating your keywords, your app keywords, your app title, even that sort of improved your chances of being at the top of the search rankings. But the problem with that, there's a couple problems with that. One is that it's not systematic. I mean, you can kind of test stuff and see how it works, but you know, you're very much at the mercy of what people are searching for.
Yeah.
Right. And so, well, you know, if that's cyclical or there was just a short-term trend or something that you benefited from, well, that's going to go away. And and then the other piece of that is ultimately these platforms want to monetize with ads. And so, you know, if they see an opportunity to do that, it's almost like you're a victim of your own

success. If they see that people are are actually deriving a lot of value from ASO or SEO or GEO or AEO, they're going to say, well, I should actually just be selling ads and monetizing that value myself. Right? If you're getting a lot of value from being discovered organically, I should claw that back in the form of ads. And that's you see that with Google search, the entire above the fold is ads. And and now they've, you know, it's it's been replaced mostly with AI overviews and AI mode. You saw that in the app stores themselves. Apple has ads in in the search results and they just inserted another a second placement, right? And and chat has ads. So the thing is like AEO, SEO, ASO, um all those approaches to drive more organic discovery ultimately get co-opted by the platforms themselves because they want to monetize that value.
Yeah, I totally agree and you're right. I mean, I think we we definitely have been talking a lot about using SEO as a indicator of where AEO is going to go. I don't know if that's super accurate. I

think that on some dimension absolutely there were early days of SEO everyone was guessing how do I get crawled? How do I flatten my link graph so that I get crawled for more content? How do I get recognized for that content? Let me enrich the content. And certainly one of the biggest topics I hear from everyone right now is my data is not good enough is you know the blocker I think to most people trying to optimize advertising any of their channels is this conversation that with AI I think we've done a very good job educating everyone that AI is only good as the data you give it and now everyone's turned to well my data is not good enough and as you look at any of the channels. I think the more that the native platforms are retaining the data and hiding that data, the more blind you are to what are people searching for? How do I optimize for that? You know, what data do I need to enrich? And so part of this is really

what channels do you control and how do you how do you bring them all together? AI lets you bring a lot of those data signals together instead of before where it was kind of all separate. Maybe I had an SMS channel, I had an SMS vendor and I could just see that, but I couldn't really tie it anywhere else. I had separate customers that were SMS customers. I had separate customers who were email customers. I had separate customers on the app store and I couldn't bring them together. And so I think as we really start to develop more of this agentto agent profile, more MCP options, we're tying more of this data together at the end I just think will be better for all of us and and our experience than we had before.
Branch is an AI powered MMP that helps growth teams see exactly where to put their dollars to bring users in the door and keep them coming back. If you

haven't had a chance to check them out, now is the perfect time to do it. They've recently updated their look and feel to reflect the innovation that's happening in their product, and they want you to celebrate with them. Head to branch.io/about and sign their virtual guestbook. You'll be entered to win one of a number of prizes, including a $1,000 Southwest travel voucher, a personalized growth strategy session with their consulting services team, or some sweet premium branch swag. Go check it out. That's branch.io/ io/about. So when you're thinking about the kind of you know AI empowered personalization regime, how much of the success that you see with that with with the personalization efforts is determined by the availability of that data and then how much then does that exercise become a function of surfacing more data of actually building the mechanisms to surface more data. So like like a product development or you know a product management actually discipline of saying look when the person hits the

landing page we need to be able to find out these things about them. We've got to create ways to find so like our product management approach should be to try to build mechanisms that they can interface with that actually give us those signals that data to then personalize against how much of this is driven by by product I guess is the question.
Yeah great question. I think in this agenda world 50% of it is going to be driven by product and 50 is going to be impacted by the customers which is kind of cool. the customer's never been in the driver's seat before. And what I mean by that is, you know, we talk, we've talked a lot about needing great data before. And the great data was great. Put a pixel on your website. Um, track all the user clicks. Let me guess if that click is a good or bad click, right? A bounce is a bad click. I don't know. Maybe I'm exploring. Maybe I liked that product um but just not today. I'm filing it away. I don't know. But we're reading the tea leaves. And now that we

