EP67: How to Fight Churn with Data Science in Your Subscription Business with Carl Gold
Carl Gold is the Chief Data Scientist at Zuora where he created the Subscription Economy Index™ and the data science behind Zuora’s Subscriber Insights. He has a Ph.D. from the California Institute of Technology and has authored publications on Machine Learning and Neuroscience. Before coming to Zuora, he spent most of his post-academic career as a quantitative analyst on Wall Street.
In this episode, Carl shares his insights about using data to increase subscriber engagement and reduce customer churn. He provides a clear overview of churn concepts, along with hands-on tricks and tips he has developed through years of experience analyzing subscriber behavior.
Why you have to check out today’s podcast:
- Discover how to use data science techniques to reduce churn in subscription services
- Learn how to create a churn dataset as benchmark metrics for subscription economy decisions
- Learn what is predictive of churn, and what can be done to reduce it
“The thing about churn is there is no one solution. If you think a customer is going to leave, the next question is ‘why’.”
– Carl Gold
Increase Your Pricing Knowledge: Become a Champions of Value INSIDER!
To sign up go to insider.championsofvalue.com.
02:18 — Backstory of Carl’s career, the stint he had at Wall Street finance industry being a quant and eventually becoming a data scientist
04:40 — The similarities of a quant and a data scientist, what they do. And the similarities of machine learning to data science
06:04 — All about subscription economy index and how companies use it
07:22 — The biggest challenge a subscription company faces when it comes to billing and where does Zuora come in to help
10:03 — Carl addressing the bias dealing with index and metrics
11:23 — Carl’s thoughts on subscription companies being recession-proof than traditional companies
15:50 — Book alert: “Fighting Churn with Data” by Carl Gold – what is it all about
23:19 — How to convert leads to customers by looking at a buyers’ journey and sequences of action leading to conversion or churn
29:20 — Carl’s best pricing advice to listeners that could have a big impact on their business
“If you can’t bill people correctly then you know, what’s your pricing plan good for?” – Carl Gold
“Some subscription companies are probably seeing boom times because if it’s a product that people can enjoy and consume more of when they’re confined to their home. Demand is supposed to be through the roof right now and they’re struggling to keep up.” – Carl Gold
“If you want to reduce churn in a meaningful way, you have to give people more value.” – Carl Gold
“One of the most valuable metrics that you can ever make for a subscription business is the recurring unit cost that the customer pays. That is one of the most important churns and retention indicators that will give you your amazing insight into the value that customers are getting.” – Carl Gold
Connect with Carl Gold:
Connect with Mark Stiving:
Full Interview Transcript
(Note: This transcript was created using Temi, an AI transcription service. Please forgive any transcription or grammatical errors. We clearly sound better in real life.)
Carl Gold: The thing about churn is there is no one solution. If you think a customer is going to leave, the next question is why. And there are many reasons a customer could be leaving.
Mark Stiving: Welcome to Impact Pricing, the podcast where we discuss pricing value and the continual relationship between them. I’m Mark Stiving. Our guest is Carl gold. Here are three things you want to know about Carl before we start. He’s a chief data scientist at Zuora. One of my favorite titles. He’s the creator of the Subscription Economy Index and he has a new book coming out in August titled Fighting Churn With Data. Welcome, Carl!
Carl Gold: Thanks.
Mark Stiving: Hey, I always start the question, how did you get into pricing? But it’s probably hard to say that you’re in pricing. So how did you get into data science?
Carl Gold: Yep, that’s fair. I’m not really in pricing except my work touches on it in several respects. So how did I get into data science? A little bit different than most people. Maybe it speaks to my generation. I did a Ph.D. in, well, it was in a department that combined neuroscience and AI and neural networks, and this was all long before neural networks and AI was cool, well maybe not that long before, but you know a good way before neural networks and AI was cool. And when I finished my Ph.D., I actually didn’t go directly into data science at the time. There was no such thing as data science and there weren’t that many jobs in machine learning. But there were a few. In the end, I was interviewed and I got offered a job doing one kind of machine learning and I also got a job at a finance company. And so I actually became a quant directly out of my Ph.D. And back then that was really the main route for a science Ph.D. who didn’t want to stay in academia. So I joined the Wall Street finance industry shortly before the financial crash of 2008 and stayed in the industry for many years, around seven years as really a Wall Street quant and not data science. But during that time, data science became a thing. AI became popular and because of my Ph.D. background, which included AI and machine learning, statistics, and all that kind of good stuff, it was pretty easy for me to do a lateral career transition and become, you know, a data scientist.
