Retail Media Vibes
Retail Media Vibes is your marketing lens on the world of shopping, commerce, and culture. Each episode brings fresh conversations with industry insiders who break down the stories driving how brands reach shoppers and how shoppers connect with brands. From big retail moments to the latest shifts in digital media, it’s your front-row seat to the strategies shaping the future of commerce.
We keep it smart, energetic, and actionable, mixing sharp analysis with good vibes so you walk away informed and inspired. Whether you’re a retail pro, a marketer looking for an edge, or simply curious about where the industry is headed, this podcast is made for you.
Join the vibe and stay ahead in retail media.
Retail Media Vibes
More Data, Better Decisions? Not Automatically
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Data overload is paralyzing brand strategy instead of improving it. With retail media networks expanding rapidly, the ability to turn massive amounts of shopper information into actionable decisions is the dividing line between market growth and wasted ad spend. Laura Weiderhaft, Director of Product at Crisp, joins the show to unpack how brands can stop hoarding metrics and start building scalable intelligence.
We sit down to dismantle the current state of retail media measurement and how teams are actually utilizing shopper data. We cover the danger of cherry-picking analytics, the OMG (Objective, Measure, Goal) framework for promotion planning, and how to properly flight budgets by Walmart week based on geographic sensitivity. Laura also reveals a highly effective philosophy on automation, explaining why marketers need to approach AI like software engineers by organizing tasks into specific "jobs to be done" rather than treating the technology like a basic search engine.
Managing all of this information is incredibly tedious, and the messy reality is that most teams aren't willing to be held accountable for the bad decisions currently hiding in their reporting. Listeners will walk away with a clear understanding of how to automate mundane data entry, the importance of aligning commercial and brand teams on a single source of truth, and why tracking secondary indicators is vital for long-term category share.
If you care about retail media execution, competitive promotion planning, and operationalizing AI, you’ll get a lot from this conversation. Please remember to subscribe and share this episode with your team to help us continue bringing practical insights to the forefront. What is the most frustrating, manual data task you are ready to hand over to an AI workflow?
0:00 Intro & Laura's E-Commerce Background
3:53 Navigating Shopper Data and Retail Media
20:47 The OMG Framework for Measurement
41:25 Building AI Workflows Like a Software Engineer
56:41 Bold Vibes & Closing Thoughts
What's up, party people? BV here, and welcome to another episode of Retail Media Vibes, a doing business in Bendville podcast. We are recording live in Rogers at Podcast Video Studios. I am your host, BV, and today our guest is Laura Widerhaff of uh director of product at Crisp. So, our topics today, we're going to be talking about a few topics that we've touched on before, but uh everybody wants to know a little bit more about, and that's about shopper data measurement. And of course, we got to talk a little bit about AI. So, with all that out of the way, let's get to meeting Laura. Welcome to the show, Laura. It's good to have you here today.
SPEAKER_01Thanks for having me.
SPEAKER_00Yeah. Um, so let's get to know you a little bit for our for our uh for our audience today. And so give us the quick version of your story and how you got to where you are today.
SPEAKER_01Sure thing. Yeah. So just been in the uh Walmart data space for the last 10 years, kind of got my feet wet at a small automotive supplier, four items, uh, about 2,000 Walmart stores. Um, but because it was so small, uh, was able to wear a lot of hats. So they were like, are you interested in learning about our shipping? Yeah. Are you interested in learning about marketing? And I ended up doing um our direct to consumer website as well as helping with um our Walmart e-commerce, our product pages, descriptions, and um all the way through Foucoma. So I was able to learn a lot about um like how things work, both direct to consumer, um, both with like targeting new audiences, um, and then also of course with uh how to work with Walmart. So after that, started working in Walmart e-commerce consulting, um, did that for about four years and really was able to work with a lot of um brands that were kind of leading the pack in terms of what they were trying to do with Walmart e-commerce. So um in that role, started out doing consulting, but realized like you couldn't really scale, um, especially with like data practices. And so I kind of shifted over into products so I could scale out the types of data that we were looking at to make better and more informed decisions for brands. And so that's how I kind of evolved into a product role because I was always a data nerd. Um I studied e economics at the University of Arkansas, and so yeah, have always just like had a quantitative background focused on data and focused on scaling and like how you can uh really think about decisions based on data.
SPEAKER_00Yeah, yeah, no, that's that's great. Um, all right, quick, quick icebreaker question for you, or maybe two, but so if you had to delete all of your apps on your phone except for one, what app would that be?
SPEAKER_01Um, I think Reddit. So years ago. And this is um, I'm gonna throw some shade at Google. Like, I just think their search results have been deteriorating over time. There's people have been like trying to game the system with a feel and kind of like slop content even before AI. Uh so like I kind of switched to Reddit as my primary uh search engine, and I like will actually search for things there before I search for Google. So I'm gonna keep Reddit. Um, and actually they have a really interesting advertising strategy too. Also a whole other conversation.
SPEAKER_00Yeah, I've actually been really interested in what you know, brands, especially like CPG brands and their opportunity around Reddit because Reddit has, you know, has had a lot of growth over the last you know few years. Um, and so it's become a much more viable platform and uh greater reach than they've ever had before.
SPEAKER_01When you talk about relevance, they have extremely niche audiences, and you do have to have ads that are uh tailored specifically to those um separates.
SPEAKER_00All right, Laura, we're gonna get now into our main topic. So, you know, we all understand that data is being collected all the time, everywhere, right? So many things that we do online, even some stuff that we do offline, you know, is leads to some form of data. So we now have more data than ever, right? And you know, the question is, you know, are we really getting better, you know, at making decisions now that we have all of this data? And then there's probably some data we desire that is not available, uh available to us. So, you know, what where where have you seen like the biggest growth, you know, that when it relates to a brand from a retail media data standpoint, it's really, you know, that brands are very interested in.
