661. Navigating AI: The Power of Our Wishes with Nicolas Darveau-Garneau

Professors fm. You're listening to the Unsiloed podcast with Greg LeBlanc. Produced by University FM, Unsiloed is a series of interdisciplinary conversations that inspire new ways of thinking about our world. So wherever you are, enjoy today's episode. And here's your host, Greg LeBlanc. Welcome to Unsiloed. This is Greg LeBlanc. And I'm here today with Nicholas Darvo Garneau, who is currently, I think you have a consulting firm that is named after yourself, right? In the business. But of course, you spent many years as the chief evangelist for Google. And of course, I had Guy Kawasaki on the podcast earlier who was, I think, the first chief evangelist that I ever knew about. And so this is a profession that is relatively rare and relatively unique. And it seems like you are continuing to evangelize, but you're evangelizing for, it seems like the field of marketing, but also for the things that marketing can teach other parts of the business. And when I was in business school, I had a very skeptical view of marketing. I went to the school to study finance. And so on marketing day, that's when people would show up dressed like Toucan Sam and stuff like that. And then I took a marketing class and I realized, oh, wow, actually, this is super rigorous and super quantitative and super deep. And this was, of course, nearly 40 years ago. And you suggest that marketing is like the litmus test or the canary in the coal mine. And companies that don't get marketing, they probably don't get anything very well. So why do you suppose that is? I mean, why is it that marketing seems to reflect the, I don't know, the true nature of the company? Oh, great. Well, first of all, thanks for having me. And it's a great question to start. Look, marketing is usually fairly straightforward to think through in terms of what you're looking for, right? And the way I think of the world now, more and more, as we're building these AI Genies, right, who are now will be able to grant us wishes. And those genies are not really what matters because we're all going to have genies, and the genies will be fairly similar to each other. But it's the wish that we select that will matter. In marketing, the wish should be pretty obvious and we could talk about that. But if your company gets that wrong, the odds of your software development team or your customer service team or your product team thinking that through correctly, if your marketing team hasn't, are really low. So as Google's chief evangelist, I met over a thousand companies in Five years. And what I noticed was that Even in marketing, 90% of the time, the teams were asking for the wrong wishes. Well, I mean, it seems like Google, their interests are aligned with the companies that are marketing through the Google platform. I mean, one might think that if they don't know what they're doing, they're going to wind up spending a whole lot more money. But I think it's probably the opposite. It seems like once they figure out how to get the most out of their marketing spend, they're going to utilize marketing more aggressively. Right? Yeah. I think it all again goes back to what you're asking the AI to do. Right. And so let's take an example. If you and I run two different charities, I run charity A, and for every dollar I put into Google, I raise $2 of donations, right? So I have a two to one ROI. Whereas you, I'll put a dollar into Google and you get five, five dollars a donation. So you get a five to one ROI. And we're both smart. So we don't have a marketing budget because we found a money printing machine. We crank this up as high as we can. And in my case, right, I'm able to actually invest $10 million and raise 20 million bucks, so I get $10 million left. In your case, since you want a higher ROI, you can't really scale it as much, so you can only invest a million dollars to get five million dollars. So you're left with four million bucks on net donations, if you will. So I'm way better off than you are now. So really, a charity shouldn't care about its roi. It should care about the donations left after it's spent its marketing investment. Obvious, right? But nobody does it this way, Right? In marketing, you know, all the dashboards we've had over the last 20 years, when you open up Facebook or Google, it shows you your return on ad spend, your roi. And including myself, like when I was entrepreneur, we've been drilled to try to increase the roi. I would just submit that that is absolutely the wrong thing to do. We should try to maximize profits. Yeah. I mean, look, there's two things. One is that if you're flying blind and you're not using any kind of metrics or KPIs, and you're not quantifying things, you're probably going to make some bad decisions. But the other is that you could be measuring lots and lots of different things and be very quantitative, but you're measuring the wrong things, you're using the wrong objectives, and you were describing in the book. It actually reminded me. I did a podcast with John List, who is the chief economist currently at Walmart, and he said that when he was at Lyft, he had to explain to everybody in the organization the difference between a marginal cost and an average cost. A very, very simple thing that you learn on day one of business school, but which somehow got lost. And the fact that companies can forget that the goal is profit, it's kind of puzzling. So why is it that companies look to these intermediate metrics or these partial metrics or these incomplete metrics and rather than focusing on first principles, which is, you know, what are you actually trying to achieve here? Right, well, thanks for mentioning the book, by the way. It's called Be a Sequoia, not a Banzai. And it's basically the seven things that the top companies I met do very differently than the other ones and only about 5% of the companies that do it. So there's two reasons why, for example, an advertiser and a marketer focus on ROI as opposed to profit. One is that's the metric they've always used, right? It's been in the dashboard forever, all day. But even Google's Gemini, right? Google's AI, she asked to define performance marketing. I did this this morning. Again, this says it's marketing with the highest roi. The word profit isn't in there. So it's just habit, right? And part of it is because until four or five years ago, you actually couldn't really optimize profits using AI and marketing. You could kind of do it, but it wasn't easy. Now it's really, really easy. So we have to throw away the old metrics. And to your point where you think from first principle, but that's like, that's the easy one. Let's complicate this a little bit, right? Where you and I are still running or charities. I've got charity A, you've got charity B, and I can acquire Kathy, she donated a hundred bucks and I can acquire for $10, right? So I have a 10 to 1 ROI, but the net donations is $90, right? 100 minus 10. And you acquire Steve, right? Steve gives you 500 bucks and it costs you $250 to acquire Steve. And your ROI is terrible. It's only 2 to 1. But you have 250 bucks left, right? So you're better off than I am. But what if Steve is never going to donate again? And what if Kathy's going to donate every week for the next 30 years and what if I can build an AI to predict that with 70% accuracy? Like the customer lifetime value of Kathy versus Steve. And Kathy blows Steve out of the water. She's 100 times more valuable than Steve. 