Mark Stiving: I have to say I’m pretty sad cause you to know when I was younger I was a geek, right? I love stats and econometrics and all that stuff, but I left it before data science became cool. Right. So now you’re like the cool kid. This is really neat.
Carl Gold: Yeah, it’s pretty good timing. I did kind of miss the first wave, though, cause if I had timed it better I could have been part of the first wave of AI and data scientists. But I was actually in the Wall Street world while that was going on. But you know, it was still a really great experience working on Wall Street so I wouldn’t change what I did.
Mark Stiving: And it’s always about trying to model something and figuring out how the world works with data.
Carl Gold: Yeah, no, I see being a quant and a data scientist as really pretty similar, I mean practically the same and that you’re just using science and math to do something, you know, that’s economically practical. Of course, as a quant, there’s a lot of specialization in certain areas of financial theory. And as a data scientist you’ve got more emphasis on machine learning, arguably. Although a lot of machine learning is, you know, finding uses on Wall Street now too. So yeah.
Mark Stiving: So what do you mean when you say machine learning and how that’s different from data science.
Carl Gold: Oh no, I don’t mean it’s different. I mean, machine learning is, of course, a big part of data science. It’s less a part of being a Wall Street quant was what I meant. Okay, so machine learning, to me it means, I guess different things to different people, but you know, it’s any predictive algorithm that learns from the data, except you’re kind of excluding classical statistics like regression and then you get into kind of these edge cases and you’re like, Oh, it’s Ridge regression, you know, statistics or machine learning. It’s all kinds of semantics.
Mark Stiving: Got it. Got it. Okay. Let’s talk about a couple of things that you’ve done recently cause I’m fascinated by both of them. One is the subscription economy index. We talk a ton about subscriptions and pricing subscriptions here on our podcast. So tell us what the index is and how companies use it.
Carl Gold: Sure. First I’d better back up and tell your listeners what Zuora is. Zuora is a software platform for running a subscription business. So if you’ve got a subscription business, you’re going to have pricing plans and products. Zuora lets you manage your pricing plans and products and of course your subscribers, your customers. It generates people’s invoices and talks to your financial system. So Zuora runs a lot of subscription businesses someday.
Mark Stiving: So can I think of it as an ERP for subscription companies?
Carl Gold: Kind of but not exactly. I mean Zuora has this model that they call like a three-cloud model where you have your CRM on the one hand and then you’ve also probably got an ERP and Zuora sees itself, I should say CRM is your customer relationship management. And then Zuora kind of sits in the middle is how they see it. So the ERP doesn’t really go away. But you’ve got because instead of selling products, you know one at a time through your channels, you have to maintain this relationship with your customer over time. Zuora maintains that relationship. Particularly focusing on the financials of that relationship.
Mark Stiving: Would it be fair to say probably the biggest challenge the companies that are going subscription face in this aspect we’re talking about is the billing and the fact that we now have frequent billing and we’re billing different aspects and big challenges the companies face.
Carl Gold: Yeah. It’s the challenge of having a, let’s just say if you have an interesting pricing plan, then it’s a challenge. If you’re billing everyone 999 a month, that’s not much of a challenge to get everyone’s bills right. But as soon as you go into like a good, better, best plan, maybe you’re giving temporary discounts. Suddenly things are very complicated and it’s really hard just to get everyone’s bills right. And then you can add pay as you go pricing, which we call usage pricing or usage-based pricing. And yeah, it’s hard to get everyone’s bills straight and very quickly. And Zuora lets any company define a complex pricing plan with like a good, better, best level discount add-ons and still get everyone’s bills. Right?
