Is Your Dashboard Lying To You? with Tim Shea

Podcast cover art for In Clear Focus episode Is your dashboard lying to you?

IN CLEAR FOCUS: Guest Tim Shea of Latticework Insights explores how brands can maximize ROAS: their Return on Analytics Spend. As marketers struggle to track data across scattered platforms, Tim explains why customer lifetime value is a nuanced story rather than just a single average number. Discover why domain expertise is the new oil, how to avoid drowning in AI-generated slop, and where human leadership remains absolutely critical for deploying artificial intelligence safely.

Episode Transcript

Adrian Tennant: Coming up in this episode of IN CLEAR FOCUS:

Tim Shea: We encourage people to think about analytics as rocket fuel that can grow your business faster. A dollar that you spend on analytics should produce some multiple in returns. You should find $10 worth of insights for every dollar that you put into the analytics machinery.

Adrian Tennant: You're listening to IN CLEAR FOCUS, fresh perspectives on marketing and advertising, produced weekly by Bigeye, a strategy-led, full-service creative agency growing brands for clients globally. Hello, I'm your host, Adrian Tennant, Bigeye's Chief Strategy Officer. Thank you for joining us. Most marketing teams believe they are data-driven. They have dashboards, reports, and often more analytic tools than anyone can name. And yet, when someone asks a deceptively simple question like, what is our lifetime value, or which customers are actually worth acquiring, the honest answer often involves a frantic scramble across disconnected platforms that never quite agree with one another. My guest today has built a firm around exactly this problem. Tim Shea is Founder and Chief Executive Officer of Latticework Insights, a California-based data science and artificial intelligence agency. Across more than 25 years at the intersection of data, marketing, and technology, Tim has worked with over 70 brands, from the NBA and the UFC to direct-to-consumer names like Seed Health, Henry Rose, and Fellow Coffee. To explore why data may be your most underused asset, I'm delighted that Tim is joining us today from Los Angeles, California. Tim, welcome to IN CLEAR FOCUS.

Tim Shea: Adrian, thanks so much for having me. I appreciate it.

Adrian Tennant: Tim, the average retail or direct-to-consumer brand keeps its data scattered across 20, 30, sometimes 40 different platforms. Paint me the picture of what happens inside one of those companies when a leader asks a simple question like, what is our customer lifetime value?

Tim Shea: Yeah, yeah, it's a great question. It's a very manic picture that I can paint for you. So I always think it's interesting, both big companies and small companies share the exact same problem, right? It's this data in different platforms. It is by no means a choice. It is often framed as an advantage. Adrian, if you and I were to start a brand like tomorrow, this weekend, we're going to start a new brand. We would immediately have our data stored in Facebook and Google and TikTok and Pinterest. We would be selling our products on Shopify and Amazon, maybe Sephora or Whole Foods. We'd be using QuickBooks, maybe NetSuite. We'd be using Klaviyo, Salesforce. And before you know it, when someone asks something simple, like, hey, the investors are asking for our LTV:CAC ratio, do we know what that is? And Adrian, both you and I would say, hold on, we're going to go download 20 spreadsheets and we're going to jigsaw them all together. And it's not going to be a matter of just simple arithmetic, because Meta and Google speak days — how much money did I spend today? — and TV speaks households, and Amazon doesn't give you the same fidelity that Shopify does, and Sephora just gives you a monthly report. And so there's all these philosophical questions we need to answer. So all the smart people in the company start running around like crazy every Monday morning just to build this report so they can show LTV:CAC on screen for everyone to look at, and sort of, you know, big round of applause. And God forbid anyone asks a follow-up question. And then we've got to go do some more data wrangling. And then Monday is just right around the corner. We've already burned eight hours of our week on just assembling a report. And so this data integration problem becomes a real existential problem for brands, especially when they reach scale. Things start to work. They get their product-market fit. They raise their A round. And now their investors are like, the goal now is to raise your B round. And in order to raise your B round, we need to be tracking these metrics every single week and make sure we are A/B testing against them, making sure that the charts are all going up and to the right. And it becomes very difficult. And oftentimes what we see is like a COO or a CFO is spending all of their time downloading spreadsheets and jigsawing them back together again. Companies like Latticework come in and we give them back their time. We automate that process. We help them see their business maybe through a different lens. Maybe they think of their business as having reached a sort of stalemate. They say, “Hey, Tim, like our profit margin is X and our cost per acquisition is Y and we've tried everything under the sun and this is what it is.” And, you know, we come in with a different perspective, having worked with 70 brands across a bunch of different verticals, and say, “Well, what if we look at it like this?” Which would have turned the problem on its side a bit. So we help them see their business through a new lens and help them graduate to their next stage of growth, wherever it is that they're trying to get to. So hopefully we can clean up that initial manic picture and we deliver them the picture where it's sort of like the dolphins jumping out of the water and the rainbows and everything.