99% of marketers will not use this data. Even if they can predict Kathy is going to be way more valuable in the long run, they're still going to focus on the short term. Well, now, is this because it's easier to measure those short term things? Right? Customer acquisition cost is something you can measure with a high degree of accuracy and you don't really need to wait too long to find out what it is. And so you can incorporate those KPIs into the quarterly cycle of reports and rewards and so forth. I mean, is that the main reason? It's precisely that, right? I mean, we're as humans, we're actually used to running large projects based on forecasts. We use Excel all the time. We forecast stuff. As a marketer, it's because these dashboards of like the precision of a metronome, right? Your ROI is 3.6111 and your cost requisition is $4.22. And the customer lifetime value of Kathy is $10,000 plus or minus 30% because we can't tell, we don't know. So that kind of unknown and making decisions based on forecasts takes a while to get used to. And so most companies can actually build these crystal balls that tell them how much Kathy is going to be worth in the long run, how much Steve's going to be worth in the long run, how much each and every individual customer is going to be worth in the long run. But they choose not to use them at all to optimize their marketing. But there's so many powerful case studies. I'll share one. There is an online insurance company in Canada called surex S U R E X. And historically, what they would look at is number of customers acquired today and the cost of customer acquisition. CAC. Right. We acquired 200 customers and the average cost was 30 bucks. And we're happy with that because as long as on average we're below 30, we're making money. And that's fine. It's a nice first step. But 5% or 10% of these customers you just acquired are going to have a massive car accident and delete all your profits and 30% are going to churn within a year. Wouldn't you like to forecast that? So Cervex built a predictive algorithm to forecast the customer lifetime value of every One of these new customers, which included the likelihood they're going to churn, which included the likelihood they're going to have a car accident. And they rated each new customer from 1 to 10, 1 being very unprofitable, 10 being extraordinarily profitable. And they started sharing that with Google and Facebook and next thing you know, within a few months we'd acquired 60% fewer of the very unprofitable customers, 90% more of the very profitable customers, and their profits quadrupled. Right. And so if you can do that. And by the way, like, predicting car accidents is pretty hard. It's a lot harder than predicting who's going to churn from their telecom bill or from their streaming device. So in almost every industry, you can actually build an algorithm to predict the value of a brand new customer you've never seen before. And the book explains how to do it. It's not that hard. The hard part is just cultural and getting used to not having precision. But once you get used to it, your results are almost always significantly better. So there's a bit of a transition time. You have to get used to it. You have to bring senior management, you have to bring a finance team on board. But I've done this a lot with a lot of companies and almost always they're unrecognizable going forward. Well, look, I mean, I've been teaching predictive analytics for a long time and I've interviewed a lot of people who you actually reference in the book. And you can predict the customer lifetime value, but it presumes that you have been in business for long enough that you have some good data. Right. Track record and you also have the belief that things aren't going to fundamentally change. I'm guessing right now that a lot of these SaaS companies are revisiting their customer lifetime value numbers. Does it make more sense when you're dealing with a startup? It seems like there's such. I mean, one of the reasons why people like to stick with the rigorous metrics is because they're afraid that people lower in the organization are going to cook the books. That seems to be a concern. And I remember when Enron was marketing to market all of their projects, the minute they signed a deal, they would more or less book all of the profits for the next 30 years on that project and say, oh, look at all the money we made by signing the deal. And then that turned out to be a little bit of puffery. Is there a concern at the people at the higher up in the organization that the marketing people are going to hoodwink them. Is that one of the reasons why they don't like these metrics? Yeah, I think the two points you made is this customer lifetime value of a brand new customer isn't particularly precise. You can get it fairly precise. Well, I meant for a brand new company. For like a company. Well, even if you're a brand new company. Right. And so let's take an example now in the meal delivery business, right, those ready meals you had to assemble yourself, and there were hundreds of those. As opposed to looking at the number of members we acquired in the cost per member, you know that 90% will trend within a year. Even if you're brand new. You can actually, you don't have to build a customer lifetime value model for 10 years, but you can predict with a pretty high degree of precision who's going to churn within the first two weeks after you get their 75% off. Right? So maybe you move from like number of members and cost per member as a KPI to number of members, you stick around after two weeks. Right. And that's very quickly you can figure that out. So the point I'm trying to make is that whatever best predictive metric you can use is often better than today's perfect metric. And then in terms of cooking the books, yes, of course there's danger that whoever is doing this is saying that AR predictions suggest that the customers we're acquiring are acquiring now are a lot better than before. And it's not the case either because the algorithm has been completely wrong and or the team is doing things that are untoward, although I've not seen a lot of that honestly. But the trick is to basically make sure that the predictive algorithm is built by finance and marketing together, right? And it's co owned and finance believes it. And the other trick, by the way, is in dashboards going forward to put an error bar on every number, right? If you predict and you just acquired, you know, 50 customers today and the combined customer lifetime value will be $50,000, put a plus or minus 30% around it and that's a lot more credible than you saying, I predict we're going to make $50,000 in the next 30 years, which is very not credible. So being humble, being honest is critical to move from this precise short term metric. The way I think of this, by the way, is like imagine an airline pilot that loses all their instruments, you know, and he says, look, we're going to be okay because I know it's -65.5, 6 degrees Fahrenheit outside, it's like, well, that's irrelevant. What