have agents and we can have conversations with our customers, the customer gets to tell us what they want even more explicitly. But also, we're getting all kind of signals that we couldn't accommodate before because we didn't have the algorithms that let us handle this unstructured data in as efficient or constructive of a way as we could before. We we didn't know what to do with the conversations. We couldn't really structure the data and read the patterns like we can now. So now customers can tell us so much more, right? I couldn't other than maybe putting a search query and two or three words into a search box. I couldn't have a conversation back and forth and tell it, "No, that's not what I meant. Please give me better results than you just gave me." So, I can actually feed in and give you feedback and tell you, "No, I really was interested in this pair of shoes, but you didn't tell me how much

they weigh, and I really need to know how much they weigh before I'm going to buy them because I'm going on a hike. I don't want 5 pound boots." Now, I can tell you that that's a data signal I maybe didn't know I needed before. It's a a data attribute that maybe I didn't know I needed before because I never gave it to the customer. So, I feel like part of this is being very thoughtful with your product on what data signals you get, how do you acquire them, how do you open it up for more feedback, and now more of it gets to go into the customer's hands and tell us what they really want, what were they interested in.
It sounds like what you're saying is there are these mechanisms that, you know, that we can give to the customers to sort of let them express themselves. I'm thinking more like in terms of the customer journey, right? You kind of brought up the example before of like someone is on the landing page and just a prompt to expose a a chatbot to them. That's just a way to collect signal at that point, right? I mean, they're giving me basically describing what they

want and then in that way I can that's that's something I can use I can tether to to to personalize the experience downstream for them.
Exactly. Maybe before you just see them clicking all over your website looking at different skirts or dresses or whatever and now if I can surface up that conversational agent or I can have Ask AI ask me anything available all the time then they can tell me well I'm looking for something to wear to Wimbledon that's very different and now I know kind of not only types of products but I know the style and what I should be recommending to you where Before I was just kind of guessing, looks like you're looking at a lot of white skirts. Do you want more white skirts? Right. What do I need to Now I can tell you as a customer. Yeah. Like who's got on their website a specific category? I haven't seen it. That was like going to the Taylor Swift concert. Need an outfit. Maybe I need like a

sparkly blazer for my husband. That's not a category. But now I can have a conversation and be more explicit and and have the have the site help me find what I'm looking for.
Yeah. And I mean, not, you know, not to mention this, a Wimbledon attendee is probably like a dream scenario for an ecom retailer.
They're probably working with a a personal shopper and they're not that, you know,
high propensity to spend there. Um,
we can all aspire,
right? You know, that's not just a potential customer just hit the landing page proposition. That's kind of a throughout the life cycle that can be the case, right?
Absolutely. I mean, maybe I was looking at Wimbledon outfits before, but now maybe we're getting closer to the holidays and I'm looking for gifting items and I just really don't know what to get my nephew. We saw a lot of that actually this holiday where engagement would start to spike with conversational because people would come to different websites and they're not searching for themselves anymore and they're searching

for someone they really care about but maybe they don't know right little kids preferences change all the time so they're like only what should I be buying for my seven-year-old nephew he likes video games maybe as a shopper I don't know much about video games and now I can be more thoughtful about what I'm giving So much more um those journeys and those purchase cycles totally change. Just because I nailed you in the first 5 minutes and knew what you were looking for doesn't mean you're coming back looking for the same stuff.
Have you seen cases where customers reacted poorly to personalization?
You know, I think that's a really interesting one because I mean I think we all see it right now where we think it's we think it's personalization, we think it's awful, right? this morning. Just this morning, I got the email from a e-commerce company saying, you know, still interested in this. And it wasn't anything I'd ever looked at before. And then there were like a whole bunch of products underneath the bed. And I was like, I'm I'm not a six foot man. Like,