Adrian Tennant: Well, as you well know, marketers fight for media budget every single year. But analytics still tends to get treated as a cost center, a line item to minimize. Tim, how do you make the case for why analytics spend should be judged like an investment, with a return expected on every dollar?

Tim Shea: Yeah, no, I appreciate the question. We are definitely out there talking about ROAS. But for us, it's Return on Analytics Spend. And look, what we do is a data science agency. In many ways, I am envious of agencies, because with agencies, there is an expectation that you're going to be buying Facebook ads all the time, 24 hours a day, 365 days a year. It's going to be always on. And we need smart people with their hands on keyboards running the ads and evaluating if they are working or not. When it comes to technology and analytics, there is a feeling of fixed-time, fixed-cost type of projects, where everyone's constantly asking like, “hey, so when is it going to be done? When do you think that thing will be finished so we can stop paying your invoices?” And we definitely have those arrangements, but we also encourage people to think about analytics as a form of rocket fuel that can grow your business faster. There should be a feeling that a dollar that you spend on analytics should produce some multiple in returns. You should find $10 worth of insights for every dollar that you put into the analytics machinery. And so we need to walk the walk and talk the talk. And so, you know, again, we go back to this lens question. It's not just that we're making things automated. It's not that we're populating your Tableau dashboard with the latest data every week. It's that we're building these latticeworks in service of coming up with insights. And those insights are based on, for me, 25 years of working in retail and DTC, working in CPG and apparel and traditional retail. So we come in with a new perspective and we say, hey, we've seen a similar situation to what you guys are doing in another circumstance. So, you know what they did is they kind of looked at it like this. They found that if they could just reduce churn for their premier customers by 1%, if they could just get them to buy a candle on their second purchase, that second purchase pays for all of the CAC for their last quarter. And people say, “No, no, no, no, no. It can't be that.” I'm like, “Let's run the numbers. Let's run the numbers and show you the step function that you will experience. Because again, it's not just the money that you would get from them buying the candle or the t-shirt. It's that they get a second go at your brand.” And they remember, hey — Adrian, I'm sure it's the same for you — you go online and you buy an iPhone case, and then the company emails you like 75 times in the next two weeks. You're like, “I don't know about your brand. I don't care about your brand.” If you could have a second chance to sell them something where they can touch and feel this product and experience your sort of brand universe, it gives them a chance to have that aha moment. They're like, “oh, Patagonia is really cool. Nike is actually a really cool brand.” And so there's a sort of soft win for what you're getting out of that second purchase. So yeah, we're really encouraging people to think about analytics and data science in a way that makes a company money, as opposed to just an expense that, you know, you're going to have to live with.

Adrian Tennant: Well, you've written that lifetime value is not really a single number, but rather a story, because a brand almost never has just one kind of customer. So why does collapsing all of that into one average lead marketers astray?