Mark Stiving: Yep. Yeah. So that seems like a really important thing to get right.
Carl Gold: Oh yeah, of course. I mean if you can’t bill people correctly then you know, what’s your pricing plan good for? So I want to bring it back to the subscription economy index cause that’s a question we started out with. So Zuora has data about a lot of amazing companies that are, you know, leaders in the subscription world right now. And the subscription economy index takes the data that’s generated on the Zuora platform and completely anonymizes it I should add. So we’re not looking directly at any company’s data. So the data is anonymized and aggregated to make benchmark metrics for the subscription economy. And so one of the most important metrics that we track is the average revenue growth rate of companies. And that’s what the subscription economy index itself is a name for. It’s an index that tracks the average sales growth of the companies on the Zuora platform.
Mark Stiving: Yeah. And so when I first saw that, you probably did the exact same thing I did. And that is every time I see a metric or an index, I say, okay, does that make sense? Are they cheating? Is it biased? Right? And so I start thinking through the subscription economy index. And the first thing that comes to mind, because you guys always publish, oh look, subscription companies are growing 20 times faster than traditional companies. I don’t know what’s the number? So I could use the right number.
Carl Gold: I have to check the latest. I think the long term average is around five times faster.
Mark Stiving: Good enough. Right? Good enough. And so we always publish that and I always have to step back and say, okay, now is that real? Does that make sense to me? And so in my mind, what I think is, it may be that the customers that you have data for are the ones that are well-run and growing well and the companies who don’t use Zuora probably don’t do as well or they’re suffering or struggling and so maybe they’re not growing as quickly. So that was my one capture of the possible bias there.
Carl Gold: Yeah, I think there could be a certain degree of survivor bias, although we keep failing companies, you know, in the index as they fail. So there’s no survivor bias that, you know, that I put into it. But it is true that I mean Zuora is a sort of a premium product for what it does. And if a company is not well-managed or not doing well, you know, one of the things they might do is cut from Zuora and use a cheaper billing platform. So there’s a fair line of reasoning there. Yeah.
Mark Stiving: So the next thing I want on that topic, I am so interested in seeing the results after the coronavirus pandemic ends because I will bet you the subscription companies survive this so much better than traditional companies do.
Carl Gold: That’s kind of what we’re expecting. I mean, I don’t want to gloat or be, you know, triumphal about bad news for a lot of people. Yeah. And I haven’t, you know, I run the data on calendar quarters so you know, Q1 just finished and I can, you know, I could start updating that data now. I haven’t done it yet. In case you’re wondering, I’m even wondering if we’ll see much impact in Q1 because in the Western economies, it only really hit at the end of Q1 and at least for a lot of B2B companies, I think, you know, January is when the deals were closing and yeah, consumer products. Yeah. There might be a spike in March, but, so it’ll take some time to see. But absolutely.
Mark Stiving: Yeah, it’ll be interesting. And that’s one of the big advantages of companies going subscription is I think it’s more recession-proof or downturn proof. Cause you know, it’s hard to call this a recession. All technically it is, it’s not caused by economic factors.
Carl Gold: Yeah, I mean, well, of course, there’s no such thing as recession-proof because you know, eventually, I mean I have some experience with this actually because the Wall Street company where I work was actually an analytics provider. And so they were actually, they didn’t call it a SaaS product, but it was essentially a SaaS product. And so we went through the last downturn and when it first hit, you know, of course, everyone was locked into multi-year contracts. And so yeah, it didn’t really didn’t move the needle in the beginning, but then when those contracts came up and the recession was still going on, that’s when you know. So subscription companies are probably definitely delayed in their response to a recession.
Mark Stiving: Yeah. I don’t know if it continues for a long time, it’ll certainly affect them, but I think that their downturn is going to be much slower or much less dramatic than traditional companies where a lot of times you could just cut and spit it off. I’m not buying that anymore.