SPEAKER_01Yeah, it's it's it's interesting because I think some of the most effective advertisers outside of retail specific media, like uh and I guess they're more retail focused now, but Google and Facebook, right? Very effective because of the data collection that they've been able to manage. Um, but a lot of it's kind of a black box, like you can target specific audiences, but the mechanisms behind the scenes aren't really clear. So it's hard to kind of see like what this audience is doing, are they changing over time? Um, the thing about these retail media networks and um like the shopper data that's being collected is that these retailers really see brands as partners. Like it's not just your spending dollars, right? You're trying to actually improve the shopper experience. Yeah. So uh that is a really interesting space. And I think that's part of why Amazon and and Walmart uh and Kroger now are kind of leading the pack on um providing uh really rich shopper data. Um, but they want brands to be thinking not only about like how are they spending their dollars well, um, but like what matters to the shopper. Yeah. How do you get closer to the shopper? How do you understand the shopper better? Because um assortment is important and you can't execute retail.
SPEAKER_00Innovation's important innovation, yeah.
SPEAKER_01Yeah, exactly. Um, pack size important, price really important. Um and what you uh execute advertising on has to have those additional considerations as well. So um I'm really excited by like what you can do. And I actually think most brands should be thinking about the shopper first and thinking about their retail media as an extension, yeah. Like how they're thinking about um what the shopper at Walmart wants. Um, Walmart's growing market share too. So they have a lot of new shoppers coming into the fold. And so um there's lots of opportunities for neuter brands um or to drive loyalty with um shoppers that me have come from a different retailer.
SPEAKER_00Yeah. Yeah, you know, the the whole data collection process is a fascinating one to me. And, you know, I've I've usually have you know held up um you know retailers like Kroger as you know one of the uh maybe best in class, you know, it's hard, you know, best in class is always very subjective. But you know, they Kroger has always had their rewards card. You know, you have to scan your rewards card in order to get rewards. So many people used it, they collected a lot of collected a lot of data, was able to use that. You know, Walmart has had to take a different approach because, you know, until recently, they didn't have a membership type of program where you had to pre-register a lot of the data or a lot of the shoppers use cash and and don't use traceable tender in order to to to collect that data. Sam's Club, obviously, on the other hand, has always been member-based data. So, you know, it's it's just interesting. Now we have access to all this data. And I think you hit on a couple of things that I think is important in what you use the data for. But it feels like now that we we have all this data, I don't know, I don't know if we're getting to better insights or not. It feels like insights have kind of been degraded over time and then the the data is being used for very finite use cases. So I I love the idea that you had, or not necessarily idea, but the comment you made about using the data for more than just tracking baskets and shopping and that, but use it for insights.
SPEAKER_01Yeah, I one of the things that I think is really important when you're trying to figure out like how do we leverage data is that like I think where you uh you can get into like a data overload um pretty easily. And I think what has to happen is you have to be really clear about what your goals are. But like you also have to think about like, okay, what's the right path to doing that? Is it better shelf execution? Are we gonna be out of stock loss? Do we need to invest more supply chain? Um, is that better uh more targeted resale media execution? Um, and if it's the shelf execution piece, like you should have certain KPIs that are really important to track. And those are different than if actually you want to grow your business at Walmart through trade-up. Um, or if you want to grow your business at Walmart through assortment expansion um or item innovation. So you kind of have to think like what's your goal? Like obviously market share, but what's what are those secondary goals and what's important for us to to track over time? Um, and like what are secondary metrics or leading indicators of success too? Like a lot of times uh like it's really hard to get an insight, especially on things like trade-up or ship switching or basket and how those are evolving unless you're looking back on a long window. So it's really important also to think like about leading indicators, what starts to tell you things are moving in the right direction. So like early kinds of signals.
SPEAKER_00And retail media is very good at that because you get click and conversion and those are good ways to But not all forms of retail media, and that's the pro one of the challenges, is that you know, retail media has expanded its definition, right? And so it it's it it's almost anything that ties back to a retailer's now included in retail media, but it all doesn't have the same data and you can't use it all the same way, right? And so then you have some gaps and you have some holes.
SPEAKER_01And that a CTV, for example, kind of um and the audio formats and those types of channels are also those are also data collection opportunities, too.
SPEAKER_00Um well that's why Walmart acquired Visio, right? I mean, yeah, it's great to have a a TV, you know, a TV company under your umbrella, but it's the data that's people who signed up, you know, to uh through smart TVs. There's so many people, obviously, when you when you set up a smart TV, you have to register. It sees what you watch, you know, it and so and as people are doing more things on their TVs than they were doing before, I think that also gives some indications into some insights or at least some data collection that could be turned into insights that could be used uh used broadly as well.
SPEAKER_01Yeah, the the growth of Love Island is interesting to me too, because that's always been kind of a commerce-focused show.
SPEAKER_02Yeah.
SPEAKER_01Um, and so like Walmart and other, I think, Amazon, of course, are understanding like, oh hey, like people are seeing outfits on these shows and they're wanting to buy them. Um L'Oreal's made a obviously a big investment in that show and sponsors to like games and things. And like they they mentioned products on the show. So but that is a data collection opportunity. Like, do those things like are you are the shows you're watching gonna like be an opportunity for consideration or discovery, right? A product and like if it is an opportunity for consideration, you want to be able to execute on that um transaction. So those things need to be in stock, they need to be available to the the shopper.
SPEAKER_00Yeah. Is there is there a a piece of data or collection of data that brands typically ask for that may be hard to get, um, in your in in your view?