Tim Shea: Yeah. As a tech person, we get a bad reputation for always trying to be the smartest person in the room and kind of splitting hairs. And we're like, “well, actually, you know, your LTV is not $100.” And everyone's like, “okay, here come the nerds telling us how smart they are.” And we're like, “no, no, no, it's not that. It's not that your LTV is not $100. It's not that it's $105 or maybe it's $96.30. It's a story. It's more than just a number.” People say, “look, I know how to calculate LTV, Tim. I'm just going to take my total revenue and divide it by the total number of customers.” And you're like, “great, that's ARPU, your average revenue per user. And that's not the full story.” And the story is that a guy walks into a store and buys a t-shirt. And the question is, well, how long did it take him to get to a second purchase? Did he come back in? Many different types of customers have different frequencies. Some of them buy one thing every week. Some buy one thing every Black Friday. And you say, well, the guy walking in the store, how much marketing did I need to spend to get that one person to come into the store? Many users have different CACs. It costs a different amount of money to bring in certain types of customers. If you're paying a premium to get a premium customer, then great. But if you're paying a premium to get someone to buy something on Black Friday at a discount, you're going into debt on that first purchase. So those discounts are part of the story. Did the guy walk into the store last week? Because if we can't pair him to someone that bought something two years ago, we're just sort of muddying the waters now. It doesn't really make sense. Those are two different types of people. The business was in a different place a couple of years ago. And not to mention, we have lots of Black Swan events in our adult lives, right? These non-recurring events like COVID, tariffs, snowstorms, other tentpole events that may have occurred. And so, you know, we're comparing maybe our business, which was a totally different business during and pre-COVID than it is today. And so it's not to say that we want to be pedantic and we want to say that our version of the LTV number is different. It's just like, can we all be honest? Can we agree that we've got three different types of customers, and those customers have different lifetimes and they think and behave much differently? And so, yeah, we tend to encourage people to think of LTV as a story, as opposed to one number that you just put onto a dashboard.

Adrian Tennant: Well, a lot of founders push back on all of this. They'll say, “I'm tired of hearing about data-driven everything. What about my intuition?” Okay, Tim, how do you answer the marketer who trusts their gut over the numbers?

Tim Shea: Yeah. And look, this is the million-dollar question. Founders do know their business really well. And oftentimes you'll bring them insights and they're like, “yeah, I know that. We've tried that before.” And we want to be really careful what specifically we label intuition. Some people say like, “well, I just have a gut feeling for a certain thing that's going to work.” And you're like, “you know what? You know your customers really, really well.” We did this exercise. I was working in advertising many years ago, and we were pitching Adidas. And we're like, “okay, cool, it's a sports brand, and here's the consumer.” And they said, “hold on one second. Adidas knows themselves really, really well. When we're talking about Nike, Nike is talking about Just do it. They're talking to a specific human being about competing, about being the best athlete that they can be. And when we're talking about Adidas, we're talking about your squad, like your homies, and your team, and camaraderie, and your town, your community. It's two different ways of talking about people.” And those comms pillars inform everything that we do. As data science and analytics people, we need to understand that. We need to understand those intuitions. There's another type of intuition, which is just like, “I feel like we did really well on Black Friday.” And you're like, “do you know that to be true? Right. Do you know that you didn't just steal that top-line revenue from October and December? You positive? Like, it was a $10 discount, right? We gave some people $20.” And so let's just be really cautious about how much margin hit we took on those various discounts, because our intuition is not telling us the exact number. And many times the exact number is the difference between being able to run payroll next week or not. And so, you know, I encourage people to think about this as very like left-brain, right-brain projects. We as data scientists, you know, need to understand that there is a different way of solving problems. We need to get along with marketing, and marketing needs to understand that there's a sort of quantitative lens for looking at things that can be very revealing.

Adrian Tennant: Let's take a short break. We'll be right back after this message.

Adrian Tennant: Welcome back. I'm talking with Tim Shea, Founder and Chief Executive Officer of Latticework Insights, about why a brand's data may be its most underused asset. Tim, let's talk about artificial intelligence.

Tim Shea: Oh my god, we're the only podcast talking about artificial intelligence!

Adrian Tennant: That's right. What goes wrong when a brand rushes to deploy AI?