Carl Gold: Yeah, absolutely. So then some subscription companies are probably seeing boom times because if it’s a product that people can enjoy and consume more of when they’re confined to their home, then you know, demand is supposed to be through the roof right now and they’re struggling to keep up. That’s what they’re saying. It’s Facebook and of course Zoom video that we’re recording this conversation on.
Mark Stiving: Yup. And Netflix, I would jump them, pile them in there too. They’re probably skyrocketing right now.
Carl Gold: Oh, and I’ve noticed some Netflix freezes in the last couple of weeks that you know you didn’t normally have. You have to be a little patient with Netflix now.
Mark Stiving: Yeah, so there’s a bandwidth problem in the whole industry right now. So
Time out for a quick ad.
Our goal at Impact Pricing is to help you become a champion of pricing and value.
This will undoubtedly increase your impact at your company. It’ll help you get promoted and make your company more profitable.
And now we have an unbelievable offer to push you even further down that path. We just released our new insider program where you get access to all of our current and future courses.
If you just take the courses we currently have released that’s over $2,000 in value. The good news is we release courses that are really faster than one every two months. That’s faster than anybody’s been able to take them. At least so far.
You get access to everything for the low price of just $100 per month. That’s a small portion of the raise that may come with your next promotion.
To make this offer even more irresistible. We offer you your first month for only $5. Try a few courses, see what you learned that’s valuable. We are confident you’re going to want to stick around.
Oh, and did I mention you get access to office hours? One hour a week I get on zoom and answer questions from our insiders. Considering I charge $1,000 for one-hour pricing consultation that makes this a no brainer.
Go to insider.championsofvalue.com. To sign up to take advantage of our special $5 offer. Use the promo code ‘INSIDENOW’.
This special offer will end at the end of April 2020. Join me and your colleagues as a Champion of Value Insider. I’ll see you on the inside
Mark Stiving: Let’s switch topics to probably the thing that you really want to talk about the most and that’s your new book that’s coming out. So you’ve got this book called Fighting Churn With Data that’s supposed to come out in August. I love the title and I have all these different ideas just based on the title, but tell me what the book is actually about before we jump into that.
Carl Gold: The book is about, well, the context is a subscription company mainly, although it could also be any company with recurring usage and you know, if you’re making your money from your users coming back, then you have the problem of churn. And for any listener who doesn’t know, churn means people canceling their subscription. The term originated with the churn rate, which came out of the telecommunications industry. I mean the idea being if you lose, say 1% of your customers every month, you have to replace them every month to keep growing and you want to more than replace them of course. So you have this turnover, they call it at the churn rate, and now churn is like a noun and a verb. Like we say, Oh that customer is a churn, you know, make a report of last quarters churns and you know, even, oh I churned from Netflix last month cause they finished, you know, the last season of blah blah. Anyway, Fighting Churn With Data is, assuming you’re a company with repeat users, it’s for actual programmers and data scientists on how to analyze churn data. Because typically a company, you know in the subscription economy or the online economy is tracking what users do on the product. And this isn’t like stalking people on the internet. This is actually, you know, really measuring the plumbing of your own product, right? You need to measure what users are doing, what if you’re going to make the product better. So the book teaches people how to use the data around churn in particular as like a survey of your users. Basically the user’s book with their feet, you look who stays and who goes. There’s a bit churns and the retained customers and the book teaches you how to analyze that for the insights that you can bring back to the business.
Mark Stiving: And so what’s unique about analyzing churn?
Carl Gold: Well, a few things in comparison to other data science problems. The biggest difference with most data science problems is actually that I don’t emphasize machine learning and AI prediction in this area. Because the thing about churn is there is no one solution if you think a customer is going to leave. The next question is why. And there are many reasons a customer could be leaving. It could be they’re not using it at all, or it could be they’re using the product in some negative or contrary way, or they had unmet expectations about it. Or it could be that they signed up for the premiere plan, but they’re not using it enough to justify. It could be that they didn’t, they don’t understand it, you know, they don’t know how to use it and they need training. So because there’s no one size fits all solution to churn, you actually have to understand the customers rather than just use an AI algorithm. So the emphasis of the book is actually on creating customer metrics that the business users who have to, you know, reduce churn can actually understand.
