Andrew Michael
Hey Carl. Welcome to the show.

Carl Gold
Thanks, Andrew. Glad to be here.

Andrew Michael
It's great to have you today for the listeners call is the Chief Data Scientist at Zuora and author of the book Fighting Churn With Data. Zuora creates cloud based software on a subscription basis that enables any company in any industry to successfully launch manage and transform into subscription business call is also the creator of the subscription economy index and responsible for data science behind Zuora's subscriber insights. He has a PhD from California Institute of Technology and authored publications on machine learning and neuroscience practice Oracle spent most of his post academic career on Wall Street as a quantitative analyst. So my first question for you call is, what is it about churn that interest you enough to write a book about churn?

Carl Gold
Well, um, it was a it's a little bit of a circumstantial thing. Just that I've worked on turns so much over the years, I felt like honestly, I had made so many mistakes in the early years, as a data scientist when I was trying to work on churn. And now I feel like I've actually learned how to do it properly. And I just look back and think, man, if I could have told myself five years ago, what I should have been doing, it would have been great. So that's why I want to write a book to, you know, help other people kind of avoid the mistakes that I made. And of course, you know, do a great job with churn because it's obviously a really pressing problem for so many, many, many companies and products nowadays. Absolutely. And I think as well, like a lot of people see it as a problem. But also when you see the flip side of retention, it's actually a huge opportunity, and it's probably one of your biggest like drivers for growth.

Andrew Michael
So today, for the listeners, we actually have a special treat. So call it will be giving away five of his books after the show. And what we'd like to do is just encourage anybody listening to the episode if you'd be interested in grabbing a copy of this book, and obviously we'll dive into details now as you get into the show. Make sure to share this episode on Twitter. And then we'll be pulling out those tweets and having a random draw for those five books for those who ever share this episode within the next couple of weeks, so we'll be doing the draw in two weeks from this episode airing. So yeah, so you mentioned like, it's happened to be circumstantial as well. And I know, you did a PhD and it was on machine learning and neuro science. I think you did neural networks, as was your case here. Is that correct?

Carl Gold
Well, somewhat my PhD thesis was really in a biological or biophysical modeling. So realistic modeling of neurons, not for AI, but for neuroscience, and understanding. But I was in an interdisciplinary program at Caltech, where we spend a lot of time doing machine learning and brain science together.

Andrew Michael
Interesting, so I'm pretty sure you probably read raychel's, what is it right causal? Right. Um, yeah,

Carl Gold
the mind. Yep. And don't don't get me started on the neuroscience stuff, because I'm actually really skeptical about those kind of claims because I spent time actually studying brains and working with brain implant technology. And I would never take one of those things in my brain. Absolutely not.

Andrew Michael
Absolutely. Yeah, definitely. I think it is very futures. It's a totally different topic. But yeah, I was just interested, like, going from going from that and studying sort of the brain itself and then moving into sort of a career on Wall Street and then, like chicken essentially ending up focusing like subscription.

Carl Gold
Well, I didn't know maybe I shouldn't say it's just a circumstance. It's really just the trend of the times because I started working on a churn problem. Something like five years ago, when a friend asked me, you know, about a startup startup that he was involved in, and then since then, You know, in coming to Zora, it's just a problem for more and more people. And so I became, you know, as Zora, obviously, we have so many customers who are, you know, have, they all have subscription companies, and they're all very concerned about churn. So through my work here is how I got to, you know, work on so many different churn studies with different customers to give me enough knowledge to write the book, but it's really just a sign of the times, you know that, I mean, in a way, the person in the right place at the right time, but it's just an overwhelming trend, but everyone's concerned about churn. But I think your point was really good about you should see it as an opportunity because it's about the glass half full versus half empty. When you're trying to reduce churn, you're really trying to increase the customer engagement and satisfaction with your product. So one of the main themes of the book is how you use the information that you collected churns to find what are the most engaging aspects of your product.

Andrew Michael
Excellent. And that's what I want to dive into a little bit deeper now as well. So you've given us a slight overview with the books about it. Maybe you want to just give us a little bit more of a synopsis quick, so we can dive into few aspects of it.