Tim Shea: Everything. Everything goes wrong. Gosh, it's one of the most interesting, fascinating, dizzying things that, Adrian, either you or I have seen in our adult lifetimes. I have never seen anything like this phenomenon. It is awe-inspiring. The only problem is, it just doesn't work. Yeah, it's not ready yet. And people say, “oh, Tim, isn't that just cope? Aren't you just splitting hairs? Aren't you, indeed, worried about your job? And aren't you just sort of creating this talk track against AI to flex your value?” And I'm like, “well, let me ask you a question. Let's be honest. You've got three different types of employees at your company. You've got above-average superstars. You have average people. You've got some below-average folks. And tell me, what types of output are you seeing from those various cohorts of employees?” Those superstars use the exact same LLMs as everyone else. And they become 10X. They become 10X engineers, marketers, finance people. They're amazing. They brought domain expertise to the table to figure out how to use the models. And they've brought their own domain expertise because they know how the world works. And then you've got these average users that are using it to generate emails and build decks and have been the recipient of these root cause analyses — a breathtakingly designed HTML document that just explains in a breathless-seeming way what's going on with the company — until you read it. And you're just like, “wait a second, this document contains nothing. This is Claude just like generating fluff.” And unfortunately, it's generating 10x as much fluff. And now all these people in the company are sort of suffocating in all this slop, let's call it. And it doesn't mean anything. You're just wasting everyone else's time. And then, you know, God forbid you have these below-average folks. We've all seen it, maybe we've done it ourselves, where you produce something using AI, and it's worse than what we would have produced if we didn't use AI. And now there is 10 times more below-average stuff, below-average code that needs 10 times more debugging, 10 times more Jira tickets to maintain. And it's a really bad situation. I think everyone's scratching their heads as to, you know, well, why are enterprises failing to get things into production? And I think that framework of LLMs as a sort of prism is a useful shorthand. Like, you need to kind of upskill your teams and you need to block out a lot of this noise. Let's call it what it is, this noise coming from Silicon Valley and this noise coming from journalists and influencers who are all making a buck off of the hype and the misinformation that they're touting about AI. So proceed with caution.

Adrian Tennant: Got it. A common thread in conversations about AI with guests on this podcast over the past several months has been the importance of keeping a human in the loop. I guess the sharper question might be whether the artificial intelligence belongs in the middle of the process at all. Tim, when is the right move to take an AI component out and put a traditional tool, you know, like a person, back in?

Tim Shea: Sure, sure. For all the nerds out there, I went to the LangChain conference last year — excellent conference, excellent platform — and I got one of the crew-neck sweatshirts. It says “human in the loop.” And I remember them being on stage announcing this as if it was like a new feature — like, “we've got this new thing and it allows humans to be in the loop with all these processes.” And I remember thinking, I'm like, “well, who invited the LLMs to be in the loop? This doesn't make any sense. Like, this was our party in the first place.” And then, as you build more and more of these applications, we can use LLMs to generate stuff like code. It's actually quite good at generating code. I can generate code and I can read the code and evaluate whether it has hallucinated, whether it's making any mistakes. I can write unit tests to ensure that it doesn't break once it's in production. And I can deploy it and I can sleep well at night because I know it's going to deterministically perform the same way every single time. It's called software, and it's been around for 60 years. As soon as you put an LLM in the middle, you get this intelligence boost out of it. It's good at certain things that we have this jagged frontier to reckon with. But the question is, it freestyles. It does a different thing every single time you run it. And so now there is a question of, well, can I go to bed at night? Or every time it gives a response, do I need to review the response before it goes to a customer? And if I work at Charles Schwab and I'm giving financial advice to a customer about where to put their stocks, there's a regulatory framework that we're living inside of. And if we're a small DTC brand and the LLM says, “you got plenty of money to pay payroll next week,” they do not live in a regulatory framework, but they do go to pay their employees every two weeks. And so I think it's a really existential question for the AI folks about, like, where specifically do the LLMs belong in the stack? Like, must there be a domain expert behind the keyboard in order to guide it to do its thing? I think jagged frontier is a very illuminating analogy. For those folks who have not heard this, the idea is that AI is amazing, right? It can write poetry, it can write code, it can teach you about quantum physics, but it doesn't know how many R's are in strawberry. And it tells people if they're around the corner from the car wash, maybe they could just walk to the car wash. That would be a more efficient use of their time. And you're just like, “well, how is it so good at quantum physics, but it's terrible about what day is it today? Has no idea.” But I think the jagged frontier is a helpful shorthand for, you know, what should we put into production? But I think people need to be honest. Silicon Valley has invested one trillion dollars into this technology. They're going to need to reap — I don't know what they need. Ten trillion? How many trillions of dollars do they need to make that worth their while? Fifty trillion dollars in return from those technologies. And so if we can only use them for nondeterministic use cases — a lot of use cases are deterministic — did we just cut out a piece of the TAM where it can only be used when there's judgment that needs to happen on the fly and when there's a very low tolerance for slip? You can't do brain surgery with it, because you're just going to kill the person that you're doing the brain surgery on. And if we can only use it to build vibe-coded to-do list apps, what is the point of this whole exercise? And so there's a lot of things still to be figured out. And I would not put it in the bucket of like cope or, you know, splitting hairs. These are real problems that AI folks need to figure out.