SPEAKER_01Yeah, the shopper data um for Walmart in particular um is quite difficult to work with. Um uh they because they provide so much that's rich, but it doesn't tie back to the individual shopper. But you can look at demographic patterns, you can look at baskets, you can look at switching, yeah, loyalty. Um, but it's hard to do so at scale. Like you mentioned this earlier, like it you it's very effective. It's like I'm looking at a couple of items. I'm looking at a set of competitors and similar pack types from my direct competitor, and I'm able to compare those and what's happening. Um, but if you're trying to scale, right there, you know, there's item limits you can all grab 50,000 items out at a time. Um, but that's where um you have to be creative about using what is scalable. Um, and so I think like shopper insights um from like the API feed where you can start to look at data at scale can actually be really valuable. I I'll share an example of something I did the other day with one of our um customers, which is we wanted to do promotion planning with the shopper data. So you can look at a history, uh shopper data at scale is every channel at every store um for every item in the competitive set that you play in. I think it is at a subcategory level. So you get pretty rich data. And Walmart says um that about 80 it tracks to about 80% of transactions, which gets to your point. I think the other 20% is presumably cash.
SPEAKER_02Yeah.
SPEAKER_01Um so you get about an 80% kind of like directional view of what's happening, but you can start to unpack things like when competitors change their prices. So looking at similar items, yeah, and when competitors um like moved to price, what was the price sensitivity at different stores? You can look at every store and how sensitive shoppers were to the price. So you can start to think about like where do you have diminishing marginal returns? Like, what's the correct price discount? Sometimes a 10% discount um might not be the most optimal discount one, right? So you can actually get to really great insights out of like the shopper data that you can scale because it has that great store level granular. And so being able to plan a promotion, understand what the correct promotion is, understand which stores are more sensitive, which means like if you do execu like want to execute the promotion, you can like plan ahead from a production standpoint how much product do we actually need to support the unit growth that we're gonna see with this promotion. Um, so there's lots of things you can do. Uh, you know, new item launch is another good opportunity because a lot of the shopper data is so rich. Like if you're say like a protein powder and you want to um release a new flavor, like strawberries and cream, um, maybe you can look at other items uh like similar, yeah. Yeah, and um come to Walmart with a recommendation about what stores you think this new item launch is likely to succeed because of the shopper data. And Walmart gives you like the long descriptions and the the real problem is attribution, right? So like how do you uh uh segment all of that data effectively? Right. Um AI is very helpful and very good at uh product segmenting. All those topics we're gonna talk about today, so that's great. Yeah, but there's lots of things you can do with it. And I think people are just kind of scratching the surface. And I think people lean a lot into baskets and switching, but there's a lot of if you do like even just the sales indicators correctly, there's a lot you can do um with that data.
SPEAKER_00So we're like what were the most common mistakes made with data from in from a brand perspective?
SPEAKER_01I would say um not being accountable to bad decisions. So like there's a lot of storytelling around data and when things are succeeding.
SPEAKER_00So like we much easier to tell a good story, Laura. Nobody wants to tell the bad. Exactly.
SPEAKER_01So like I would say like if anyone's cherry picking data and not being honest about like what's not working as much as they are honest about what is working. Um and I see I see it increasingly. I think brands are evolving in such a way that like they want to understand like what's not working for the shopper so that they can show up better and um for both the retailers that they're working with and yeah, um, and for their shoppers. But yeah, I I think um cherry picking is I think especially, and I see this in you know, like you have to be work with your partners and um retail media partners and be very firm about like we need to see the things that aren't working as much as we need to see the successes. So you have to be very direct about like what your reporting needs to look like and like how you get an honest view of what's really happening.
SPEAKER_00Yeah, yeah, yeah. That makes I mean that makes a a lot of sense. Um so if you were gonna give you know three questions every marketer should ask before getting, you know, before using data or three tips, what would what would those be when working with data?
SPEAKER_01Yeah, I would say um the first question would be like do you understand the goals of your business? Like are you like are what is what you are trying to do aligned with like your commercial team?
SPEAKER_02Right.
SPEAKER_01Your brand team. Yeah. Like can you get on the same page about um like what the goals of your business are? Right. The second thing I would say is um like how do we how do we be as fluid as possible? Like how do we really use data to drive decisions? How do we not lock in? Which is difficult when you're thinking about budgets. Yeah. How do you give yourself as much wiggle room to actually respond to data as possible?
SPEAKER_00So um Yeah, there's a weird cadence there too between what do you react to and when? All right. Yeah. You know, do you react to data that, you know, real time? Are you working with an agency and they're, you know, you're you meet with them once a week or you know, and it's it's there's always timing is always a big challenge, I think. You know, timing of the data to when you get the data, yeah. Plus then how do you how and when do you act upon the on the data.
SPEAKER_01Yeah, like how long do you give a test and learn before you decide whether it's effective or not? Yeah. Um, I I think that's hard to answer, and it depends on like what vertical you're in. Obviously, like velocities vary, and um paper towels are a very different like product than um say like snacks. Yeah. So uh I think the time varies depending on kind of like um how much influence uh or how spread out the shopper is. Like paper towels, you there's not a lot of queries that people are searching. They're just they're searching for paper towels. With snacks, there's you know, probably 50 different queries you could like start thinking about if you're if you're thinking about both on-site and um like where you're targeting.
SPEAKER_00Yeah.
SPEAKER_01So um yeah, uh, there's no right answer there, but uh I think you do want to make sure you have enough like um wiggle room in your decisions that you can respond to data, like that you're not locked into a specific strategy that's not working.
SPEAKER_00Yeah. Yeah, one of the things I learned long ago, and I'm not saying I'm very good at it, but what I learned long ago is that like you can have the data, but you also have to be able to tell a story with the data, right? You know, just taking the data at the number, you know, whatever the number is, and just trying to figure it out and then figure out what it actually means and what impact it has on your business and all that, right? So right now, now we're trying to, we're actually bridging into measurement. So we're gonna get into that topic here uh now. But you know, it it does telling that story around the data uh takes definitely a certain level of skill um and be able to, you know, fine-tune like you know what data is most important to you.