you really want to know is that it's, Your heading is 350 degrees plus or minus 30 degrees. You're way better off having an imprecise heading than a very precise temperature. And so most marketers measure the temperature, not the heading. Yeah, I mean, I don't mean to push you so hard, it's just that whenever I read a book where it seems to be saying something very obvious, I always wonder like, why do we need to say this over and over again, right? Like, what is it? There's gotta be some, like, deep reason why companies ignore basic insights. And another one that you point to is this idea of treating every customer the same. So companies will use averages rather than identifying the differences in these customers, some of whom are quite valuable and some of whom are, you know, probably negative value. Yeah, I mean, like if you ask a thousand CEOs, which I have, what percentage of your profit in the next five years will come for your top 20% more valuable customers? You'd be surprised how few know that. Right? It's around 20% or so. And then if you ask the follow up question, which is, hey, like, what did I eat for breakfast? And who are they? Where do they live? How did you get them? What was the first product you bought? Why are they so valuable? And what can we do to other customers to make them as valuable as these customers? The answer will be a blank stare from 99% of CEOs. The ones who know this and act upon it completely outperform. So I'll give you one example is if you look at the entire history of optimizing websites or optimizing an app, we call it landing page optimization. For example, for websites I ran for startups, I did thousands of these tests. They were all wrong. I messed up every time. Why? Because what I looked at was the conversion rate of page A versus page B, right? Hey, well, this page converts better than that page. Well, that's an indication that something is better. But is it more profitable? Like, is it converting better? Because Everybody's buying at 90% off and they get free shipping for life, right? Or so what's the margin per visit? Even better. What's the customer lifetime value per visit? So the trick isn't to test or design your customer experiences for the average customer. It's to actually build customer experiences for your best customers. Very, very few companies do it this way. Right? And so once you can predict the customer lifetime value of a Brand new customer or an existing customer, the entire corporate strategy should revolve around it. And to your point, you know, people like Peter Feder been in saying this for a while and many others. And so this is not a new concept. Where the book I think really comes in handy is exactly why you're not doing it. What are the obstacles in front of you and exactly how you get around them. Right. And so step by step, like, hey, finance is worried about this or your CEO is concerned about that, or you're worried that the algorithm won't be right. So here's how you do it. So there are very practical ways to get this done. One of the biggest concerns people have over optimizing for customer lifetime value is I'll make more money in the long run, but it's going to cost me two or three quarters in the short term of bad numbers. That is absolutely not the case. There's a way to move to long term profit optimization while still making short term profits go up. The book explains how to do that. So it's really if the concepts are well understood in theory, people freeze at the execution and at the cultural changes they have to make. And you also mentioned pressure from stakeholders. So it seems like folks who are not in the marketing space, they don't always understand what's going on. Is this a failure on the part of marketing people to communicate with the rest of the organization to kind of explain to the finance people what's happening and why the finance people might be looking at the wrong metrics? Yeah, listen, I'm not going to point fingers at one or the other, but I would say having a CEO, a CFO and a CMO who get together a lot and talk about growth, profitable growth, long term growth, not just this quarter's growth. There's nothing wrong with this quarter's growth. But you're really kind of lining the strategy to your point, Right? Marginal profits or marginal costs are defined together. We agree on what that is. The first chapter of the book goes into like, well, obviously you should optimize profits, not ROI. But then it goes into like the 17 ways to optimize to calculate profit and how nobody agrees, right? So like, okay, so you know, there's a CFO and a cmo because I've seen so many marketing dashboards and then marketing team is jumping up and down like we've increased profitability by 30% and then the CFO takes one look like, no, no you didn't. That's not right. And so, and often, by the way, CMOs are not invited at the table. Right? Fewer than 25% of CMOs are invited to board meetings. The average CMO tenure is something like 18 months. So I don't know if it's the CMOS fault or the CFO or the CEO. All I know is these three people should be best friends and work together, not across purposes. Using a joint agreed upon dashboard of what is supposed to be optimized by marketing. Now, oftentimes we now see these new roles like Chief Growth Officer or Chief Revenue Officer. What are those new positions mean? I mean, why are they necessary? How do they fit with cmo? Yeah, I think the traditional CMO was doing branding, communications, performance marketing. Right? And that's a subset of the levers you have, especially in the digital economy, to grow the business. Right. And so what is a digital onboarding process? Right? You're a customer, you see an ad, you click on it, you go to the website. What happens next? What is a set of SMS or texts or emails you get after that? What is the AI that you're going to build to be helpful to customers? How do you get a customer who's just bought one thing to buy again? So that whole kind of customer journey, usually historically had been split between the website team, the email team, the marketing team, and putting it under one roof under the guise of a Chief Growth Officer or chief Marketing Officer with larger purview is often a very good idea. So what you want to build is a growth team, right, that has performance marketing, brand marketing, it has website design, it has a B testing, it has email marketing, it has all these things. And then the team doesn't own the product or the product is still owned by the product team, but they own the onboarding product. Right. They decide how the customer views the product in their first few interactions and then what happens next and how to sequence that journey, how to personalize it. That is a really powerful set of tools. And if you have one person running all that, right, Was the one really thoughtful KPI, ideally customer lifetime value. That person has a multidisciplinary team, not just marketers, but data scientists, AI engineers and so forth, then you can do a lot very quickly. And so I think a big part of marketing now is attribution, trying to figure out how to allocate resources and which levers are actually moving the needle, which ones are getting the customer to profitability more quickly, maybe keeping them more profitable