what are these products? I I've never bought anything bright red in my life. Why are all kind of wrong at the same time? I mean, where's the harm? Not really going to be bothering me. But I was like, this is weird. I think that most people are so shy about personalization that we haven't really gone to the explicit version of personalization. We haven't gone really deep. In fact, the balance I see right now is people are so cautious to be overly personalized. I think a lot of money is being left on the table because you're reading the signals and saying, "I'm just going to make a really small change for a really small amount of people." You end up in this AI is meant for us to scale, do personalization at scale, and I'm going to hold it back to a very small amount of scale and see if it's okay. See if

everyone reacts okay to it. I think that consumer behavior is changing. you can see people having more conversations being okay with their data conversations being used to to inform further further conversations. So I I'm curious where we're going to go here on if we're going to become more and more explicit. And that's when I think personalization has a a chance of of going wrong is when you start to get really explicit and you try to be very definitive about an answer and you get it wrong, which is why I think, you know, AI is so good when it's there's no one right answer. So, I can kind of get it right.
So, we'll see. I think people have been pretty cautious to date on personalization. So even when it's wrong, not that bad.
Yeah. I mean, I think I mean, my position on this is is is that first of all, like there's a lot of cases like the one you just cited where, you know, you get the email and it's for something

that you never engaged with or interacted with and you just dismiss it and don't think about it again. But then you might get the email for something that you did browse or whatever and and that feels like personalization. But again, it could just be totally random. I mean because it's just this the things that are irrelevant you just ignore, right? And that's where I think a lot of the confusion arises around people thinking their phone is listening to them. I mean that that truly is just ignoring the the 99 or whatever the 97% of ads that you see that are totally irrelevant and then anchoring to because you recently talked about taking a cruise on Disney. Anchoring to the Disney cruise ads like oh they must be listening because I was talking about that today. Well, that only resonates because a you were talking about it today, so it's top of mind and b because it's it's not noteworthy for all the ads you saw today and there were thousands that just were totally irrelevant, right? And so it's you're just ignoring the totality here. You know, the other thing is I I do think on that I mean people they they truly do appreciate a personalized experience that's more

relevant to them and and it's it's really just more about the context being appropriate or not. I mean, per personalize some, you know, ecom product that I would like to buy, a retail product on Instagram, and people love it. But, you know, if you chose the wrong content, if you chose an unfortunate context to do that, then yeah, people would bristle at it.
You're totally right. I mean, context is super important. If this were my bank, they better get it right. If this were healthcare and I'm going into my doctor, don't give me the don't give me the wrong medication or the wrong diagnosis. Right? You have to be absolutely precise that in this area the harm of personalization or I I got it wrong. The the the risk of getting it wrong is less. So it's it's going to be harder for me to have an adverse outcome or offend someone. It's possible. Um but I think we're we try to be really thoughtful and not as explicit in in those areas. But there are industries where it 100% matters. Um, and you can't

just have no one right answer or I kind of got it right. You have to be perfect,
right? And and that's nuance that needs to be recognized. I mean, there there just there, like you said, I mean, there's just there's there's circumstances where there's no room for regret.
Yeah.
And yeah, that just that requires a lot of thoughtfulness and care. And then there's there's circumstances where you can be experimental and actually there's just very little downside to getting something wrong. and and and there I mean I think in those circumstances or in those contexts people are just they're very forgiving of a misfire. Right. Actually I don't I don't I don't want a baseball cap. I don't wear baseball caps. You shouldn't have recommended that to me but at the end of the day I mean that didn't really derail my life to see that. Right.
Yeah. Exactly. You can recommend baseball hats to me all day. That's typically what I live in. soon.
That LTV model you're using to optimize your campaigns, even the most accurate one in the world won't guarantee better performance. Google and Meta run on signals. Developing and deploying them