SPEAKER_01Yeah, uh it's not a one-person job.
SPEAKER_00Like yeah, right.
SPEAKER_01I mean, like you need all the team like bought in to what you're trying to do. You need your if and if you're focused on resale media in particular, you need what your partners you're working with, who's executing your ads but then they need to be bought in the thing.
SPEAKER_02Yeah.
SPEAKER_01So like how do you kind of like in order to tell a good story, have everyone aligned on what their role is and how they're helping support telling that story.
SPEAKER_00Yeah. Um yeah, so let's let's talk a bit about measurement. So I've always you know thought about measurements, you know, from the standpoint I I use the acronym in my head OMG. So objective, measure, and goal, right? So, you know, what's your objective? And you hit on that when we were talking about the data piece, like being clear on what the objective is, right? And there may be, you know, maybe it'd be a business objective, there may be a campaign objective, and then there may be a media objective, right? And those all deserve a certain type of measurement or series of measures to determine what how you're going to figure out if you've yet met your objective. And then there's the goal. So each of these have their own goal, you know, whether it's a certain, you know, uh new to brand uh measurement or whether it's you know, you have your benchmarks for your click-through rates and you know, or um, you know, sales velocity, whatever, whatever it is, you have your goals. You have these, you know, these three things that work work together. I don't think we always go through the discipline of being clear on what is your objective in uh down the line. It's like, well, we're just gonna do what we did last time. Okay, well, a lot of things have changed since then. So, you know, from your standpoint, you know, how did how should measurement work, right? So you know, it it it does seem size sometimes like measurement is figured out after the fact after you ran the campaign instead of up front, and not enough time is spent on that.
SPEAKER_01Yeah, uh I totally agree. And I actually think um there are I think AI can be very helpful at letting you um get more green large so you can have an overall objective, but I actually think you should be pretty um targeted across like what you're trying to execute. So like even like different items should be promoted differently and have different measurements. Yeah. Um, but in order to do something like that, you really do have to set um up how you want to measure um ahead of time. But so like I think for a brand, um, like if you're trying to target branded keywords, um, like probably what you want to do is promote items that allow shoppers to trade up and not just like hit your opening price point. So like you would want to track a trade up uh on your campaigns that target branded keywords, right? Um, but if you say are um trying to grow like new to brand and we're in a big um big price sensitive environment now, um you in those campaigns you're gonna want to like promote your opening price point items and uh like track new to brand, track switching, um probably try to align the media like in-store promotion with retail media promotion. Right. So um the other mistake I'll say um is like just splitting out digital and that being the only thing that you're tracking. Like it's very important to think about how retail media connects to consideration um the for transactions that aren't just online, but yeah, also in store. So tracking store um metrics as well, um pretty important from a measurement perspective. But yeah, I I think like strategy is the most important part of measurement because that helps you decide what to track before. And and that's where it's easier to track things than ever, like data's flowing quickly if you have a good um ad partner if you have a good data partner for the scintilla data. You know, you should be seeing metrics flowing in at a daily grain. Um and you can see indicators quite uh early in the day. So um yeah, but again, just that strategy piece coming in.
SPEAKER_00Yeah. So I feel like you know, measurement data. I mean, it's it was hard, it's kind of hard to separate these actually these two topics, right? You know, we're talking about measurement before. But you know, one of the things when I think about you know, my experiences in measurement is being in being influenced by conversations with certain brands that don't trust the measurement, the measurement, you know, a measurement technique that was used a lot of times. You know, you have two parties. You have a brand and you have the retailer. And the retailer is providing data in measurement, they're they're measuring, they're providing data on how something performed. The brand is saying, well, that's not what I see. Right. Um, and maybe even saying the way you're going about the the measurement here doesn't align with our measurement techniques. So therefore, we can't trust this, therefore, we can't use it, therefore it's not a value, so therefore we don't participate. Let's just say that's a obviously an extreme case.
SPEAKER_01Yeah. I think one of the things that's important um is to look at that broader set of metrics. Like um, Roas isn't gonna tell you the full story, and it's like a time limited story too. Like if you convert one shopper, you might get their loyalty for you know the next six purposes or they might stay in your basket or they might be added to subscribe and save.
SPEAKER_02Yeah.
SPEAKER_01Um, and so just looking at one metric at one point in time is quite difficult. Um, but you have to think about like how are you looking at like a longer time horizon?
SPEAKER_02Yeah.
SPEAKER_01Um, and also like are you focused on the right goal? So like if you are just um kind of using your budget as a blunt object, right, then um it's very difficult to actually like maneuver. Um and there are there are things like ROAS you can spend on branded terms and like just increase your ROAD because people are gonna purchase your brand if they're they're searching for your brand. So there's things like that where it's like uh if things are really malleable from a measurement perspective and they're easy to influence um by just like putting dollars in different places, then you know that's something that I would be skeptical about as both a brand and a retailer. Um I don't know, are impressions really an or like I think impressions are often like not a trusted source of data? Is that a North Star like for anything?
SPEAKER_00Right.
SPEAKER_01Like what does that actually tell you?
SPEAKER_00Um allegedly it should give you an indication of reach.
SPEAKER_01Yeah.
SPEAKER_00Right. And so if you're, you know, in in today's day and age where there is no mass reach channel anymore, right? And so you've got to be able to, if nobody knows about your product, then how are you expecting to convert them when that time comes, right? Yeah. So it does play a role. You know, whether the numbers, you know, it's one million versus two million or nine million versus ten million, you know, those numbers obviously, you know, it's it's based upon the methodology and and so forth. So, but yeah, I mean, it it's I I think the idea here is again, it's one data point amongst a measurement story that you're telling, right? Yep. We got this much reach. We also saw a certain cat a certain audience behave in a certain way. And we know that because of these indicators from our our measurement and the data that we were able to collect. You know, one of the one of the one of the things that I have heard and been in the room for is, you know, the whole conversation about modeled versus like test and control. Right. So where you have your holdouts and, you know, and and and you know, some brands do not like modeled data. Um, there's a lot of reasons behind that because it's more predictive versus actual data. The test and control is maybe a little bit more reliable, but it's not as easy to deploy. You know, what are your thoughts around some of the methodologies to get to uh you know, you know, measurement that is helpful to a brand?