while you have them holding onto them longer. Right. I mean, what do people get wrong with most attribution models? I Mean, I remember I used to teach the problem with Last Touch and some of these more basic rudimentary models. I think most companies have moved on and become much more sophisticated, but it seems like they still have trouble figuring out what's driving what. Yeah, that's a great question. And I think, listen, we know a lot better than before in the age of TV and print what's driving what, but it's still partly opaque, it's certainly not perfect. And as Apple has made changes in the past, for example, and other privacy changes have happened, it's gotten a little bit harder to decide for what works and what doesn't work. The best advice I can give people is use the best method possible, one that leverages AI. Take an off the shelf product. Google has one, there's others. You compete. You may want to tweak it a little bit for your own purposes, but don't spend too much time on this, right? It certainly shouldn't be Last Touch or First Touch. You should have, you know, each touch point should have some value and you should let AI decide where the value resides. You shouldn't let Facebook or Google decide, right? You should let a neutral third party decide. But more importantly, much, much more importantly, you should do constant testing, right? So you look at the totality of what you're doing and first, by the way, whatever you're doing today, the first test, you should run in a small place in the world, right? You take a small state, you should like build the biggest, baddest marketing campaign you can. You should build what I would call a Sequoia, right from the book, meaning invest in everything, even if you're not sure it's going to work. So you like, you call your Google team, you call your Facebook team, you call, you know, your other teams, your account teams will work with you and they say you should do this, this, this, this, this and you just go crazy. Like you're probably over investing. There's probably non trivial amount of money that's wasted, quote, unquote. What you're going all out, right, in Iowa, doesn't cost you that much money. But you're just testing what happens when you have no restrictions. Ideally, good things happen. It's like, well, we're actually making a lot of money, more money in Iowa now than before and it's way more profitable despite the fact we know there's waste in this investment. But we've gone all out and look at this like we're making twice as much money as before. Okay, great. So now let's Find the waste and pull back. Right. And test removing this YouTube campaign or this Facebook campaign and see what happens. Are we making more money or less money? Before and after? So one of the big mistakes marketers make is they never get to the Sequoia phase. They're just like, they have a little thing they're building. It's okay. It's like a. I don't know, they're investing enough and it's really, like, efficient. It's a bonsai, not a sequoia. And then they try to add on, you know, 2% on top of it, and it doesn't work, and you freak out and they remove it. They're always pruning, like, they're always like, the bonsai is always pretty, but it's not getting bigger. Let it go right? In a small state and just build a big, massive Sequoia and then prune it. Right. And find the things that don't work. And hopefully that Sequoia you built in that one state is now twice as big as the previous bonsai. And you replicate the Sequoia that has been pruned just the right way in every state. Right. So this kind of testing, as opposed to changing your attribution model every three months and now making wild swings of like, hey, Google is really valuable to us. Oh, no, it's not. Yes, it is. No, it's not. It's not the best use of time. Testing stuff, testing your next hypothesis is the best use of time. So, I mean, that seems to suggest that companies are. There's more companies spending too little on marketing than ones spending too much. Right. It sounds like when you find something that works, rather than just like, all right, we got it. Now let's fuel that fire. There seems to be like, well, we don't have the budget for that. Like, I mean, that thing about that you mentioned somewhere that you should not ever go into a marketing campaign with a fixed budget. Yeah. I mean, it's profitable. Right. If you're investing money. Right. Let's go back to our charity example, right? When you were investing a dollar and you were getting $5 in donations, that's a money printing machine. Right. So if you open up your Google account and your Facebook account and it says, hey, you missed out on 30% more donations because your budget ran out. Yeah. That should never happen. Right. And so I call this, don't turn off the money printing machine to save on electricity. Oh, we have that university. We do that all the time. We're like, oh, we got a budget. And that's the budget and no more budget. Come back next year and ask for more budget. Yeah. And you know, understand that the older, more entrenched, more rigid the organization, the harder it is. But to the point you made before, in order to get this free flowing budget, you have to convince finance. Right. So marketing, clearly at this point, if you're there, right. It's obvious that marketing and finance are not in agreement because if they were, none of the budget would flow. Most sophisticated performance marketers don't have a budget anymore. That battle has been mostly settled. There are some counter examples, but I'd say it's like 70 to 80% settled. The bigger issues are the fact that even with an infinite budget, you're trying to focus too much on efficiency, not growth. And number two, you're certainly focusing too much on short term as opposed to longer term. So replacing short term ROI with customer lifetime value profit. If you do that right, that's when your marketing budget typically quadruples because you are seeing things very differently. And now you're making a lot more money in the longer run. And again, there are tricks to make sure that you don't lose money in the short run. But as you make that investment grow and as you keep acquiring the more valuable customers, they keep stacking on top of each other. And next thing you know, you got these revenues that keep coming at you without having to acquire more customers and the revenues grow exponentially. So that's the real trick, is just stacking high CLV customers on top of each other, on top of each other so that your churn goes down a lot. You know, profit per customer goes up a lot. And next thing you know, you can actually shut down marketing for a while and the money just keeps coming in. Yeah. I was just talking to a friend of mine who works at an organization nearby and she was saying that her boss is under a lot of pressure. And so the sales team's job is to just collect logos, right? Get as many new customers in the door as possible so that he can report to his boss that things are moving forward. And of course this means shortchanging the long term customers and not investing in the relationships and just getting a bunch of numbers that they can, you know, show. Right. And I mean, this is the oldest game in the book. I still don't understand why anybody