at the right time in the right format for every platform. That's its own discipline, one that has to learn alongside the auction, not behind it. Openoro, SEM Rush, Miro, they all work with Voyantis for exactly this reason. Because the model was never the hard part. It's everything that comes after. Signal engineering, testing, maintenance. If you're ready to close that gap, visit buoyantis.ai. That's v o y a n tis.ai. How have you seen clients implement conversational shopping? Are there specific verticals for for which that works best? Just having that chatbot and actually talking through the purchase. H yes I this is like one of my favorite topics because I think for the first time in like 26 years e-commerce is getting a complete revamp and there's so much talk about having conversational and now it is all of e-commerce going to

go to conversational and I think that aside from verticals I mean there are certain kind of attributes to a vertical that makes it great for conversational shopping I mean if you've got a very large assortment If you're very particular, maybe it's B2B and you really need guided selling because customers aren't exactly sure the they know the make, model in their car, but they don't know the exact part that goes with that. You have a higher chance that the conversation is going to be something you want on your website. But really, I see it coming down to the purchase interaction and the customer journey. And so, first, like, not every purchase requires a conversation. If I went into the grocery store to go get milk and the people at the front stopped me and said, "Can I help you with something today? You know, we have something on sale." I'm going to be like, "Oh my gosh, I'm out of milk and I this was a 30-se secondond purchase, right? It's that Instagram impulse purchase. Don't don't interrupt my flow

because if you do, I'm likely just not to make this purchase." But there are a lot of situations where it becomes high consideration or I want to I want to learn more. I want to have a conversation. And websites are really limited real estate. No one's going to scroll. Like even if you think I've got infinite scroll on a category page or I've got a product page and you can scroll down those Amazon product pages are I don't know how many rail pages long, right? They go on forever. No one's scrolling. You really do have limited real estate, but a conversation opens it up to infinite real estate. Now, I can ask you anything. I can find out about reviews without reading the reviews. I can find out about this the weight of those shoes, even though you don't actually have that on the site anywhere. I can find out the fit of the product. Is this shirt does it is it tight or is it loose

fit shirt? Is it cropped? Uh maybe I can't tell from the photo. So I can kind of open up all of those conversations and that data and a really limited amount of real estate. So I think from a vertical perspective, I think a lot less about oh this is travel or this is an apparel company. I don't know if this is applicable because every user's definition of high consideration is so different. what's high consideration to me and my budget and what I love or I have sensitive skin so I'm way more interested in the details on cosmetics than the next person it really changes the conversation and means that I think every vertical has the potential for I need to engage with some subset of my users in a different way
here we're talking about like on the on site right so they're on the retailer

site you know and this is like the opportunity to sort of engage with, you know, a chatbot that kind of functions as like a personal assistant. You could think we're talking in the retail kind of e-commerce. Let's talk about like let's just narrow it down to like the uh clothing, right? So, I'm on an ecom
retailer's website and there's this opportunity to engage with a chatbot and and use that at like a personal assistant. Like if I'm thinking about it from the retailer standpoint, there's obviously probably improving conversion because more sort of readily route them to the thing that they're, you know, likely to want to buy. It's there's not much discovery needed. They don't need to click around and they might not they're less likely to get frustrated and just leave because I'll just lead them right there. But there's probably other stuff too. It's I mean, you know, you think about like maybe it's finding more accessories like other things that they also are interested in as a result of like kind of getting this information. It's not just a matter of more efficiently leading to them to that thing and getting them to the destination, but probably in enriching the shopping experience, giving them exposure other things maybe they didn't even know they were interested in.
Yeah, I think you know the the two

surprises I've seen the most when people have put conversational shopping on their website is number one is returns. And so your purchase confidence as a consumer goes up much higher when you can ask and engage more questions. And so before when I'd buy those shoes and then I'd get them and I'd find out, man, okay, turns out they run large and they weigh a lot. And if I went back to the website and I really digested all the information, turns out they told me that it runs large at the bottom of the page in a review that I didn't read. Um, but maybe they didn't tell me that it it weighs 5 lbs. And so there's this returns aspect where now I'm taking people on the product page and I'm increasing their confidence to either buy or not buy. In some cases that conversion rate not going up can also be a healthy thing in conversations is that I can actually help users at that moment