Mark Stiving: Yeah. So I could imagine. But first off, let me say that’s not what I thought the book was about. Believe it or not. I actually thought the book was going to tell me many, many examples of companies that have identified churn using data and how to solve that.
Carl Gold: There are case studies and there are a couple of case companies where Zuora customers that appear throughout the book showing you, yeah, this is what the real results look like. This is what a real company found.
Mark Stiving: Yeah. So I think that makes a lot of sense than to teach me how to do that and what I’m imagining this looking like is I’m going to look at December’s churn and so I had a 1% churn in December. That means I had 99% non- churn in December. So I’m trying to compare the 1% of people who did churn to the 99% who didn’t. And to do that, I have to go back to November and October and September and look at the behaviors of those customers over that timeframe.
Carl Gold: And if the way you’ve said it, I mean, well, I don’t know how technical are your usual listeners.
Mark Stiving: All right, let’s go for it. We’ll try to simplify it in a second. I want to hear the details.
Carl Gold: I mean what you just described is basically what is taught in the book is how would you make a data set from observations of past times when people stayed or left and the big element that you were alluding to is the calculation of the customer metrics because that’s what tells you what has this person been doing for the past month or the past three months or whatever it is.
Mark Stiving: And so are you teaching the customer metrics? These are the right metrics to be watching for.
Carl Gold: I teach a system for designing and exploring customer metrics and it’s really, if there’s a, you know the subtitle of the book could be like the science of customer metrics. It’s not quite going to be that they have subtitles, something along those lines, but because there is no one size fits all solution for the customer metrics, it depends on what data you’re collecting. It depends crucially on what are the value-creating activities of your product and how closely you can measure them. Because in some companies the value-creating activity is very easy to measure and in some companies, it’s intangible like your enjoyment of a streaming video is intangible and the company that’s hopefully trying to stop you from churning will have limited information to work from. They can look at what you watch, they can look at what percent you completed. If you actually liked or shared something, then that would be like a real treasure of information. But most people don’t even bother with that. Yup. So then there are other companies if you think of it like a CRM or a Zuora kind of a billing product, the user of the product is generating financial value and it’s actually very easy to measure the value that the customer generates using your product. So there’s a whole range of situations. But the amazing thing is actually the techniques in my book apply across the board. So it’s kind of like a framework for thinking about metrics and designing them.
Mark Stiving: Yeah. So let me flip this upside down for just a second so I could use another set of words and then I want you to give me those words for this circumstance. And so a lot of data scientists today are focused on how I convert leads to customers. And so we do that by looking at a buyer’s journey. And we model, we assume that buyers are going to go through this step first. I have to be, you know, there are many different models, but I have to know that it exists. I have to like it, I have to evaluate it, I have to. And so the data scientists or whoever it is that’s working with them is creating a behavioral model of how they expect buyers to behave. Now let’s flip that upside down and say you have to be doing almost the exact same thing, but it’s how I expect users to stay. Right? What’re the words that I’m looking at?
Carl Gold: How do you expect users to generate value or to experience value from using the product and you’re right, it’s the exact opposite. There is some difference though. And I’ll tell you, I mean the first one that comes to mind and I often you know, discuss this with people is that from the buyer’s journey, people are more thinking about sequences of actions and they’re concerned, you know, about what order they might go through a website or you know, read certain content, you know, once the person signed up, cause this is a recurring usage situation. I tell people to just focus on aggregate statistics. Like how much are they doing things and not so much the order? People often ask me things like, Oh, what? What do you think about the word, the final moments? Or you know, I don’t usually think about the final moments, honestly, because this is in a way, just based on my own subjective experience. When you get value from something, it’s a process that occurs over time. And if you’re not getting value from a product, it’s usually also a process that occurs over time. It’s not like one bad experience will make you cancel, but for sure a sequence of sub-optimal experiences without enough really good experiences over time that will, that is what will make you cancel. And so usually in the churn scenario, the straw that breaks the camel’s back is it’s just another piece of straw. You know, you have to look at the whole journey, the whole aggregate of what they’ve been doing to see where they got value or failed to get value. Yeah.