would be fooled by this. Right? I mean, yeah, I mean, unless you have a kind of lead scoring model, right, which kind of scores each of these new logos coming in. And that model has been built by neutral third Party, right? The data team working with the sales team, the finance team and the marketing team. So you get a new logo, right? XYZ customer just signed up. Hey. We predict that this new logo is worth $2.6 million in the next 10 years. So that's unless you have that, then nobody knows. If you water it and prune it, like fertilize it and so forth. Right? You have to keep working on it. But this is the model's best expectation, everything else being equal. The problem is it's even worse than what you and I are saying. Not only can you get logos that will churn fast and we'll just waste your time, but next thing you know, like if you're collecting any logo, you don't have a good system to score these logos, you're actually going outside your ideal customer profile, right? You're grabbing new companies. You want new features that you haven't built before. You're yanking your engineering team to build these features. Nobody else wants these features because this customer is not in a sweet spot of what you should be getting. Next thing you know, you've got all these one off features for, you know, 30 different customers. Your customer success team is pulling their hairs out because they've got to do all sorts of custom consulting. So it's much worse than just, you know, you're wasting your time and there's churn. The entire company becomes wildly unfocused. If you are just collecting logos and nobody is putting a critical eye of two things. What is the economic value of these logos? And two, are they actually logos that we want? Yeah. Well, before we turn to getting more value out of your high value customers, I want to just ask one more question about just getting customers. We've seen a huge spike in the number of data scientists and analysts and quantitative folks moving into marketing. I mean, sometimes I think if you're a cmo, I mean, software engineers and data scientists are the bulk of your employees at this point. But it seems like the way they're trained is very different from the way business people are trained and the metrics that they value. Things like accuracy and area under the curve and lift and stuff like that. I mean, those things have nothing to do with profitability. And so is it any surprise then that we have to kind of remind them that profitability matters? It seems like their training was never about the profitability. Yeah. That's why having a centralized growth team under one leader that's multifunctional, that has one tpi, which ideally is customer lifetime value Profits and then all the hypotheses are centralized, right? So here's a list of 150 ideas we have to improve CLB profits and then a neutral third party who's really knowledgeable source these things by likelihood to improve customer lifetime value and how hard is it going to be to do, and so on. And then the growth team is just constantly finding the minimum viable test they can run to prove or disprove the hypothesis. And they move as quickly as they can. And then, and only then, like do you get a data scientist to do work, right? So if there's a hypothesis that's proven through a simple test, I'll give you a quick example. Maybe if we completely customize our emails using a very sophisticated AI tool, we'll have a higher, you know, open rate, click rate, purchase rate, customer lifetime value rate, which is all we care about is customer lifetime value. Okay, that's a nice hypothesis. Why don't we take 100 emails we're about to send and customize 50 by hand and leave the other 50 as is and prove the point, right? As opposed to buying an expensive AI solution and implementing it for six months. You should never, ideally, right, you should never go into a complex project that requires huge data lift or huge IT lift or a third party integration without knowing for a fact that it's going to work from a business, not just technologically, but, you know, the lift. Like, you know for a fact that this thing is going to be a 7% lift because you've done two or three tests before, you're investing that 5 million bucks to run it. So now like the team is doing all sorts of testing and then they're really implementing the things that are really valuable and whoever's running that project. So now we agreed, for example, that this project to personalize emails is super valuable, that it's going to deliver $16 million of CLV. Therefore we're willing to put $3 million behind it. And the project manager for that project has to put CLV in front. Keep reminding the data team that CLV is key. But the good news and the bad news at the same time is that, I mean, I've seen some demo of Anthropic recently that could actually do this right out of the box, right? I mean, like it would go into Selflake database, you know, take all the customer data, run the clv. I mean, we're still gonna need data scientists, but if you prompt the AI the right way, it's gonna make sure the data scientist is also doing the Right thing. So hopefully the AI will be a good coach. Data scientists to make sure they keep the eyes on the business metric. But honestly, it's a project management's job, right? There's one dashboard we care about. We're looking at one metric. It's clv. And here's how the data that we need, we don't need all the data. And here's the algorithm that we need. It doesn't have to be perfect. As long as it's 60% accurate, I'm okay. And so go. Right. So that has to be spelled out very carefully or else the data engineers are going to do what they've been trained to do. It's not their fault. Yeah, I mean, I heard you use the word prove, but I mean, you're not really talking about proving in a scientific sense. I mean your comfort is you're being comfortable with a level of confidence that something short of foolproof, right? Yeah, yeah. I mean, yeah, prove to the best of your belt. But look, if that email idea is 60% likely to work and the upside is 60 million and the investment is 3 million, that's a proof, right? I mean, you've already proven the case because the upside is so much bigger than the cost. If this is a Vegas casino, you should make that bet all the time, right? So prove enough that the investment's a no brainer. It shouldn't even be close. If you're like, you have to, you know, dot the I's across the t's after a test or two to prove that the profitability will be there. You probably shouldn't do it. Now you talk about how to get more value out of your most valuable customers. I mean it's first of course you have to identify who they are. But I mean, it seems like the airlines know, right? I mean, they know, hey, you're silver, you're gold, you're platinum. Right? I mean they make it very clear to you. That's very clear to everyone in the organization. Why don't other companies do that? Why don't they understand, oh yeah, this is premium customer. This is the customer that you know. It seems like if they do have that stratification, it's very simplistic, you know, I know with my bank, it's like if you have $100,000 balance, then you're in one category below 100,000, you're another category that doesn't really capture the CLV. I mean that's