#6 Dr. Markus Wübben | CRM Democratization | Customer Lifetime Value Mastery | AI Integration Challenges

Foreign.

Hey there and welcome to Inside CRM. I'm your host Jessica Jansen and I'm so excited to have you join us. Today on this podcast we dive into the world of CRM with experts from all kinds of companies, from fresh face startups to billion dollar unicorns and solid established corporations. Our goal to share practical insights and strategies that you can actually use. Each episode is a quick 20 minute chat packed with tips and tricks to help you get the most out of your CRM. So grab a cup of coffee, get comfy and let's get started.

I think for me you were one of the persons that quite early were pretty proactive and also loud around CM and its impact, talking a lot around customer lifetime value and I think back in the days I really appreciate it. I still think we need to be louder but I think you were one of the first movers that really pushed it and tried to get attention around it. First of all grateful for that and with that maybe not everyone is aware what you have done in the past, what's your background, so it'd be great if you can give like a little introduction to yourself.

Sure would love to. Thanks. Thanks so much for the warm intro. I'm Marcus. I've been working in CRM or with data for about 19 years now. I started 2005 with a PhD. Currently I'm actually consulting companies on CRM strategy. But what I've been doing at least for the last nine years before that I was running a company called Crossengage. We're a customer data and prediction platform. The last year to a Dutch company now uses the CDP in their entire ecosystem basically. But many years before that I started in a loyalty program called Deutschland Card. Like we had a physical card and you would check how many glasses of marmalade somebody would be buying and people were really like excited. Oh you're really like you're lasering the customers and you really know everything about me. Okay. A couple days later there was Facebook and Google and all these players. But it was insane what people were actually how afraid they were about these things. But started my career there pretty fast. Knew that corporate is not mine. So I went to a company called Rocket Intranet where SRI Lauren Heidelman hired me. He's now running Project A as well. And I was building up the CRM departments of all the rocket ventures back in the day. So mostly were E commerce ventures, Dafiti in the South Americas, Linear Jumia, Lazada, Zalora, the iconic Lamoura in Russia. Back in the days Zalando, of course, all these things. Oh actually so the one thing why I'm so passionate about CRM is I did my PhD in basically CRM and we found out back in the days already that the value of a customer is driven by the retention time. So as the, the time the customer is with the company. So why don't we do everything around this? What I noticed actually also back in at the Rocket Times was all about quick sales. Quick sales, right. And then Sierra was sent out a newsletter, see how many revenues we're making. And back in the days it was actually viable. Right. Because we could just pour in new customers so easily. It was so easy to buy new customers and put them in the funnel. We are also doing work with Groupon. It was just insane. Like how many people like we're losing on the newsletter side just because we didn't do any great CRM but they were able to pour in new customers. Fascinating. Was actually that that CRM really was undervalued. And one of the reasons I thought was that back in the days it was the data platforms that were really bad. We had something called Mailchimp in 2012. I don't know if you still use that. Yeah, right. It's still big, right? It's just if you compare that to what Facebook and Google could do in terms of segment targeting was just a joke. I remember we were in the Rocket office and the people were on the acquisition side. We were running these kind of campaigns on these kind of segments like super interactive and what they purchased and what they did on the website and all this stuff. And it's like a gender. Right. We have two versions of the newsletter now. It's actually one template. It's dynamic. It was a bit like that, right? I was like that can't be true. Right. Back in the time at Rocket basically understood that first of all, CRM is totally undervalued in most companies. Actually not on the old, very old fashioned companies, the catalog retailers and stuff. They've been doing this for 20, 30 years, right. Because they wanted to optimize their catalog sending and stuff. But for the new companies it was just crazy how undervalued CRM is. It's not that case anymore. It's so cool what you are like presenting every, like every month of how companies, even startups are investing in CRM. And I think it's really cool to give them a good head start that back in the days it really wasn't that way. And this brought me to the, to the idea of Building a customer data platform. So platform we have all kinds of data in it's behavioral data, it's attributes around this. It's. You don't need SQL to get the data, but you can actually do this with a UI user interface. I don't know, back in the days it was like you had to talk to BI to get a segment and you would.

Back in the day.