Carl Gold
Sure, let me hear, here's how I can explain it with Well, my own story is that when I started doing chernykh analyses, I had the I mean, I was trained as a data scientist, you know, I had a machine learning background. So when you're that, when you're trading that way, you think the answer to everything is to make a predictive AI model or like a machine learning model. But I have a lot of failures in getting any traction, you know, doing that kind of analyses for companies, because what would happen is I make you know, some kind of predictive AI model, and then no one would really use it because what I learned is different about churn from other problems that people apply AI to, is that there's no one size fits all solution to churn. So if you think about an AI problem, like, you know, spam emails or something, if you detect a spam email, you know, you put it in the spam folder, and you're done. But if you think someone is at risk of churn because of your, you know, your prediction, well, what do you do? There's a lot of different ways to reduce churn and you need to, you know, use appropriate one for each customer. So that I realized was why no one wanted a predictive model that just said, Who's going to turn what they really need is to know the details about customers in a way that they can use it to to drive action. So what I found people what really do, I mean, people we can talk in a minute about you know how people are fighting churn. I'm sure you've covered it in other podcasts. But what I found is that people come in company These are really using customer metrics to, to segment customers for churn, they'll look at, you know, some metric of their customer and say, Oh, these people are not at risk. And, you know, these people are at risk. But what I found was that a lot of times the customer metrics that companies were using were not really well designed for this purpose. As like a data science person, you would, I would typically get the customer metrics, that would be the input to the predictive model. But then I would discover, hey, those customer metrics, were not that good in the first place. And no one's really using the predictive model. So maybe I should focus on making better customer metrics. So that's actually the main theme of the book is how to use your data about churn to make really great customer metrics. That really and a really great customer metric is going to really be associated with the value that people receive See on the product. So when you segment customers by a great metric, you really, you really are segmenting them into the groups based on how much value they receive. And then those kind of metrics can be used to make you know, your interventions to reduce churn, you know, more data driven. Absolutely.

Andrew Michael
Yeah, I think what you're saying it makes total sense. And it's definitely something like I've come into and been guilty of in the past is like thinking that a churn prediction model is sexy, but then when you really think about it, what are the actual use cases of it is mostly really trying to save a concert already at risk, as opposed to and it's very reactionary, so like, you can only really act on it after the fact as opposed to really understanding like, what are those key drivers and key inputs that are going into that that you can actually try and encourage and work towards So you mentioned these company and customer matrix sec. What do you advise is like a typical process then for company really trying to get a grasp on this in We're trying to understand what they should be tracking to move the needle on turn, which is a lagging metric, but how do you suggest they go about really trying to understand what these metrics should be?

Carl Gold
Well, okay, how to go about it is really, you know, start in chapter two of the book, and work your way through, but I'll try to, you know, put it in a nutshell. I mean, the, the techniques I teach, assume that you're already tracking interactions with customers. And that's, it's not really covered by the book, because that's a technology you know, you you can buy, you can buy that from many different companies, you know, things that will allow you to track and I want to also clarify that when I say tracking, I'm talking about usage of the product, not, you know, track stalking people on the internet, right. So everything I talked about does not involve personal information at all. It's only involves one And measuring, it's really measuring your own product, right? How many people are coming to each page, if it's a website, how many people are using each feature, if it's a SAS tool, or how many people are watching each video, or you know if you're a streaming video service. And so those are your basic events that you should be tracking. And in that case, the most important events for tracking, it's different for every company because it always has to be the events that are most closely associated with getting value. And that's different on every product. I mean, there are some general General, general statements you can make. For example, I mean, logins are almost always one of the weakest measures of engagements because well, someone logs in, they haven't accomplished anything, right? You really want to track what do people actually accomplish on your product and if it's, you know, if your product is for making documents you might track document creation. If it's for say streaming media, you get into this thing well, watching a video is like a good event to track but even better would be completing a video because that means the person actually, you know, liked it. And then if you track likes, then tracking likes of videos would be even better. So all those different events and interactions you can track form the foundation. For what? For what comes after that part makes sense? Oh,