a very crude metric. Why don't companies have a really good metric and then why is it that they under invest in making the high value customers happier and keep them longer? Yeah, I mean, especially with the tools that we have today, Right. In most industries it's pretty straightforward to forecast the customer lifetime value of an existing customer. To your point about the bank, right. There's some subtleties. If you have $100,000 deposited in the bank, that's a start. But even better is somehow me understanding that you have 2 million bucks at some other bank that I don't know about. Right. And so your customer lifetime value is way higher than what I think it is. So there's some tricks on how to get this right. The banks do ask you, they do have to connect your other account. Right. I've noticed they do. So there's ways, right, to kind of really peg you from 1 to 10 or 1 to 100 on how valuable you're going to be. It just goes back to what they are willing to do with it. Right. So if you're not willing to do a lot with it because your metric isn't clv, that there's not a lot of point in building a CLV model. Right. And so maybe they'll prioritize, you know, your customer service, email or your phone call if your CLV is high. So that's a pretty straightforward use case that's been around for a long time. So they have a crude model to rank you from 1 to 10 and then you get ahead of the queue. Okay. So that's. And by the way, they typically just give you a different phone number. So that, that's one crude model. But they're not running the whole business around this. Right. If you look at some of the best companies in the world, what they do is, well, they acquire the most valuable customers first, right. You also acquire the other ones, but they just pay less for them. So they acquire very valuable customers, a disproportionate percentage of the valuable customers in their industry and then they do everything in their power to increase that clp. Right. Amazon prime is a really good example and there's lots of other examples like that. And I'll give you my favorite example of all time. So I was working with a fashion retailer, well known fashion retailer, and they had to build a CLV model for every individual customer. So they had a good sense for how much you were worth in the next 10 years. And it wasn't just, you know, how much you would buy, but they got it down to the margin, but they actually could figure out what you're going to buy. Approximately and the margin of each type of sku. So they could tell that Greg would buy, you know, $20,000 worth and 12,000 bucks of profit, and Nick would buy $10,000 worth and $6,000 of profit over the next 10, 15 years. So they could do that. Well, if you have that and you can run that model a lot, like every day, for example, and you can see the change in greg and Nick's CLV, right? So a month ago I was worth $10,000, but now I'm worth 10,500. Hey, we've added $500 of value to Nick and we only invested 50 bucks to do that. Therefore, this is good. So imagine thinking this way, right? And you now implement an AI recommendation engine on your website that recommends another piece of clothing for somebody who's already bought clothing. And the AI has been told to optimize and increase conversion rate, make people buy more stuff. It hasn't been told to increase customer lifetime value. So what is it going to do? So if a woman, for example, has bought lots of dresses, you're just going to see more dresses all the time. So we were doing this and the AI worked like it increased all the numbers. Revenues went up, profit went up, the promoter score went up. It was good. It was good. There's nothing wrong with it. I'll tell a merchant, look at this and said, I don't think that's right. I think we could do better. And the way to think about this is if we keep selling dresses that we would have bought dresses, they're eventually going to churn and they're just not going to be with us in the long run. We should pepper these dress recommendations with other categories as well. Now they're maybe less likely to buy, but if they do, they're going to be worth a lot more. So we tested this whole idea, right, with email marketing. And we ran, we had a test group and a control group, and we saw that the customers who got the emails with more diverse categories bought a little bit less, but their CLV went up by 45%. So we retrained the AI algorithm and said, hey, yes, send a lot of the same things people have already bought, but 30, 40, 50% of the time, send something completely different. If you go to Netflix, by the way, it used to be you've watched this, therefore you like that. Now their whole page is like, hey, you've watched a lot of Rom com. What about, you know, watching a little bit of horror? Or what about, do you have a child? What about watching some children's shows, because presumably I've having to work with Netflix on this. But presumably if you just watch the same show over and over again, you run out and then you like more likely to churn. But if you've got you're watching six different categories, the odds of you churning are probably pretty low. So that kind of insight, where you're optimizing for the longer term, the value of existing customers moves mountains. Yeah. I guess the question is how do you find these leading indicators? Because it seems like when you finally realize that your brand has suffered or you finally realize that you've alienated a customer and they've attrited, it's often too late. I mean, I have a friend who works at a bank and they had this customer attrition model. And initially the customer attrition model used net promoter score. And so if the person had a low net promoter score, then they predicted they were going to trip. But of course at that point there was nothing you could do. So you need to know somewhat earlier that you needed to do something to hold onto them. So how do you find these leading indicators that maybe you're doing something wrong or that's going to damage either the customer relationship or the brand? Yeah, sometimes you can't. Right. But ideally you would build a as real time CLV model as possible, meaning it changes every day. And so it's not based only on how much money has been deposited into my account, but it's also tied into maybe your last customer service interaction. Right. Or the last time you use your debit card or what is your activity on your credit card? Like the model. Some of these banks obviously do this, but the model has multiple flights of data which as you use all of that together into one kind of coherent customer lifetime value predictive model. It may say like, hey, Greg is worth $10,000 and then the next day Greg is worth 9,500 bucks. And that drop may be enough to trigger an alert because Greg is in our high value category and he's dropped a lot. And so trying to get that most immediate signal possible is really key. And then having a plethora of interventions that you can use and knowing what the next best action is for you, that is going to move that CLV up. So that's actually tough because to move CLV up, I have to please you, but I can't give too many things away either. Right. And so most, in my opinion, and I haven't studied this in depth, but most recovery programs now sometimes. Well, first