SPEAKER_01Well, Walmart actually has made test and control um pretty easy in the scintilla product. Yeah. I mean, again, the scale is always a problem, but they have a test and control model.
SPEAKER_00Yeah, well, what role does scale play? So I think you know, for those that are listening that, you know, don't understand, like if I lit if I if I have 10 people in my pool versus a million people in my pool, what impact does that have?
SPEAKER_01Yeah. I mean, I think um it's hard until you start to parse down. Like, um, I don't think you should start measurement at scale. Like you should start with specific tests on specific games.
SPEAKER_00Like a thousand and or yeah.
SPEAKER_01Yeah, yeah. So you can like refine your strategy. Like, I I think there's like a scaffolding you can build to get really good at this over time. Uh, and so it's about starting with, okay, like how do we like take a small test? Um, and you can do it through the test to control or through, you know, just um a test and learn. Yep. Um, either works, but like you start with something small where you can get very specific about what your strategy is and what your goal is and what metrics you need to execute. And once you kind of figure out like how you can tell the story at that scale, that's when you can start to scaffold up and apply the same principles. But again, like different items are gonna require different metrics. Um, everything behaves the same. So you can't just shift and lift um everything, but like getting really good at one uh building one muscle, you know, helps build the rest. Right.
SPEAKER_00Yep. Yeah. So what would you say are some basic principles around measurement that anybody should take away?
SPEAKER_01Um I would say like be consistent, right? Like changing your strategy too often, um, that makes it quite difficult. Um, I mean, leave room to be nimble and make different decisions, but always like, you know, have a shared source of truth. I would say again, everyone should be looking at the same numbers, have a source of truth. Like if one team's looking at things one way and another team is looking at things another way, and you're like talking uh across each other, um, and not like looking at the same thing. So I would just say, um, like everyone needs to be on the same page about what we're measuring and why. Right. And what those measurements are telling us. Um, and so everyone should be informed. So like it's worth spending time like talking about what you're measuring and why. It's worth everyone on the team understanding what one metric moving in one direction means and what those what decisions should be made. Right. If it's if this metric is green.
SPEAKER_02Yep.
SPEAKER_01Or if it's red, like what decisions are we making if this is green? What decisions are we making if this is?
SPEAKER_00Yeah, decision tree based upon those metrics, based upon you know, benchmarks or you know, whatever that, you know, whatever the goal is, right? So just in the OMG model, knowing the goal, then it's almost a series of if-then statements, right? So if this happens, then we will go this route. And in pre-planning that, I think a lot of times, just again, my experience, so it's not necessarily everybody's experience, but everybody's so reactionary. They got 15 minutes to to plan for something that some of these, some of these core things get forgot about or don't have time, got time for that.
SPEAKER_01Yeah.
SPEAKER_00Then at the end of the campaign, there's this pressure to, okay, what did we learn? Well, shoot, all we have is what we normally do. We didn't really look at it from the same lens. So I think really honestly, uh one word I would bring to this discussion is intentionality. Be intentional with what you're going to measure and what those measurements mean in order to adjust your business. And that could be quote unquote real time, could be after the fact, whatever. I think taking the time. This is our objective. Everybody aligned? Yes, we are aligned. This is how we are going to this is this is how we're going to measure. These are our goals. If these, if this, if this happens, we're gonna move, we're gonna change our creative. Yep. Right? If that happens, we're going to stop running media on this particular channel because obviously it's not helping our business, right? And so but it's rare, at least in my opinion, it's rare that you see that level of discipline because it's like, oh, retailer just called me. We have to have a campaign up in 15 minutes, or somebody waited two months too late to launch it to kick off a campaign or kick off a media plan. And so now everybody's got two weeks to get the media plan done, the creative done, and out the door, which is you know, a uh a less than ideal time. And so there are two things that get cut out in that situation: the measurement plan and the QA. Those are the two things that get lost in that. And so those are two things that are probably the most critical at the end of the day, right?
SPEAKER_01So, you know, so yeah, I'll layer on that that I think a lot of people think about measurement as something that happens after the action. Um, a lot of people are like, um measurement is about what happened, but measurement should equally and maybe even in like a greater sense be about what you should do. Yeah, right. And so, you know, there are things that are you can do really early, especially in retail media, um, using shopper data to help craft what campaigns you're gonna run. Like, and you know, this is where like you can start to think about scale because you could do something like budget flighting and you could budget flight by like pack type by Walmart week. And you can really do that because you can see your market share by Walmart week and the shopper data that that Walmart makes available. And so like you can get really granular with how your budgets like should be allocated. And it's easy to say, like an allergy season, I'm gonna, you know, budget more for my allergy brand than my toothpaste brand. But you should I so that's easy to say, and I think probably anyone executing retail media for anyone who has an allergy brand is probably doing that.
SPEAKER_02Yeah.
SPEAKER_01Uh, but you should also be doing that for micro moments. And you can even think about like geotargeting, like, you know, back to school starts at different times across the country, right? And so when you're thinking about geo-targeting back to school, like look at where certain items are starting to move in certain stores with that historical share data. Like, where do you need to, where have you been able to capture the most share? Where is it really important to get in front of shoppers? Where are sales the highest? And you should, I mean, like you could really flight budget by Walmart week.
SPEAKER_00Right.
SPEAKER_01And brands should be thinking about doing that more.