Andrew Michael
absolutely. I think it's something we might have a question. Yep. It definitely does. I think like early on, we interviewed Heidi Gibson from GoDaddy. And I love what you touch on now is really about measuring value. And I think in some cases, it's really obvious and really easy to do you know, if somebody is coming to YouTube, for example, to watch a video like if they've watched the video, the video, the venue is clear there for the startup, so it's not as clear cut as well. I think Like, for example, a website builder, like GoDaddy itself is one thing. Yes, it's you want to build a website so that you can say, once you built a website, that's achieved part of the goal. But really, you're not building a website to have a website, you're building a website to drive sales to do close bookings or, and really, it's sort of the end goal and the end value. And the closer you can get to measuring and tracking the actual value that your customers receive from your product, obviously, the most powerful it's going to be in terms of a predictor of if people are going to stick around or not.

Carl Gold
Yeah, absolutely. So some products, it's going to be easier than others. So it's a good example, you know, the website builder, you know, GoDaddy or any company like that, you know, creating a website is one thing and you would want to track that. But then also great metrics for customer churn would also be the website traffic, you know, how much how many people come and if you can even measure transactions. You know, the That would be the holy grail, but not, not every system is going to allow you to directly measure the value that customers receive that way.

Andrew Michael
Yeah.

Carl Gold
But with the you know, using the techniques of analysis, basically, what I show you is you can turn almost any measure of customer behavior into a measure of customer health, simply because what you almost always find is that the more customers you know, do different behaviors, you'll you'll see a relationship to turn that usually plateaus. So it's pretty easy to see what the healthy state is. So almost any metric will can be common indicator of customer health, but then you have to remember Okay, is this metric really directly causing engagement or is it just correlated with engagement? So I mean, this is a question of correlation versus causation is well known. There's no escaping it. Yeah, but but you can work around it is the thing. I mean that the book, I really don't dwell on that at all, because you can use the important thing, metrics for churn. I mean, here are the important things that you need what I think to make a customer metric really great. It should be really related to retention and churn and you should be able to see it in your data pretty easily using the techniques in the book, a great customer metric. It also needs to allow targeted interventions by the people who are trying to reduce churn so that it should allow people to, you know, segment customers into healthy and unhealthy groups for the purpose of interventions. The metrics have to be easy to understand because generally it's not going to be a data scientist, or you know, a technical person trying to reduce churn, its people in marketing and customer support. will say more about that in A minute. Well, that's actually the last point actually about, you know, good customer metrics. A great customer metrics is If more than one department can use it. So for example, actually, a good example is, you know, churn forecasting or churn prediction. The risk of churn is a good metric, but it's really only appropriate I think, for one department, which is like a customer success function, because they're the ones who are going to, you know, reach out to a customer who's in trouble. So naturally, they could use a metric of you know, who's in trouble. But customer metrics aren't the type Prevention's I advocate aren't just those kind of one on one interventions. I mean, it doesn't have to be reactive, it could be like you know, it is kind of proactive if you reach out to them. But the thing is, before even the customer you should think about the product because the best way I think, and I know you said you said similar things on your other podcasts. But the best way to reduce churn is to have a better product or a great, you know, a great product. Yeah. And you can you can use the customer metrics that I'm talking about to really figure out in a data driven way, you know, what are your best features, and who's using them, who's not using them, which is so then you can make look alike features or you might just want to, if you figure out one feature is really driving engagement, you might want to make it more prominent, you know, make it easier to find if it's not already easy to find. But then when it comes to people who aren't using your great features, that's where your marketing department should get involved. And they should be sending out occasionally campaigns to make sure that people know about all your great features. I mean, talking so I'm talking about campaigns to your customers not for sale. So this is marketing, working on, you know, like a customer success project. And but when you send the campaign to get people using the great features, you don't just want to spam everyone who has a high churn risk, you really want to target the people who aren't using the feature. So that's back to the exact, you know, the churn risk forecast. It's not usable, really by the other departments. It's only usable in customer success, which is still good. You know, so I'm not completely trashing it. I'm just saying a really great customer metric, more than one department, you know, would see a use for it. So