of all, a Lot of work has been done at the bottom where the customer is not recoverable. And that's just a huge amount of effort. And sometimes we just have to let customers go. But in the middle or the top, sometimes we create an increase in that promoter score at the expense of clv. So what you really want to do is you want CLV to go up and that promoter score to go up or at least stay flat. If you're making more money but the customer's not happy, that's not good. If the customer's happier but you lose the money, that's also not good. So there's a sweet spot where the MPS and the CLV are both moving up and finding the interactions with the customer that both delight them but also increases profitability is the job of what a CLV improvement strategy should be. Now, is there a way to measure the impact that a customer might have on other customers? Because if they're really happy, then they're going to influence other customers to buy the product? Yeah. A lot of sophisticated, especially online companies have a really good sense for the average kind of virality coefficient, for example, Right. They understand what drives that they can. Also, I've worked with one company at least, that can create a virality coefficient per customer. They know, for example, that you, Greg, are more influential than me. Right. And so they actually bake that into their customer lifetime value. Hey, I've acquired Greg and after a couple of weeks I could see that Greg is already doing something that suggests he is going to bring in other high CLD customers and therefore recursively. Right. I'm going to say Greg's CLV is 20% higher than I thought originally when I first acquired him. So, yeah, you can get to that, that level of sophistication, honestly, I think it's nice. Right? But just doing the basic CLV work where you're acquiring the most valuable customers and then you're also trying to turn your like second most valuable customers into the most valuable. Those two things. And the third thing, building a customer experience based on high CLV customers as opposed to the average. Those three levers, like, are so powerful that you don't really need a lot more for the next couple of years, if you follow the program in the book, just do those three things. And then the fourth thing I would say we haven't talked about is branding. Imagine like you're a sequoia, not a banzai, right? And you're acquiring the most valuable customers in your industry. You're increasing the CLD of your existing customers faster than your competitors. Your entire customer experience is built to delight not just the average customer, but the most valuable customer. So you make very different design and product decisions. And then lastly, your brand is more beloved and improving faster for the most valuable customers in your industry than your competitors brand. So historically, branding was this kind of mysterious. Your introduction to the call. We had people wearing weird costumes. It would, you know, smoke, you know, thin cigarettes and drink vodka and tell us our brand should be that those creative people are still very necessary because the right creative is really critical. But now we have highly sophisticated tools to test whether or not these creatives work or not. Right? So you put an ad into YouTube, you go into a small state, you don't have to put a lot of money into the test. You also don't have to put a lot of money to the creation of the ad to use AI very wisely. So your production costs are lower, your test costs are low. Then you're like, hey, I'm putting this ad into Iowa. Is there any evidence that my brand is doing better in Iowa? Are the searches for my brand going up relative to my competitors? Is the awareness of my brand going up? Is it consideration for my. Is there anything I can point to? But if the answer is no, change the creative until there is some evidence, right? So we did a lot of that at Google and we were able to very often double the number of searches for a brand or double the awareness of a brand in a small state and scale that to the U.S. right? And then now you start investing money, right? And you invest money on YouTube, on Facebook, and then you put it on TV again. Back to the concept that you shouldn't, like, pray that something's going to work. Like before you make the $10 million TV investment, you know, with very high degree of confidence that it's going to work because it's working on YouTube, it's working on TikTok, it's working on Facebook, and you've tried, you know, seven different creators and this is the one that works. So there's a whole way of doing what I call modern branding. The other concept, by the way, is, which is a little controversial because it's not the only thing you should do. Your brand should be improved for everybody. Eventually everybody who's in the market for you could be in the market for you. I believe that. But maybe in the short term, you could focus on making your brand a lot better on today's highest value. CLV customers in your industry, if 10% of the customers drive 90% of the value today in your industry, grab those 10%, invest the money in those people. Your cost of media will be cut by 10x. And since they're 10x more valuable than average, your campaign is going to be a harder X more powerful per dollar you invest. So that's a really good start. And then if you choose to go down the CLV chain and target the second most valuable tier customer, then the third more and eventually you can target everybody. But you should probably start with the most valuable customers. They are searching for your brand twice as much. So there's a lot of rethinking to do about branding in the age of digital. That and again, I was talking to the book, in the book I quote one of my Google colleagues who's still there and she said we're not seeing anybody really do this. They're still thinking about branding in the old TV ways, right, where you produce creative, you put it on YouTube and mod TV and then you measure, you know, using a media mixed model six months later. And so that is probably not the best way to do things nowadays. Do you think that companies could do a better job of sourcing insight from their customers? I mean, I know when companies are startups, they're always interviewing customers, they're always asking customers questions. I mean I remember when Amazon first launched I could email the team and say, hey, you know, I think you ought to have a shopping bag that didn't disappear after you purchased stuff and you'd get a response from them. And now good luck trying to ever get in touch with Amazon. But I mean when was think at least with the B2B companies that there would be a way to source insight and information that isn't simply passive, right? You're not just looking at click throughs and you're not just looking at activity. I mean, how can companies get insight, maybe qualitative insight from their customers? Yeah, there's a myriad of ways. But the most important thing before you think about tactics is to again differentiate now the word customers, right? The average customers from the most valuable customers. And so for example, if you pipe in your customer lifetime value data into Facebook, into Google, very soon you'll start to see what those types of