Okay, yeah, you have some data and then you're going to be I do this. But, but that was the idea of building a cdp, right? Bringing the data to the marketer and actually make sure that we have meaningful segments in there. And another love that that I have is for customer lifetime value. I believe in that metric. It's very, it can be quite challenging to implement that. But on a basic note, I believe this is the one metric we really need in, in, in CRM to understand our efforts and effects on the value of a customer, not only the single transaction. And maybe why I'm so passionate about that is we once ran a test at a company that there was a CRM manager calling us at Cross Engage and he was saying like hey, management wants to send another newsletter. We already have four a week but they want the additional revenues. And they said okay, additional newsletter will give you additional revenues. So look, let's do this. And the CRM manager was really worried. It was, there was already a high subscription, complaints were high. They're just classical. Sales was dominating CRM. I don't know if you know this fact, like the sales department has a lot of power and they come to you and so that's in another one because they have these short term goals, right? We're going to understand that, right? But what we did actually we went and applied a CLV metric and what we did was we were predicting who actually should receive less emails. Right? And what we did, we then run a test and we had this one group who only got one newsletter that week and the other ones got four. And we used that with machine learning. Like we tried to understand machine learning and what we found out, these two groups didn't differ any at all, neither in the conversion rate nor the money over a certain period of time. For the simple newsletter, yes, we got some downturns but if you look, we looked at it for eight weeks, we didn't see any difference between four newsletters and one newsletter between these groups. It's just crazy, right? So it's basically what is the name is Tiamonica? I don't know, you just spread the revenues right across more Newsletters. That's what you basically do. And we unrevealed that using like a long term metric. And that was lifetime value. Basically. Lifetime value is a little bit misunderstood. It's not always the lifetime. You can also say what's the lifetime of the next 30 days? 30, whatever, six months, three months, whatever you need that to be. It doesn't need to be. That's sometimes a misunderstanding in CLV that it's like this one metric. What is Jessica going to be worth to whatever Zalander in the next 20 years? That's bullshit. Right? You can also look what's it going to be? What's our worth in the category shirts in the next 30 days? And that might also relate to your question, what you had early on in terms of understanding who's likely to convert on a coupon or should be receiving one. In the CLV metric you have two things. You have conversion likelihood and you have the amount the basket value. So it's basically like how probable, what's the probability of me buying next month and how much is that basket going to be? So if there's a probability that says for 99% Marcus is buying but only for €1, I probably wouldn't give a coupon, I would probably do a bundle to increase the value the basket. But if I see there's only like a 10% chance that I'm converting, but my basket will then be very likely be very big, then I'd probably give them a coupon because I need to push the customer with the clutch. And this is what I find so fascinating about the customer lifetime value. It's all in there. It's not one metric. You can just bend it and stretch it to certain timeframes to different categories, to subsegments, all this stuff. And then it gets very flexible. But it's also can get very complex. And that's what I've. Yeah, that's what I've loved doing for, for a long while. And that's a little bit about me, what I'm thinking about, what I do and I talk.

But as you were talking about customer lifetime value also what I noticed, we always talk about it, but I also asked like attendees in one round of the meetups, like what kind of KPIs are you using? And customer lifetime was mentioned 11% of the time. So what I see is a lot of companies struggle with implementing it and using it. What is your experience there? Why it's difficult for some companies to do that.