and understand kind of which product they're looking to get. The second I would say is this notion of voice of customer. There's really no place on your website that you can converse with a customer. So, finding out that your customers really cared about the fit of the shirt and what size is the model wearing in the photo or is this product easy to assemble? Well, I had a bunch of assembly information, but I didn't realize like I needed that particular being able to have all these conversations and see that your customers, they're clueing you in. And what we what we see over and over again is these trends where you look and you say, "Wow, there are like really five things customers really wanted to know and they can't find it on your website." Now, they're getting they're getting that actual voice instead of maybe two words in the in the search bar. Yes, I knew everyone wanted black dresses. I

didn't realize that what they wanted to know was the the hem and the length of that dress and that that was a really really big deal to 80% of my customers. You know those channels your colleagues keep bragging about, the ones getting all the credit? Yeah, they might be doing squat. Attribution makes every channel look like a hero, even when it's a zero. Incremental tells you who's actually doing the work. It's like a lie detector for your marketing budget. Start using Incremental today. Get your demo at incremental.com. That's ncrmn.com. Mention that you came through the mobile dev memo podcast for a special 15% discount for the first 6 months. I want to switch gears to the advertising use case. So, how can advertisers utilize AI based personalization tools outside of what's available on the actual ad platforms themselves? What do they have available to them to improve their advertising outcomes?

The hardest part about these large platforms really is how do I how do I obviously get the best return, but how do I deal with the fact that on any of these platforms, they may or may not share data with me. It's siloed data. And so as much as I can get data from these different platforms and channels, at least what I can do is be leveraging the AI to create better audience strategies. Whether that's unique to particular platforms and recognize that certain platforms have different audiences, but creating that better audience strategy per platform personalization, personalizing the actual destination from those platforms, making sure that my messaging and targeting and creative are actually personalized and correct for that audience or that person. and making sure that I've got a really strong feedback loop on that. So now I

can do this more in real time behavior or real time testing and really get quickly to which audience strategy, which content strategy is actually winning on which channels and optimize much quicker where I think before we would spend maybe a month of data and analysis and then say okay turns out the AB test that we just run on personalization B1. Now we can know within a day and be optimizing. That's a lot of money that you're leaving on the table if you're running these AB tests or control tests for a long period of time.
Is there a risk that you know any any approach like that where you're analyzing the data set on your own and then and then using that to adjust the the platform levers the platform settings. Is there any any risk that that that those sort of are in congruent or like because a lot of the the platform is well we see now this bifurcation of you the very large platforms that that all run these totally automated

systems people call them black boxes right so you've got you know Facebook and Google were kind of like early early leaders there even you've got Reddit Pinterest Tik Tok they've all got essentially like endto-end automation and you know I think with most platforms following suit is there any risk that what an advertiser is doing on their own then kind of works against or undermines behind what the what the platform is doing in this kind of blackbox environment.
Yeah, I think there are there are two big risks there. You bring up a great point because your native platform does have configurations, optimization, different data, very mismatched audience logic, maybe audience definition, um different goals that they're trying to accomplish and optimize for, right? All of these algorithms are giant math formulas with different coefficients or different goal setting that they have built in. And so, you know, I think this is the world we live in today, which is why customers feel like it's such a disongruent experience based on whether they're on

your site and went to your brand in particular or whether they're on one of these native platforms. I think you've got a lot of that off-platform personalization not necessarily being complemented by the native tools. And then I think you've got the other things with these native platforms is you've got usergenerated content. The userenerated content is opinion based, not always factual. At the moment, it's what someone thought the answer was, but it wasn't. No ill intent. They just got it wrong. for it was their experience and that experience isn't relevant for me, it's relevant for someone else. And so you do get these conflicting signals. I think right now people generally understand that. But but you're right, if you're building these strategies, you need to be very thoughtful about how much weight you give a signal that you're not confident in or if you use it at all and what do