SPEAKER_00Yeah. Yeah. You and I both come from an agency experience. Um, both worked with big clients. In my experience, I've had very mixed results when it came to sharing data back with the agency to be able to figure out if things work. I've had situations where it's been better and have had a better measurement story to tell. I'd love to hear about like what your experience has been around, you know, brands that do share data and and so that you can measure appropriately or and don't, and what the big differences between those, those happen to be.
SPEAKER_01Yeah. I mean, it's a matter of um how accountable can you be to the results that you want. Right. And so what's really important when you're working with big brands is like accountability to the results that they're expecting from your partnership. So like everyone needs to be on the page, same page about what those results should be.
SPEAKER_02Right.
SPEAKER_01Um, and then like you need to be accountable about like how much you're able to drive the results with what data that you have available, right? If you're if you're a partner. Um, and it it's it's helpful like working with big brands. Like I try to um make sure we are honest about what isn't working too.
SPEAKER_02I yeah, yeah.
SPEAKER_01I remember early in my career working on like with a large brand, we were recommending not executing spin on branded keywords because it was just a ROS play wasn't as incremental. But um, you know, actually that hurt the sales performance of items like within a you know, a couple weeks we were like, oh, you know, like a quate is or uh a great value shows up at the top of your branded keywords if you're not sponsoring there.
SPEAKER_02Yeah.
SPEAKER_01And so like you really can, you know, push private label down the search grid a little bit. So you it actually does make sense. I mean, what's the you know, you can think about what's the right bid on those keywords because you are more relevant. Yeah. Um if someone's searching for your brand. So like maybe don't overbid on those keywords, but like it's not the right strategy to not bid on those keywords. So it's like you need the sales story to be able to know that not bidding on branded keywords is not actually the right decision. You should actually reserve some budget for that. Um, otherwise, people are gonna search your terms and see private label right at the top.
SPEAKER_00Yeah. Yeah, I think the key word in what you mentioned was accountability, right? How can I be accountable for results that I don't get the data for? So if if you're gonna be the gatekeeper of this data and we're making a recommendation, how can we, you know, manage this campaign in partnership in order to drive the results truly, because we're flying blind for the most part, right? And the better you, the better you are at sharing and the more confidence you are you are at what you're sharing with your partner, you know, be it an agency, ad tech partner, whomever that is, we can then be more accountable for the outcomes. If we don't, if we it's a black box, then it's hard for us to do anything more than just report back on, hey, this is what our click-through rate, these are our impressions, this is, you know, you know, whatever data we can collect row as and whatever it is. Yep. But if you really want us to dig in and really t say what happened to with your business, that's what you know that that's we we need we need better partnership there.
SPEAKER_01Yeah, and that's why it's important to be knowledgeable about even the data that you don't have. So if you like as a retail partner, uh and more stuff done a great job of um releasing more and more data to media partners, I'm really excited. The more media partners have access to uh some subset of um the scintilla data now. But like it's important to still be knowledgeable. Like you still can't get the shopper data um some of it at scale, like the demographics and the baskets and the switching and the assortment deep dive. Um, but it's important to be knowledgeable about those. So you can say, like, while we don't have access to this, you know, we would recommend like these are things that you could look at to see if it's driving this. So like you shouldn't even uh be thinking about how brands should be measuring if you're if you're gonna come as a strategic partner, even if you don't have access to the data.
SPEAKER_00No, that's actually a good point because I do think the specificity there is important. So sometimes it's like, oh, well, we want all your sales data. Well Why? Yeah, and why like it that seems so general. But if you come to the table and you say, I need to know X, Y, and Z, these, you know, these three metrics, oh yeah, well, I can get you those three. But if you just say sales data and then, you know, somebody's trying to figure out how the heck am I gonna, you know, navigate that and just I'm just gonna give you everything because you asked for everything. What do you really need? Right. And so sometimes, you know, in this in this plethora of data that we live in now that we kind of started this conversation about, like being specific about what you really need and being knowledgeable about what you really need was an extremely good point because that that is truly gonna get you better results than just say, Oh, I need your sales data.
SPEAKER_01Yeah, yeah. And even if you don't have it, you can help brands understand, like, you know, are they meeting their goals? Right, right. Um, so I the the best you can show up as someone who like cares about the results.
SPEAKER_00Yep.
SPEAKER_01Like that someone's paying you to get, um, you're gonna be a better partner.
SPEAKER_00Yeah, absolutely. So there's been this technology that's come around in the last few years. It's two letters, it's it's supposed to have changed everybody's life. But one of the things that I do think it has helped a lot with is, you know, data, data processing and reasoning, and that is AI, right? And obviously, everything we've talked about today, really, between the data and the measurement, you know, AI could potentially play a role to support that. I'd love to hear your perspective on what role AI has in the data and measurement side. And then we'll get into a little bit more about AI too.
SPEAKER_01Yeah, um, I think AI is only as good as like the systems you set up around it and the prompts that you set up around it. Um AI um is so like I think most uh frontier AI companies have focused largely on coding. But what that means is um AI is very good at writing and executing code. And I think that's how even non-technical people should be thinking about it, which means like if you're just asking an AI simple questions, it's gonna come back with simple responses.
SPEAKER_02Yeah.
SPEAKER_01But if you're thinking about measurement and you're thinking about like AI as uh a partner in making decisions and a partner in strategy and a part a partner in um helping execute things that are more granular, like if you're doing weekly budget flighting, hard for a person to do.
SPEAKER_02Right.
SPEAKER_01Um, a scaled AI system could can actually do those things, but it's not something that happens overnight. You have to think about it um like a software engineer. Uh, and so I I think like the biggest thing is like similar to what we were talking about earlier, which is like start with something small, like automate something small. Yeah. Um, and then you can start to scaffold something that looks more like an AI system and be thinking about like, I'm not just asking the AI a question. I'm thinking about like why am I asking this question? What's my goal? Um, is there a way that if I'm asking this question once that I'm it's not just a one-off answer that I'm getting? Yeah. So if I like this result, how do I benefit from this information that I just got that I liked on a regular cadence? Yeah.