It's basically it's two Things. So the first thing is how immediate you can measure changes in clv. So if you have a good system, yes, you can do this quite, quite fast, right? You can see the increase in customer lifetime value if you send out a campaign. But if you don't compute that or you don't have the data to do this, basically the CLV thing is a bit of a long term effort. You have to wait three months to see whether stuff has an effect. And that's a problem, right? You tell management, right, wait three months and then we'll see whether what, what our million dollars of investment did, that's a problem, right? They want to see immediate results. And that's why these conversion metrics are so popular. Because you see, I put in 100 and I get out 200. Revenue doesn't matter, no matter the customer, right? But you put in 100, you get up to 100. It's so easy to compute the RI return on ad spend, whatever. So the CLV metric is very hard for that, right? Because if you don't do this right, it's a long term process looking at what really changes. You can actually go into like very deep in a CLV and immediately see the changes. But you have to be predictive. You used to, you have to use AI and it does a prediction and a prediction is inherently wrong because it's a prediction. But you can still, you can do this, right? You could get more immediate feedback, but it's tough. And the second thing is CLV is not this one concept where you say this is this one thing and it works for every company. I just was on a panel with of the marketing browser and I was talking to Elsa Dietz and she was at ebay Kleinzag. No, it's clients client. And she said this is so tough for us to model even B2B customer lifetime value because the processes there are so complex that I don't really understand what is my cost. What is actually the revenue? What is revenue? What is actually retention in this thing? It's so hard to do so because customer lifetime value, what is it? It's the lifetime of the customer to a brand and that what's the value? The value is basically the margin or the revenues minus the cost divided by some discount factor, right? That's always future. It's always future. So there are a couple of problems in there, right? Who's the Customer in the B2B? Is it a buying center? And I think you had this question earlier on as well, right? Do we know what the definition of a customer is. Yeah, a lot of companies don't really know that. So in a B2B setting is. Oh, no, it's like. It's Klaus. No, it's Jessica. Jessica makes the decision. Yeah, but Klaus has the money. So who's the customer? So it's very difficult, Right? Or think about something like Media, Martin. They have, they sell, I don't know, Bosch washing machines. So if I go into a store and then I buy this Bosch washing machine, am I a customer of mediamarket? Am I a customer of Bosch? Am I a customer of both? Who am I the customer to? So it's not that trivial to understand in many settings who the customer is, as I said, in the B2B settings. So that's very difficult. The other thing is, can I actually look into the future? Because a lot of people look into the past in terms of customer lifetime value, and that really doesn't reflect what customer lifetime value is all about. It's always to the future. So what you very often see is you see formulas that go into the path, which is average customer lifetime. So how long are people going to stay? What's the average margin? And that's about it. Right. And you just multiply that and that's basically your number. So you have one number for every customer or maybe for a customer segment, does it help you anything? Not really. Right. So it needs to be very individual. So you need to be a CLV for you, that increases complexity quite a bit. So it's a. So it's these things, right? So who is a customer? How can I look into the future? Do I actually have data? Do I understand how to model the future? What are my. Do I do? You know the. If you do a. If you do a sale, who's an E commerce, who sells goods, do you know for every customer what the product margin is? Basically there it starts right then. Now we get into the problem and where the, where there's the problem with clv because we're talking about margin and not about revenue. Right. Because that's inherently in the customer lifetime value. It's the profit. If we don't even understand what the, what the margin is you're getting from one single product you're sending out. Well, how can you actually compute a customer lifetime? It's not possible. Right. If there's a margin of 10% or 90%, that complete changes the lifetime value. And then you're back to okay, I look at revenue and then you're back to, okay, I look at the revenue of the newsletter this is like kind of the chain that happens in this thing. So there's even more complications. So it's about understanding what margins are, understanding actually how long customers are gonna be with you. For now, it's not that. So if you are Netflix, it's simple to compute whether somebody's still a customer. I'm just writing them, say Netflix, I'm out. Price increase. Didn't like it. In the best case, if you're in a hotel or E Commerce or whoever, it's transactional and you're leaving the hotel, how is the hotel supposed to know that you're going to come back? So how do I model? It's hard, tough, right? It's statistical modeling again. And then we get it complex and then it gets complex and that's the big problem. Clv who is the customer? It's different in different business models. B2B versus B2C. Then it's different in different verticals. Whether you have E Commerce or whether you have kind of subscription basis, then you need to be able to look into the future versus just going to the past. And if we add all this stuff up, given that it's a metric that if you don't use, it takes a long while to change. People are saying, I'm going back to revenue producer. And that's the problem.

I would say, is there any company type where you would really recommend to put the effort in to do that calculation?

Yeah, a subscription base is super simple. Yeah, it's because you have pretty steady margins, steady revenue. You can actually compute the churn quite easily, predict the churn quite easily and then you do subsegment CLVs. So you say, hey, for this cohort that we won with this campaign or this kind of viewers have this kind of cl. So it's super simple to then compare cohorts. But the, the, the, the measuring in terms of, in a subscription business is significantly simpler than a transactional business. So in a transactional business it gets harder, but you need data points. So if you have a loan, that's another complexity. Complexity. So if you have long purchase cycle, I don't know, it's who does furniture, six months or a year, whatever, you don't have enough data points to go there. If you have a business model that has a lot of data points there certainly makes sense. And you need past data. Otherwise it's tough, you can't really do anything. So again, subscription business check should be super straightforward. Never simple, but straightforward. If you have a transactional business, it's more than you need the data points. If you don't have the data points, you have a long purchase cycle. Super tough if you can't map any data. And that's another thing. You don't have a lot of data to run this. It's also like difficult. Like you can start CLV with just transaction data and gives you a good overview. But if you really want to do clv, you need more behavioral data. I give an example reactivation campaign. So very often we're sending it only on the transaction date. Right. So last transaction was 30 days ago, 60 days ago. But isn't there a difference whether somebody has put something into the basket in the meantime or not? Right. Two different. Completely different group. Right. The one was still engaging, it just didn't buy and the other wasn't. Wasn't even engaging. Right. So you can't model this into your CLV and your campaigns, then it's hard. So being able to track a significant amount of data is another thing.

Yeah, yeah. There's one question behind you. Do you have any thoughts on retail media and pushing for more or bad placements? Because I have the feeling that this often harms cov.