you want to use from those platforms? What are some cutting edge applications of AI based personalization? So, you know, what is what is Bloom Reach doing? What are you know, what are your clients asking you for? What are your clients having the most success with? Like what's what's coming down the pike? What's kind of like the most exciting kind of frontier here in AI based personalization?
I think we're starting to see some of them. They're just not as prevalent. Like I'm I'm super excited about the conversational shopping agents. You see it with Roffus. I love the magic apron at Home Depot because it's just not my um I'm not an expert there. So, it's very helpful. But I think what next really come I'm very excited about the the agentto agent experience. The fact that we can really start scaling I think in in marketing in particular right now you're kind of limited by how many campaigns you can come up with and run. But the fact that we're moving to agentic campaigns, the fact that I can

do not only more campaigns than I could do before at micro segments instead of broad segments, but I can also feed in signals that I couldn't feed in before. And the AI will optimize for those outcomes. just means that I think our customers are getting a better experience with those campaigns, but I'm going to have way better outcomes than I could have with the fact that I don't think anyone ever feels like their marketing budget is enough, right? We never sit here and say, you know, I have enough people and enough time and now there's going to be more throughput on that there. So, I think, you know, really the kind of the agentic automated marketing, the campaigns that we're going to be able to come up with, it's one of the things I'm I'm most excited about.
Have you seen anything that surprised you, like things that your clients are doing or or just applications of this that had, you know, dramatically better results than you would have expected? Just just something that that happened that kind of took you by surprise.
You know, I guess I thought, you know, I've got one customer I work with who

does a lot of the Funko Pop products. They do a lot of gaming products and really started experimenting with these agentic campaigns and I thought that I thought they'd get a little bit of a pop from it. You know, I hoped right that we would get a little bit of poverty. But the thing that I didn't factor in, the human element of it is when you get into certain of these categories like gaming or like pop culture, your fan base is so important and so bought in that getting that right drives so much more of an impact than if you kind of got it right or if you threw them into the population of general. I mean, I guess that, you know, there there are and there are certain universes that don't play well together. I mean, if you like DC, but then you I don't even know the superheroes, so I don't even want to attempt to this analogy, but I guess they're not all the same. And um you don't want to get that wrong with your

customer base. And so, they really leveraged AI, got it right, and that the pop was a lot bigger than I thought it would be.
That's actually a really interesting thought. It's kind of like, you know, AI, a lot of these AI tools allow you to to go beyond the paro rule, right? I mean, you know, you can actually say, well, no, I've got I mean, it's it's it's relatively costless for me to actually take this to 100%. Versus before, I was like, look, the trade-off is you kind of run into a a cliff at at the 8020 threshold. It's not really worth going the extra mile. But when you've got kind of a fan base of certain categories, that's fanatical. going to you know taking that to the extreme actually does sort of really unlock all the value and it's almost like an inverse brado rule right for some of these groups and
it's the long tail right there's always been this holy grail of like h I wish I could optimize the long tail but as you mentioned like that extra 10% it's going to be monumental effort so forget about

the long tail I'm just going to 8020 it and focus on the top 20 get 80% of the value but now I can go after the longtail. And even though each one of those is maybe pennies,
right,
in aggregate, it's a huge impact
and unlocks all the value for those fanatical, you know, fans, but but who are probably have like a higher willingness to pay and, you know, and and that creates a pretty uh compelling opportunity. That's that's I never really thought about that. That's a really interesting idea. Uh Christine, uh this was great. Appreciate you coming on the podcast. How can people learn more about Bloomreach? How can people uh engage uh with your company?
Yeah, absolutely. I mean, we're at bloomra.com. So, uh, you know, follow us on LinkedIn, come visit our website. Be happy to chat with you.
Cheers. Well, thank you so much for your time today.
Thanks for having me.