SPEAKER_00So how do you set up something that scales and you know is is easy to replicate over each time, right? Yeah.
SPEAKER_01And AI doesn't do that well on its own.
SPEAKER_00No. You can ask AI five It needs context.
SPEAKER_01Exactly. Yeah. But you could ask even with context the same five questions and AI. might respond differently yeah five different times but code does not respond differently five different times so you have to think about how to set up like what question are you trying to answer what result do you want how do I have AI that help me be an engineer yeah that helps me build the process that gets me the question that I want which is hard I mean it it is a a different skill set yeah but uh that's what where you have to start thinking to really unlock like what AI can do for you.
SPEAKER_02Right.
SPEAKER_01It can speed up workflows with one yeah one off answers for sure.
SPEAKER_02Yeah.
SPEAKER_01But where it really scales and like you start to get those like 10x like where you're really thinking about I can do really granular things. I can like multiply my own work is when you start thinking about becoming a software engineer. And yeah I'm not saying build end to end applications.
SPEAKER_00I'm saying how do you yeah I mean I think I'm picking up what you're saying about being a software engineer and I'm sure a lot of people who are listening are very intimidated by saying that right it's like I'm already I'm already freaked out about AI. Now I need to be a software engineer in order to do it. But I think you know with with a software engineer there is an order of doing things in a certain certain way. And I think if you think about it from more of a process mind mindset, it will allow you to work with the AI AI better. So you know one of the things that I tend to do when I use AI, I don't typically I don't just jump in and start throwing down a prompt. I do take time and I'm actually if I'm not putting it on a piece of paper, I'm actually thinking about okay, what do I want my outcome to be? How can I break this down into small chunks? Because one thing I know AI doesn't do well is big processes, especially processes that you want it to do over and over again. If it's all one process, it has lots of opportunities to go off in the wrong direction. Because AI is really based on predictive models, right? It makes one poor prediction and it's going to go totally off in another direction, right? So if you can chunk it into like if I so and that's how how I think I think okay I want I know I want to get over over to you know my end result. And the first thing it's going to need is it needs to know what I'm asking it to do. And it so it's going to need context. Okay, where can I get my context? Oh I have a PDF I have maybe something in my email or I have like and so then I start to I start to build the stacks to get to where I'm going. And then I am the person that's connecting those pieces together. I think that's what you're really illustrating when you're saying being a software engineer because it's not about coding because honestly you know with coding comes syntax right you have to say things in a very certain way. With AI you don't right and there are different ways to communicate that get better results. However, it's natural language you can say whatever you want it'll figure it out for the most part and you know it may give you wrong answer every once in a while but for the most part it'll give you what you're looking for. So I I think it's the systematic approach of a software engineer's type of mindset. But for me it's actually making me think deeper because I am thinking about aspects that I'm probably not normally thinking if I just have to you know build a deck or I have to you know write a paper, you know, write a uh a a white paper or something like I'm have I have to think about it from you know a lot of different angles composition of it like I have to think about all of these elements and bring them together and and work with the I AI to get to that desired result. And so I think it is it does allow me to probably think a little bit deeper than and it's doing the hard work right it's doing the hard work and I think everybody will agree like the low level work AI should do for you, the bigger thinking work you should do for yourself and you can be supportive with AI in that process.
SPEAKER_01Yeah one of the biggest mistakes that I see um and uh crisp's like core product right now is an AI uh agent tick platform and I think the biggest gains from AI are the things that you dread doing. So like I see people trying to tackle the first thing they do in the platform is like I've never been able to do this uh and they have kind of like a lofty goal of something that's quite difficult for even their own analysts to do and things that are hard for humans to do are so hard for AI to do. Right. Uh and they they take time and they they uh require you to build quite a bit. Um but there's so much that AI can take off your plate that's just like I'm I have to go pull data from these two different data sets and drop them in an Excel file. And that Excel file populates this report that I want like if you're doing stuff like that, AI can 100% do that. So be thinking about like what is frustrating for you every week what's like annoying and you wish you didn't have to do and it's just data entry is that for me. Data entry I like I absolutely yeah despise data entry even if I have the data and I have to copy paste it like no AI here this is my data please put this in an Excel sheet for me right and format it in a certain way yeah so those are and even like deck building like uh one of the things that we've been kind of emphasizing with the brands that we work with is you know like a lot of folks will go if they're doing like a competitive analysis they'll go to Walmart.com and take screenshots they'll go through PDPs like maybe they're looking at like their competitive assortment or new items that launched and they click through PDPs take a screenshot put it in the PowerPoint take a screenshot put it in the PowerPoint. And like actually AI can just go grab all those images for you and drop them at a PowerPoint. Yes. And that takes 30 seconds versus like you spent you know 10 minutes. And that's just like you know kind of shitty work honestly. Yeah. And that's something you should think about like anytime you're doing something that you're like hmm this is pretty like manual. Like think about delegating that to AI because it's not just our tool like Copilot I think could probably do a good job of scraping like that's not unique to a one specific tool but be thinking about like what is routine and what what is manual because those are the first things you should do with AI. And you can't you can scale you can actually like get the processes together to do things that are hard for your your analysts to do you can do that with AI um but it it does require humans in the loop it requires building it requires the strategy and the decisions and um deep retailer knowledge.
SPEAKER_00What would be some examples of some things that someone who could you know start to use AI and then you take it from that to the to the next step and do do more advanced works. Because sometimes I think people see the the you know people vibe coding and you know you know doing these very complicated things and get intimidated when you know usually if you just start small and then you kind of stair step your way.