Andrew Michael
yeah. And that definitely makes a lot of sense, I think, because I think there was I was alluding to the beginning, when I mentioned in terms of the churn model, the team that's most useful is going to be the customer success because they can then go in and try and save the account. What would be like sort of the alternating view then on that and would be a metric that would potentially be scale across the team, would you think about looking at potential accurate tension prediction model? And is that something you've looked into? Or do you find that it comes to the same sort of conclusion?

Carl Gold
No, I mean, no, it really comes down, I think, to, you know, good or great metrics. And if you just take a simple metric, like let's say, you know, let's say your product, I mean is for the, you know, editing documents or something like that, or different kinds of documents. I don't know, I'm just making up an example off the cuff. Um, if you have just a metric of you know, how many documents per month customers making like in different types. Then, once you do the analysis that I show in the book, you can easily see what's a healthy level for each type of doc. It maybe it's like 20 documents a month, is where customers look healthy on that metric. Well, so then, you know, you know, what you're driving to for the marketing and engagement campaigns and also the customer success team. has a clear goal to because you tell the customer success team, oh, these counts are risk of churn. Right? Well, then they're like, Okay, what do I do for these accounts? So you're just back to the same question. You need metrics with a known kind of level of health, unknown, you know, point a known target for, I need to come up with a better term for this. But I think we got one going, for sure. So usually, I mean, the churn risk forecast can be good, but it's always in combination with other things that will tell you how to act with that particular customer, or what group or not well, you will segment the customers you'll almost never do anything, you know, one by one, but you'll segment the customers in a group that could benefit from a certain kind of treatment, you know, maybe a maybe a class you know, it could be like, you know, some kind of webinar on how to use the product better. Yeah.

Andrew Michael
I think as well like the thing with the churn prediction side of things, is it Looks to try and find and identify the behaviors of users that are about to churn or that are at risk of churning, as opposed to again, like going back to the the alternate view of really focusing on what are the actions and actions that users are taking that are retaining and sticking around. Because when you think about sort of, like churners, and output metric, and it's definitely a lagging indicator, and like actions you take today could only be seen maybe 612 months from now. Like, it's really about understanding what are those input metrics, and I think that's what you're alluding to, in terms of, like the the value side of things with figuring out like, 20 documents. And yeah,

Carl Gold
well, there's an important trick with that, actually, that's that I teach in the book, and I'll explain it as best I can. The podcast, you mentioned like, you know, turn is a lagging indicator. So one of the key techniques I advocate and teach in the book is that you actually pair observation of church With observation of the customer from a time before the actual churn took place, because typically, if you think about it, if you look at a customer like the day before they cancel, their behavior has already changed, right? They might have stopped doing using many features. And maybe only using a few features that are relevant right before they quit. If you have a pattern like that, it would actually make it easy to predict who is going to turn on the day before they quit, but by then it's too late. So the trick is actually when you do these analyses is to pair the observation of the churn with metrics from a time significantly before the turn well and the time before, hands on the product type. If it's a monthly product, you know, month to month, then you would probably look at, you know the metrics right after the last renewal or maybe one week After their last payment in three weeks or four weeks before their next payment, and then you would observe customers, you know, three weeks out and then look Three weeks later who actually turns and that helps to control for that, you know, lagging nature of the churn for a product with the annual subscriptions, like a SAS product for businesses, and this is how we do it at Zora, we, you look at the customers, typically around three months before the renewal, and that's the metrics that you're looking at. So hopefully that three months before renewal is a time when the customers haven't made up their mind yet, they're still you know, evaluating before their next renewal. And that's a that's a technique. It's really kind of a trick but I mean, just, you know, a tech need to make your analysis more about the leading indicators of churn, although you can never perfectly you know, control for that problem.