Yeah, look, a great question. And that's the. I think the biggest challenge that we have in CRM is we need to balance two things. The one hand side, we need to balance the salespeople who want to make sales and who need to make sales. And they're usually incentivized on making sales. You need to sell off these many units. Right. And that's when you get your bonus, Jessica. Right. And then I'll just set it on a campaign. Sounds good. Right. So this is what the salespeople do. Right. And then they push. And on CRM we have the responsibility of maintaining that relationship. And this is the big problem. And now this is sometimes the problem. Why CRM is not always on has the decisive because sales and says but we need to sell another 50,000 units and otherwise we won't make our business plan. So yes, it does. I think it's a constant struggle to find the balance around it. Was it. I think this is one of the bigger challenges and I see this with a lot of companies when I do consulting as well. It's the balance between, let's say the customer experience and on the other hand side, the push to sell something. And I think this is one of the big challenges that we are having in CRM to make sure that we can maintain this customer relationship. Because acquiring new customers has become very expensive. A lot of companies have Understood that when I get into consulting, probably say, now we have all these customers, let's make something out of this, right? Let's send another newsletter. That's sometimes the answer, in fact, but this is the balance. And so you're absolutely right. Very often it's not. And the problem is the systems are not connected. And that's actually a super interesting topic. How this is going to unroll for the next couple years. How are we going to use AI and CRM if we don't have any connected data now? How are we going to make sure like a customer journey is, let's say AI optimized? Because we can't. You've done great work in customer journey mapping, but it's also like a lot of work for it. You have to reduce to certain main journeys. It's not like super individual. With AI we have the opportunity to create a great holistic customer journey. But if we don't have the data and there's some third party provider who sends an email to you that screws you up and you like, I'm done, you're annoyed, then we can't optimize that. What I did in my last project was I the success metric between the CRM departments and sales was a balance between the sales traction that they had. So the sales that we were making minus a certain cost factor, times the number of unsubscribes, minus the cost of acquiring new customers because they were gone if they were unsubscribed or they were churning so they could push hard in the sales department. But if they had more like a lot of unsubscriptions and there was like I think $20 or €20 that they said was an unsubscribe, so if they push too hard, their metric would decrease significantly. So even though they made a million, but they just lost 800,000 customer potential, then they get penalties for this, right? And then we finally can compare like how good these sales units are taking care of the customer while still making money. Because then you have a ratio because you have this 1 million that you made, but effectively you only generated 200,000, right? Because of the lost customers, then you have an effective rate of 20%. And then you can just compare that and say, hey, sales unit A, why is this 20% and you have 80%, what are you doing different? And then they start talking because they get the penalty. But if we go to them and say, hey, with this customer relationship and let's important that we are just maintaining that, otherwise they get pissed off yeah, let's send another newsletter.

A lot of companies are talking about synthetic profiles, which is really crazy. So basically you feed the AI data on your customer base and then they create these profiles and then you basically perform your market research and your customer experience queries on the profile.

Yeah, I think it's a perfect bridge because I also wanted to go into the AI talk with you.

I love it, love it, love it. I'm a friend of short answers, so let me show you. So it's a great question. It's a really great question because all the advancement that we are currently seeing in AI is built on top of the amounts of data that we are currently seeing, especially the large language models. Right. So the large language models, they currently saying we've already been trained on the entire Internet, we need new data. It's crazy, right? So this is a problem for them to get new data. They learned the entire Internet. It's crazy, right? And pictures, right the way the picture. There's so many data around this. Now look at CRM and look at what, what has happened since 2018 with GDPR. We're all concerned about how long we store data and not only the GDPR legal side, which there is a case for, but also still technically the data is like spread around and it's a different system. So the last consulting project they had sales cloud from Salesforce, they had the marketing cloud at the service cloud. They didn't really know what customer was actually which and they couldn't really match that.

Although that's what they sell.