Adrian Tennant: Well, for all the talk of tools and dashboards, what clients seem to want most is leadership, someone who knows what to do next. Tim, as the technology gets more capable, where does human judgment become more valuable, not less?

Tim Shea: Yeah. Look, I think maybe this is the trillion-dollar question. You know, I'm out here saying that domain expertise is the new oil, right? So, like, a lot of folks said data is the new oil. Great. Those folks are wrong, right? Let's say. And then I think that, like, you know, maybe data is good for the foundational model companies and they need a lot of data. They need probably more than they currently have. I think companies know what data they have. We ran through a list of it earlier in this podcast, right? And they've got to get that data. It needs to be nice and neat. So does that make it oil? Or does it just make it a thing that they have to reckon with? I think a lot of companies are saying compute is the new oil. It's all we need, more tokens. We need more GPUs. You're just like, “look, man, I'm selling fizzy water to Sprouts. How much GPU do I need for this?” And in order to really sell fizzy water at scale to make the company really work, I need the domain experts. Like, I need to know the people that work at Erewhon. For those folks who don't live in New York and Los Angeles, there's a store called Erewhon that charges like $23 for a can of fizzy water. And there's a line wrapped around the corner to get your drink on the shelves. And you need to know those people. You need to know what to say. You need to know the branding to get through to them. You need to know where all the bodies are buried, like how have companies in the past succeeded and failed. I need to wake up every morning. I need to run the company. And so folks come to me at Latticework Insights and they say, “hey, Tim, we've already built all the technology that you have on your website. We've got our Snowflake, we've got Fivetran, we've got Tableau, we've got all this stuff. But, man, nobody can agree on, today is Wednesday. And the marketing people hate the data people and vice versa. And what do we do? The data pipeline keeps breaking and there's no trust. Tim, seems like you've been in the game a long time.” And I'm like, “yeah, you know, flattered for you recognizing that. But, you know, we're going to come in and we're going to upskill everybody. We're going to make sure the data people speak marketing. The marketing people need to know when and how to use data, not just the last minute before they run into a meeting with the board. And we need to have folks speaking the same language. Like, can we agree that we're trying to increase LTV by shortening the time between first and second purchase? Can everyone agree that this is what we're focused on?” Because if what we're talking about is whether we should do an Instagram post with Addison Rae, does that ladder up to faster second purchases? Or is that a flashy thing that maybe we wouldn't award a can for? If we want to win awards, it can. There's a very well-mapped-out path for that. I think leadership is a big piece of this in terms of helping other human beings get along and fire on all cylinders, how to grow companies quickly in different domains, and then how to do the hard technology work to get these systems to work.

Adrian Tennant: Well, you mentioned a moment ago, of course, you did work in advertising a while back. And I think clients have asked agencies to do more with less, probably for as long as either of us has been working. Now that agentic artificial intelligence can automate parts of what account teams and analysts used to do, what does that mean for how an agency really earns its keep?