Andrew Michael
Interesting. So what you're saying then just target understand this clearly. So if we take the b2b, SAAS and you mostly yearly plans, you're taking a look at the yearly, but at the behavior that they show three months before that renewal is about to happen. And then are you looking at those returned and those that didn't and comparing the differences between the two?

Carl Gold
Yeah, I mean, that's the that's the essence of the method. Yeah, look, look a good time out from the renewal. You got to include both customers that renewed and those that didn't, so that you're comparing them? And I mean, of course, I mean, this is a you know, this is a book on data science and programming methods. You're not in the book, I'm not going to tell you to like compare the customers one by one, your cohorts and compare the cohorts. Yeah. But again, it's all cohorts and people who are in an annual renewal case, it's all cohorts that were three months From their renewal at the time that you measure them.

Andrew Michael
Very interesting. And would you recommend like sort of focusing on this sort of an analysis as opposed to really looking at customers who were still with us in the second year and seeing what behavior they had versus customers that weren't? So what would be, in your opinion, like the difference between the two views of one sort of just before training, and one actually after turn has occurred, but also focusing really on the behaviors of those that are still around? Yeah, well,

Carl Gold
you can do different I mean, additional analyses, for example, you know, you could make do one thing only on your one customers and one thing only on your two customers, I usually don't actually do that. I do include the length of time that someone has been a customer, as, as you know, another metric that you can apply this thinking to and what you'll typically See, everyone knows this is that churn falls over time. Well, for most companies, I should say, turn decreases over time. And it's not always bad if it doesn't, doesn't, that's another story. So I mean, the length of time they've been a customer can be a separate dimension of this kind of, you know, analytic process. And also, yeah, you can divide customers into different groups, but I usually don't do it because honestly, I find that the behaviors itself are the most telling thing. And if you have a customer who's, you know, six months in versus 18 months, and if their behaviors are the same in terms of you know, the value they're getting from the product, usually doesn't make that much difference that one is, you know, one year in versus two year and, of course, the one problem with you know, looking at the relationship between how long they've been a customer insurance is It's not particularly actionable in the sense that you can't make someone be a longer term customer other than by waiting and having them not sure. Right? Yeah, it is useful. You might see peaks in churn risk, like, you might like for a monthly product even you might see that risk dips in the first year and peaks around one year from sign up. And that can be actionable information that you know, you should be reaching out to people you know, right before their one year anniversary, maybe get them in a webinar or a training or anything, you know, but yeah, the length, but the length of time the relationship between the length of time insurance is something that people obsess over from a financial perspective, because the profile of churn versus you know, some fiber lifetime is it's important to understand for finance, but it's not actually as helpful for you juicing Sure, in my opinion?

Andrew Michael
Yeah. So that makes a lot of sense. I think from the yearly perspective when you think about sort of the monthly and there is some things we're probably saying the same thing but just in slightly different ways is like when you think about sort of monthly SAS business and you have a customers been paying you for like nine consecutive months, which is typically a higher than your LTV or your current sort of average lifespan would looking at their behavior. So this is what I assume in terms of like having somebody who's in quote unquote, a successful customer because they've stuck around for a long time and they still continue to pay and use your service. You would still say okay, it's much better just to focus on the actual behaviors that users show to turn as opposed to looking at like what successful customers look like as well. Or would you want to be doing both.

Carl Gold
I'm also an advocate of simplicity in turn analysis in the sense that That I mean, the most important thing for most customers is just to make metrics, get great metrics for your customers, and get them in front of the the churn fighters, which are like the customer success, the marketers, the product creators, right? They're the ones who are going to reduce churn any information, which is usually just, you know, the customer metrics. And when you try to get into more advanced stuff, like, oh, the first year customers versus the second year customers, they might not, they might not end up using it. I mean, it's one of those things where Yeah, if you're Netflix, you should definitely do a really detailed analysis of everyone every year, because you also have a data science team of thousand people, right? But if you're a smaller company, and you're the only data person, just focus on getting great customer metrics, and don't don't worry about a hairsplitting between first year, second year, you know, get the metrics in the hands of The fighters make sure they know what's a healthy level for the for those metrics, which isn't hard to do, and then let them do what they're going to do. Because, you know, the marketers and the customer success people, they all operate with different different constraints, you know, different opportunities, depending on the kind of product you're on. So it's hard to generalize. Yeah, that,