Right. Everything is connected. 360 degree user view. Right? That's. Yeah. This mystery of 300, 360 degree user view. So, so I'm trying to. I'm wondering how we're going to benefit from the advances in AI and CRM if we cut off our limps all the time. And that's data to me, there's still too big of a problem in terms of the availability of good customer data and having this in a great location, understanding what data is there. And my last consulting project began, they're like they had huge amounts of data but they didn't know about it. So one unit had this and the other one unit had this. At this first thing is I think this is a huge problem in, in, in CRM and I think the discipline of CRM is going to suffer. Not on the content optimization part because there we can use picture generation that's easy text generation like companies like Neuroflash and whatnot, who like Ink, Product Description blah, blah Blah, that's soft. But on the segment side it's really tough. And now I think the only way to go forward is actually synthetic data. But the problem in this with these companies is now how are you going to, how are you going to produce synthetic profiles if your base data is not good? If you don't have any, let's say seed data, that's really good. If you had 10,000 customers which are like really great and you understand exactly what you did and it has this full profile, I'm pretty confident that you can build a lot of synthetic profiles. Right. So what is the problem actually in this? So the AI models, they usually use these neural net neural networks. Looks like little brain, right. Before that it was machine learning, it was logistic regression. Like super simple things. You have all done regression in university, most likely assumptions in there. So mathematical models, super simple to use. AI is using these neural networks which need these huge amounts of data. They need them, otherwise they just spit out crap. You can actually still see that a little bit. On the GPT3 versions, German was still very bad. So I wrote a poem for my mom. She's the best mom. It was still awkward, but in English it was perfect. But he doesn't speak any English. So I had to take the German version and make it up. Why? Because of the availability of the data. Right. And the training data. So long, long story short, for these new models, you need these huge amounts of data, you will need synthetic data. But if we still don't have any base data, that's super correct. I don't understand. I don't know how to extrapolate good profiles from a shitty source. And that's where I'm still struggling with, to be honest. But I think the biggest challenge in CRM for me for the next year is if we don't want to get stuck like all the other disciplines are really making huge progress. Robotics, logistics, self driving cars in the factory, all the. All along the value chain. Why? Because there's steady stream of technical data that's coming in all the time. Super standardized is super. Well, just us in kind of the customer data field. We're just chopping up our limbs all the time. Less data. And I was working, we had a prediction engine in our software and we needed at least two years of data to reliably make predictions. Two years of data. But every customer think about this, right? Do you guys have that? Who has two years of past data? That's not too much, right? That was only two, right? Just think about seasonality. If you really want to model seasonality. You at least or five at least. Not only two years or three years, but. So there we go. This is the biggest problem, I think the biggest challenge that we have in CRM. And we need to work on this not only on the technical side, but also I would say in the cultural side because I think and I again disclaimer. I like data privacy and this is all fine, but the, the power that I see that some of the data privacy officers have gotten in their companies and they're just, we're staying safe. Please, no, we can't really do this. And please do this. Four checkboxes for every channel. You've got this long list and it's. Come on, combine this in one checkbox, please. And then it'll have all the consents. No, this is the small things, I think what we need as CRM where we need to make sure that we are pushing a bit on. On our side. Otherwise we it's how is CRM going to look like? Because what we actually want in CRM isn't it like that? We want a machine that tells me hey, there's the journey you should actually look at. There's a customer segment that's underserved. There is some certain behavior that you might need to trigger. If somebody buys this or that, then why don't you react this way? Isn't that the way we want to work with software in the next coming years would be cool, right? That's fine. But why wouldn't you do this in a. You could still do this in an analytical system right? Now, what would be the benefit of AI in this aspect? Don't get me wrong, I'm interested. Okay. I think the interface. I think you're absolutely on the right track. Like to wait to talk to a system like this and not doing this technical segmentations of has purchased three times but not in the left. You know what is kind of cool? And then we come back to the problem, right? If you need the base data and the base data needs to be very fine granular otherwise the model cannot look at the pattern. Right. It needs to see a lot of these incidences. And that's the issue. That's exactly what I'm trying to point out. Otherwise I can see a bright future, right? You just give you. I was always dreaming even across engage about this the system to the marketer. I log in and it says in the morning, Marcus, these are these three things you need to do today. And if you've got this, you can go home Basically I was. That was my dream. But it was like, it was very hard. It's very hard. Usually not because you don't have the theoretical model or you. But the data wasn't just there. Still not there. Even in our platform. Even though we had behavioral data, we had attributes, but still didn't have this across. We didn't have the call center data in, which is important. And then actually the acquisition data would be interesting as well. Which channel did this come from? How much was the. What was the cost of acquiring this customer? Blah, blah, blah kind of stuff. But it was still not there. The potential is there now with the machine. Like the AI is there, but the data is the problem.

Should we cut it off here? I feel like the data is the.

Problem.

That we definitely need to work on that. So thank you so much for sitting here and also answering those questions. I think we could talk for hours here. Was really great listening to you.

Alright, that brings us to the end of today's episode of Inside CRM. I hope you enjoyed our chat and picked up some useful gems to take back to your own CRM toolbox. Thanks so much for tuning in. If you liked what you heard, do me a favor and subscribe to the podcast so you never miss an episode. And hey, if you've got a quick sec, check out our website for more great content and updates. Until next time, I'm Jessica Jansen and you've been listening to Inside CRM.

Take care.