SPEAKER_01Yeah I would I the biggest thing I work with um brands on when they get into our tool is like thinking about personas or uh in product we call it jobs to be done.
SPEAKER_00Yep.
SPEAKER_01Uh and so you actually should have a different like prompt or process for different types of jobs. So like think about if you're a category manager you should have a different job or a different agent for attribution. Like because that has different rules like how should attribution work like what rules should it follow like what's the item like universe of items that it works on. Like how do you look at your categories? How do you look at your subcategories like if you're starting to blend a bunch of different contexts like an agent's very bad at one-shotting like you're not gonna get a great analysis if it has to do attribution and then it has to figure out what data to bring in and then it has to like run the analysis those three things like one agent for attribution you know one agent for um like the analytics engine what do I need to measure every week what do I need like execution one for that's good at building decks.
SPEAKER_02Yeah.
SPEAKER_01So think about it if you had like a team uh that was able to support you think about like the different ways you would break up your job um and start to think about like how to tailor like what context is important for attribution.
SPEAKER_02Yeah.
SPEAKER_01What context is important for drawing mods?
SPEAKER_02Yeah.
SPEAKER_01What context is important for measuring whether mods are successful or not. Like those are three different things. Yeah. And trying to have one kind of AI that has the same context doing all of those things. Yes. Um like you're just not setting yourself up for success. But like build in um doing it this way where you're thinking about like jobs to be done and you're thinking about splitting up AI into the different workflows you do. Like think about how every response can build on itself. So like if you're just one shotting or one offing um Monday morning reports every week like that's less valuable than if your agent can go back to a Monday reporting morning report that was saved from four weeks ago and it said hey your replenishment in stock on this item is below target. And then the Monday morning report that you run this week can go back remember that that happened check on it and like make sure you're you're back on track.
SPEAKER_02Yeah.
SPEAKER_01So like it's about stacking context but if you're stacking context across all these different domains it's a mess. But if we're stacking context where like I have this workflow it's a Monday morning report needs to do this for me. This is my goal I need to be able to like speak intelligently to my merchant catch things that wouldn't be caught until after the fact and also have like visualizate top top line sales, visualizations that are consistent. Yeah like you can stack those and have them build on top of each other and then remember like you can start to create like a brain or a knowledge base that like really improves the result over time.
SPEAKER_02Yeah.
SPEAKER_01So that's the other thing that's hard to think about with AI, which is like the first result is not the best result.
SPEAKER_02Yeah.
SPEAKER_01So if you're using that first output, yeah you got to take some time but it really does build on itself and it really does get better if you invest um in like what you're trying to do as if it was an assistant that was helping you.
SPEAKER_00Yeah. Yeah yeah makes makes sense. I appreciate that tip I think you know just again starting small building upon that you know and then start to bring in the things that you that you need. Yeah I I do like the idea of you know maintaining uh a library of context you know you're saying like you know report after report after report that's probably a good tip for me a lot of times I just have it you know overwrite the the previous report but if I just had it do a new report and then just let it look at that folder it can go back and look at all the the the reports and then trends over time.
SPEAKER_01Yeah I uh I keep even a personal log of everything I did on a given day and I don't do that manually I'm like look at all the tabs that I looked at look at all the emails that I wrote look at the Slack messages that I sent compile it into a log I never look at the log but then I can ask AI to go back and say like hey um I'm looking at this result like was there anything that I did in the last couple weeks that yeah like influence that so you can kind of draw and that's a lot of context to build that would be useless without it like I would never do that unless AI was helpful. But like you can actually start uh to hoard I recommend hoarding data like now that now that AI is seriously like makes it easy to read. Obviously it was not like it it was too big of a universe but yeah I mean I can go back and like I have a question did anything from my meetings in July like did this ever come up did anyone ever say anything about this and like uh in a past life I would have had to like click through the trans. Yeah but they I can just like search it 15 seconds come back and say like oh yeah like go read this part of this meeting and someone at like 20 minutes said this if you want to rewatch it.
SPEAKER_00No, it's good. Yeah yeah yeah cool cool all right well I appreciate they appreciate helping out with uh covering you know the AI and all the things that are uh changing there so we're gonna wrap it up with some bold vibes so this is uh our traditional wrap up game where I throw out a statement and you give me your gut reaction don't think okay don't think it overthink it just give me what you think all right uh all right so more data does not automatically lead to better decisions your thought yeah I mean I agree with that um if you don't know again like why you're looking at data um and you don't know like what your goal is or what your strategy is I I don't think the data can be helpful. Data is most valuable when it challenges what the business already believes yeah I like that I like um having my thinking challenged by by data and I like try to think of it that way but good measurements should tell a brand what to stop doing not only what to do more of yeah I I think um overall allowing yourself to be able to like pause and change strategy is is a good approach. Human judgment will become more valuable as AI becomes more capable. Yes yeah cool all right well that's it for bold vibes appreciate you playing along so thanks uh thanks so much Laura and thank you for joining me today uh it's been great having you on and uh sharing your perspectives and your and and your knowledge today so before we get out of here is there anything you would like to share or plug?
SPEAKER_01Yeah I'll um I think one thing I'd like to share is um there's a vibrant local music scene here my partner uh Roger Barrett is a wonderful um local music promoter.
SPEAKER_02Okay.
SPEAKER_01Uh and so I would say like if you're buying tickets to the AMP or like the big shows of the momentary think about also supporting like smaller shows at Georgia's and they have um the momentary is a great bit indoor venue called the roadhouse that um brings in smaller acts and sometimes you know those are what lead to your favorite bands coming to town too. So I'd say buy tickets to see local music um and local um support local artists.
SPEAKER_00Yeah support local I absolutely love it so um all right well um so for everyone watching and listening thank you so much for spending time with us today if you want to catch all of the episodes of retail media vibes you can visit retailmed com. Thank you so much for listening and as always I promise to do better next on BV out.