Andrew Michael
I think, I mean, like, the thing that just sticks in my brain is like this is because it's sort of like, again, the back thing focusing on the problem as opposed to the solution. So like when you're focusing on sort of their churn event and trying to figure out how to avoid it as opposed to focusing on a retained state and encourage it so like, that's why for me exactly. Scrubbing is like a little bit counterintuitive, in my mind, at least, but I'm slowly coming

Carl Gold
around. Well, I think No, no, no, but yeah, my approach definitely is to focus on the engaged state, you know, and look at you know, what, what do people look like in that? stayed and try to make customers like that. Okay,

Andrew Michael
so that makes sense. But then maybe let's talk about a real life example now and it's war. So maybe you like if you can be a little bit specific in terms of the metrics that you the good customer metrics that you've come up with for the team and how the different teams are using those metrics.

Carl Gold
Sure. Well, maybe probably better not Zuora. But I can talk about the case studies that are in my book, there's three companies that are agreed to be case studies in the book. One is named broadly and it's a it's a SAS product for managing a business's online presence. Another one is called clip folio and clip folio is a dashboarding tool for metrics. I definitely recommend you check out broadly and clip folio and a third companies versus sure their telco in Canada telecommunications over the internet. So let's just take one of those examples, you know, like, say, the the clip folio, it's a it's a dashboarding tool. So your basic metrics there are going to be just how many? How many dashboards? You know, are they looking at how many dashboards Do they have. But the truth is, those are only kind of basic metrics and simple counts of how much people use of things only go so far, because you end up with a problem with a company like clip folio, or any SAS product that you have bigger and smaller customers. So you have big customers who have, you know, 1000 employees and they buy 50 seats, and they are going to look at a lot of dashboards and you have small customers that have you know, five employees and they buy two seats, so you know, selling by the seat. And so just looking at the number of dashboards, or the number of you is limited because you can't tell the difference between, you know, a small company that looks at a lot of dashboards and a big company that's not using it. So then you get into more advanced metrics. And a good example there is license utilization. So that's what you get when you take the number of active users and divided by the number of seats sold. And that's an even better metric. Because, you know, it applies across these different sizes of companies. Yeah. So to make you know, more examples like that, so have like, you know, a good metric and an even better metric. Well, for example, for the telco versus Sure. In the book, you see the more calls customers make the less they churn, right? Yep, that makes sense. Now, let me tell you something surprising right here. I'll ask you this one first. If customers pay more in terms of like their monthly fees. Do you think they turn last or turn more?

Carl Gold
Yes, it's almost always customers who pay more churn less. Yeah. And that makes sense, especially if you think about it in the SAS context where you know, a b2b customer, if they're paying more and means they're a bigger customer, they have 100 employees, instead of, you know, five, and a bigger customer, a bigger company is going to churn less well, for several reasons. One, they're less at risk of going out of business. And to they're going to be more invested, you know, in the product that they bought because they they're paying more they had to train more people.

Andrew Michael
Process security, all sorts of things to get things approved. Yeah,

Carl Gold
yeah, exactly. So once you get a big customer, they're going to turn less. But so that makes a problem with interpreting pricing. So here's where a great customer metrics come into pricing. Instead of looking at just the price customers pay In recurring revenue, look at the unit price they pay so for versus sure that metric would be the cost per call. And that's not a true unit price because the product isn't sold by the call, you know, it's sold with a monthly recurring fee. But you can still calculate the price per call that customers make in that scenario. And there, you actually see a really great metric and a strong relationship to turn where the more the higher the cost per call, the higher return period. So that's where you actually see the truly impact of your pricing. And that's really powerful, because I mean, that a metric like if you look at, you know, a cost per unit cost metric and see what's the healthy level and what's the not healthy level. It's really powerful across all levels of your customers. And it's also really great information for your pricing. We didn't even talk about right sizing pricing, but that's a way to reduce churn. So, yeah, I'm I advise people against discounts for reducing churn, they generally don't work. You use discounts to get people to sign up. And if you're giving out discounts to reduce churn, you're actually really you're undermining your pricing

Andrew Michael
and your neck tenure brand.