Tim Shea: Yeah, well, you are definitely correct to assume that people have been asking agencies to do more with less since the beginning of advertising agencies. People have been saying advertising is dead for as long as advertising agencies have been alive. And for some reason, advertising agencies still exist and in many cases thrive. And so clearly, they provide an impeccable service. They are in the business of creating alchemy, especially for these creative agencies that come up with the “just do it” campaigns that sway audiences. Should they be doing more with less? Should they be doing more? Because all their competitors are going to be doing more. They all have these tools. This is not a tool that has only been given to a certain sector of the population. Everyone needs to do more. A colleague of mine did this excellent AI-generated commercial where they created this character. And I think it was a crypto company. This character advises people to file their taxes on time. And then, “hey, by the way, all that Solana coin that you bought, that's capital gains. You need to file Form 32A by this date.” And they were like, “we need this character to show up in many different contexts. So it needs to look really cool at a coffee shop, and it looks cool in a yoga studio, and it needs to look cool at a baseball stadium. And, by the way, we don't have the budget to shoot all those commercials. Is there anything you can do to make this happen?” And the agency says, “yeah, we can. We've got people that are just gangster — creative directors, editors, VFX. We know people that know how to cross the blood-brain barrier and get a message through to consumers. And we'll do it. We're going to do it. And we're going to do it because we know that everyone else is going to do it too. They're going to create — this campaign is going to live not just on TikTok, but on Instagram and YouTube and TV and out of home, all different aspect ratios, you know, the different cuts, and that's table stakes.” If we cannot get it to all those platforms — in the barbershop and the baseball stadium and all these different things — that's table stakes. And so they need to do more, maybe with more in some cases, maybe they do less with less. It takes really, really smart people to figure out how to respond to that question from a brand that says, “it feels like we don't need to fly to Costa Rica for that car commercial. It feels like we could probably just vibe-code it into existence. It needs to look good though, right? Can we agree it needs to look good? You can't have a little baby shaped like an orange peel, and he's dancing around. It can't look like one of those videos. It needs to look really good.”

Adrian Tennant: Great conversation. Tim, if IN CLEAR FOCUS listeners would like to learn more about you and Latticework Insights, what's the best way to do so?

Tim Shea: I love collaborating. I love being a sounding board. I think when people are struggling with hard problems, I would say search me out on LinkedIn. The company's called Latticework Insights. You can go to the website; it's latticeworkinsights.com. Find me, Tim Shea, on LinkedIn. Connect with me. Ask hard questions. I'm really interested in collaborating with retail and DTC brands. We tend to work with mid-market folks, $50, $100 million and up. Folks that are struggling with data and AI questions, when they're just like, “oh God, we've tried this a couple of times and it didn't really work. Everyone else seems to have it figured out. Is there anyone out there that we could talk to that could help us figure this out?” It's a great signal to pick up the phone and call Latticework Insights.

Adrian Tennant: Tim, thank you very much for being our guest this week on IN CLEAR FOCUS.

Tim Shea: Adrian, thank you so much for having me on. I really appreciate it.

Adrian Tennant: Thanks again to my guest this week, Tim Shea, Founder and Chief Executive Officer of Latticework Insights. As always, you'll find a complete transcript of our conversation with timestamps and links to the resources we discussed on the IN CLEAR FOCUS page at bigeyeagency.com. Thank you for listening to IN CLEAR FOCUS, produced by Bigeye. I've been your host, Adrian Tennant. Until next week, goodbye.

Timestamps

00:00: Introduction to Analytics as Rocket Fuel

00:23: Welcome to IN CLEAR FOCUS

00:45: The Data-Driven Illusion

01:16: Meet Tim Shea of Latticework Insights

02:02: The Manic Data Landscape

04:07: The Existential Data Integration Problem

05:20: Analytics as an Investment

08:29: Understanding Lifetime Value as a Story

11:06: The Role of Intuition in Marketing

13:30: The Importance of Human Judgment

15:01: Challenges of Rushing AI Deployment

22:04: The Value of Human Oversight in AI

25:18: Agency Adaptation in the Age of AI

28:28: Connecting with Tim Shea and Latticework Insights

29:16: Conclusion and Farewell

Other Episode

View All