customers are doing, what pages they're going to, what products they like. You really from a pure like data perspective, you start seeing a different data stream than before which was your average customer's behavior. Now you're seeing your best customers behavior. So that's very different. And then once you circle who those best customers are, then There's a number of tools. Right. I'm on the board of a company called Alita out of Toronto, Canada. And it's like Customer Experience SaaS, a software company. And as opposed to sending surveys to random customers, you can create the community of your most valuable customers. Right. And survey them over and over again and get some longitudinal data. And so the community can really give you insights. And then there's a myriad of other ways. Right. You can of course talk to customers. You can do all. But the most important thing isn't the tactics so much as moving away from the customer to the segment of customers who matter the most. And that's going to change everything because everything you're going to learn, like the most valuable customers in your industry do not want the same thing as the average customer. And so you're going to hear some very, very different things than if you're just surveying the average. Yeah. Well, last question. I think you end every chapter with some lessons that other parts of the organization can take from these insights in marketing. And I used to teach a course on HR analytics and a lot of the people that were speaking in the class, they had started off in marketing and then made their way over to hr and they would use all the same tools. They would talk about employee churn and they talk about employee lifetime value and employee acquisition costs. And I mean, to what extent can the things that you're highlighting in this book apply to procurement, apply to hr, apply to sales, I mean, to all the different areas of the organization? Yeah, they do for the most part. Right. Sometimes have to be modified to fit the specific requirements of the department. But a couple of thoughts. One is if fewer than 1% of marketers are following all the best practices in my book. Right. And marketing is arguably the easiest thing to optimize to long term profits. Yeah. You said less than 1%. I couldn't believe that number. Yeah. So if that's the case, when you can bet a lot that if your marketing team isn't doing what's in the book, the rest of your departments are not because they're harder to figure out in HR is harder to optimize than marketing in many ways in terms of what's the right metric and how do I measure it. So that's point one. Point two is having said that, if you really think it through very carefully, you can do some pretty solid procurement. This is the example of my book where the UK health department procurement teams changed completely its TPI from lowest cost to best long term benefit. And so one of the examples they gave is like incontinence diapers for people who have incontinence problems. And so if you tune the cheapest diaper, you're going to save a lot of money, but you're going to use two to three times more diapers. So they switched the KPI to like the lifetime costs of these diapers as opposed to the one time costs. And according to them, this is really hard to believe, but according to them, they're saving £500 million a year by making this one change. Right? HR, same thing. Right. If your KPI is time to hire or cost to hire, and you know for a fact that in your industry a top 10% employee is typically 10 times better than the average, then you're missing the whole point. So can you identify top prospects ideally really quickly, but it's not quickly when you're hiring them at least exposed facto. So you can actually track the revenue contribution, the profit contribution of each employee in their first 18 or 24 months? Yes, it's hard to do. Yes. It's not perfect. But again, going back to don't measure the outside temperature on the plane perfectly. Measure the heading of the plane imperfectly. Right. So every department should start with what they wish they could measure. What's perfection? I wish I could hire superstars. I don't really care how much I spent to hire them. I don't really care how long it takes. But I want to hire superstars and I'm willing to invest more time and more money to do so. Okay, how do you track whether or not you're hiring superstars? What are your best ideas? Right. And then how do you refine that over time? And then how is it? Would the CFO agree to doing that? Would the CEO agree? And if they did, you change your metric. Right. Call center average handle time, terrible metric, and so on and so on. On time delivery. Right. For your delivery team, is that the right thing to do? Now, when you have some customers who are worth 100 times more than some other customers, maybe you want to be way early for the most valuable customers and you could be slightly late for the average, I don't know. But it's certainly not a metric that is been designed for the average customer and for efficiency. Right. You're trying to move the customer's value up. You're trying to move the net promoter score up. Right. So what is the right KPI for a delivery team that's going to enable that to happen? Now? There are ways to do it. Well, I know Google's always been talking about the 10x coder and the 100x coder. And so I think our universities, we maybe should start thinking about the 10x student, 10x faculty, 10x alumni, and maybe we can rethink how we reach out to all those different groups. Well, wrap all this conversation. I know we have to go, but wrap all of this in a layer of AI now, right? Everything we said is useful, but if you believe like I do, that AI will 10x or 100x what we currently do, but it doesn't know what to do. It doesn't know what we care about. It can do things, but we have to decide what the KPI is. So if you give an AI the wrong KPI, you ask it to be more efficient in the short term for your marketing, for example, it's going to acquire a whole bunch more customers at a much lower cost per customer. These customers will be terrible and it's going to go in that direction very fast and very efficiently. So the whole point of AI is that we're building these amazing genies, but the power isn't in the genie. The power is in the wish. We have to think very carefully about what we ask these genies to do for us. By the way, I actually think that's going to make us even more human. We have to really do some deep ethical thinking, some deep value thinking, and really give these AIs the right wishes. And then I think we'll thrive and be fully human. But if we give these AIs the wrong wishes, then AI can send us into a tailspin not only as companies, but as a society pretty fast. Well, this has become a whole lot better at doing the wrong thing, right? Yeah. Yeah. Well, Nicholas, thanks so much for joining me. The book is called Be a Sequoia, not a the 7 Growth Secrets of the World's Most Successful Companies. Thanks again. Thanks, Greg. Thank you for tuning in to the Unsiloed podcast produced by University fm. If you enjoyed today's episode, please give us a five star rating and review in your favorite listening app. To listen to our other episodes, please visit our website@www.unsiloedpodcast.com.