Carl Gold
But you but there is such a thing as how if you have like a good, better best pricing structure, like a basic standard and premium plan, it does absolutely make sense to get people on the right plan. So it's not giving someone a discount. If you downgrade them from you know, standard, the basic. It is revenue churn, so it's still a form of churn, but it's a lesser form of churn. But anyway, that a unit cost metric is actually if there's one thing if you're listening to this podcast, and you're like, what's one takeaway that I can apply to my business, calculate a metric which is like a unit cost and it's Almost always a very powerful metric for understanding engagement. And I think people really think this way because like, you know, when I'm thinking, evaluating my Netflix subscription, I'm really thinking, Oh, how many episodes or movies did I watch last month? And if it's like one, I'm like, Damn, that was an expensive movie. But it's pretty. I'm like, wow, I'm getting a great deal from Netflix, right? Push me. So people really think that way. And I think I mean, that's I'm not a psychologist, really, even though I studied neuroscience. So definitely unit cost metrics. Again, that's like taking the metrics to the next level. It's not a basic metric of usage or what they paid but it's a you know, a ratio of what they paid at what they use. Yep, that's really powerful for for segmenting, and it can drive action, you know, in multiple, multiple parts of the company from understanding that one metric.

Andrew Michael
Very cool. I love both examples. Though as well as sort of like proxies, and you mentioned the point as well, like some businesses have large companies and small companies using this service. And it's not always easy to distinguish. And as specifically most with me when we think about privacy and not having to be able to enrich data and pulling from different sources are really love the way that you sort of be able to figure out like, Are people getting value out of the service and using the idea of licensing versus utilization, or in this case, as well like sort of the cost per call? I think it's really really insightful.

Carl Gold
Yeah, thanks. Yeah, I love it too. That is turned into a really interesting area. Honestly, I have no idea when I first started thinking about churn that I over the years, you know, find so many interesting angles on it.

Andrew Michael
Absolutely. Yeah. Because I think it's like I'm also thinking about in the context I currently the business intelligence team and how john and for us as well, like we're an all in one solution. So we have any legs and feet. Tools. And we do things like heat map session recordings, like polls, surveys, and so forth. And get one, it's not always easy to distinguish behavior, like you say, between like a large and small company when you're just looking at the data and the numbers without having that reference point. And then two, as well as like, how'd you establish like, are these companies actually receiving the value that they're looking for you and like, can they justify, again, sort of that price? So this is definitely something I'll take back to the team to sort of try and see which areas we can maybe make our metrics more. I can say like stringent and focused. So next question I want to ask is sec typically ask this question to everyone on the show and I want to hear your input as well. So if you had to start a new job or a new company, and you arrive and you see Turner attention is not great? And the CEO is giving You the job to try and help turn things around for the company. And he's given you 90 days to try and sort of show some results for the company. What would be your process? Where would you start? And what would be one or two, the first things that you would action?

Carl Gold
Well, honestly, I don't know if this will be the best, the best advertisement for the techniques in my book, because the first thing I would do is just look at the backlog of all the things that, you know, you're trying to do so and when you look at the backlog for a typical product, you might find things like, oh, the Android version doesn't work, you know. And you have to make sure there's no real bombs in your product that are, you know, just torpedoing it. I mean this stuff about great customer metrics and making sure everyone's getting value. It assumes that you don't have any really obvious fails going on, but most comes We'll know about these from QA and, you know, customer feed back and tickets, and they just won't have done something yet. So because really always the best way to reduce churn is a great product. And if you have some situation where like, you know, one of your mobile versions doesn't work or something, or, you know, you gotta you gotta deal with that before you worry about. I mean, no one's going to get value if they can't even get the app installed. Right?

Andrew Michael
Absolutely.

Carl Gold
So and first 90 days, that's pretty tough to say, to actually make a change insurance, because it really depends on you know, what are those, you know, one of those product gaps. Now, if you don't have major product gaps, and you have a significant churn and you have a churn problem, then it probably has to do with competition. And that's where you need to really then you really do need to get into the metrics and I mean if i if i if i Found insane my first month at this job that there were no serious product gaps or things that were just like obvious, you know, killers, then I would, you know, turn to the techniques I advise in the book, which is, you know, look at what events are being tracked, you know, make sure they're suitable, make, make customer metrics from those events and you know, see what, see what you learn. And then you can start planning the interventions. It could be, you know, product improvements, or changes in the layout, or it could be, you know, things like engagement campaigns. So, in that scenario, you might be able to get some results in the first 90 days if you spend your first 30 days making sure there's no product gaps, your second 30 days doing all the data analysis, maybe in your your the last month of your 90 days, you could start doing some targeted campaigns to to drive some more engagement. And so that would be my locked by 90 day plan, you know?

Andrew Michael
Yeah. So it's obviously it's not an easy question to answers or not having an understanding where the company is at and what sort of product you're dealing with. And like you said, like, turn itself is such a nuanced problem, you can't build a model foot because it's not as easy as as BAM model that applies to any company and every situations. Definitely multifaceted. So last year, we're running up on time. One last question for you. And, like, What is one thing that you wish more companies would look at when it comes to churn and retention?

Carl Gold
I don't know if there's, it's hard to say just one thing, but I mean, the overall theme is definitely to look at it as not, you know, reducing the churn but increasing the engagement and value. And so if you wanted to say one thing, I guess I'd go back to my you know, you know, unit cost metric because that's really going to separate it, you know, Who's finding the value and who's not for any product with a price, you know, that is so and you know, in general insurance fighting techniques apply to free products to you don't have to have a subscription. But if you are on a recurring payment plan, then definitely, I mean the unit cost metric as a way to really see you know, who's getting value.

Andrew Michael
Absolutely. And I think even on free products that you're still paying a price and you're paying it with your time, typically so

Andrew Michael
Cool. So very, very good Carl. I really, really appreciate having you on the show today. And once again, for the listeners call is going to be giving away five copies for for his book fighting turn with data. So you'll hear some of the techniques he discussed today and a lot more as well with that. So once again, make sure to share this episode on Twitter and we'll be sure to grab your name and added to the draw, which will be announced two weeks from today's date and five lucky people will get a copy of his book Call really, really appreciate having you on the show. Thanks so much for joining today. Maybe you want to just leave us with final thoughts and how the listeners can keep up to date with your work.

Carl Gold
Well, one thing I just want to point out, so no one, you know, feels like they've been misled is the book is not actually complete. There's currently four chapters in an E book. And if you if you do get one of the ebooks, you'll get the four chapters that are available now the fifth is coming out soon, and then you'll get around one chapter a month in an E book. And then the book will be the pub, the hard copies will come out. We're targeting June of next year. But the book is about 80% written at this point, but you know, publishing you know hard copies takes a lot longer than making the ebooks and also parting thoughts. I do have a blog website. It's just fight turn with data calm and that's where there's you know, There's like a video of a conference talk I gave and some blog posts and you know, different resources that I found. In fact, you'll be there will be a link to this podcast by the time you look at it. So fight. Sure and with data.com is the website to follow up.

Andrew Michael
Excellent. Yeah, and sorry, I it's good that you pointed out that I've always found it very interesting is all sorts many publications. And you're able to launch a book early, get some feedback from, like, sort of your beta readers to some extent. But again, like the the content itself, like Paul said, Is 80% there, and you will at the end, as well have access to that book. So thank you so much, again, for joining and thanks for sharing with the listeners wish you the best of luck now and the completion of the book and endeavors going forward.

Carl Gold
All right. Thanks, Andrew. Thanks.

‍