Mark Stiving: So it’s hard to say like a buyer’s journey, that it’s step by step. It’s more of a gradual, we’re now not getting the value we used to get. So, okay, now I love this, but here’s what I want you to do next. And I wouldn’t surprise you if you do this. Once I can predict how much value somebody’s getting from my product based on their usage patterns and how they’re, I guess we’ll just say usage patterns. Now I should be able to say to my sales team, these are the people you want to call on because they’re about to upgrade or they’re ripe to upgrade to the next level or they need to get more usage going. And if we guided them we could get them. So we throw it into the customer success department. Are we doing that as well?
Carl Gold: Yup. That’s pretty much the idea. The book doesn’t go into the details of how to run a customer success team or also, you know there’s going to be a marketing component too. You might start with what they would call an engagement campaign where you’re marketing to your existing customers with useful tips and information. I’m sure you’ve gotten some of those. So the book basically gives you all the data those people need to, you know, do their thing. But I don’t go into how you design an email campaign or you know, run a customer success desk. Cause I speak from my own experience. I’ve never run a customer success department. I’ve talked to many heads of customer success departments, but I’ve never been one. Yeah. Okay.
Mark Stiving: So I’m going to rename your book by the way. You don’t have to rename it, but I want to rename your book to determine how much value users are getting from our product. That is really the essence of it. I think Fighting Churn With Data is a little catch here. So probably stick with that. But it really is the essence of the book. I mean, cause I emphasize very early on that there’s no gimmicks or tricks with churn because the people who are already using your product, they know what it’s like. So if you want to reduce churn in a meaningful way, you have to give people more value.
Mark Stiving: Yup. Yeah. I love that. And the reason I wanted to rewrite your title, at least for me and my audience, is because value has everything to do with pricing. So that fits us perfectly. Okay. We are so far over time, but you promised me that you had a contrarian view on data science that I absolutely have to hear before we shut this down. So let’s hear it.
Carl Gold: Well, I think I already threw it out there, at least with churn and there I say, you know, save the AI for later and just focus on actually using your own brain and understanding your customers. So that’s kind of contrarian and it definitely makes it harder to sell a book like this because most of the conferences are looking for, oh, what new buzzwords in technology or are you promoting. And I’m like, well, little sequel, Python, maybe a regression.
Mark Stiving: Oh man. So I just did, I do these little videos and I just did this little video the other day on how I get this feeling that companies are so focused on data now that they forget to go out and actually talk to their marketplace. Actually, I listened to people. And I think you’re kind of in the middle. So if I were to go AI and machine learning, it’s like I don’t even want to think about how people think. Now you’re saying I want to think about how people think, let’s use the data to figure it out. And I think we also need to make sure that we’re out talking to people who just churned out and figuring out why the heck they churned out. And are we representing that in the models we’re trying to test for in the churn models?
Carl Gold: Yeah, absolutely. I wouldn’t at all say you shouldn’t talk to your customers. I mean I do say at some point in the book that you know this churn data, it’s like a survey that you don’t have to ask people, you know, and that everyone answers every time they renew or churn everyone answers this survey because it just takes a little more technical expertise to read the results out cause you have to, well it’s kind of all in the book, you know, the design and the metrics and everything.
Mark Stiving: Awesome. Carl, I have enjoyed this so much. Thank you. But I have to end. I’m going to ask you the trick question, especially since you’re not in pricing. Do you have one piece of pricing advice you’d give our listeners that you think could have a big impact on their business?
Carl Gold: Sure. I mean I have probably had a few, but I’ll have one that I would give is really straight out of the book. One of the most valuable metrics that you can ever make for a subscription business is the recurring unit costs that the customer pays. So if you take just the price that they pay per month or year or what have you and divided by whatever important activities say, it’s like videos watched, you know, so it’s like your dollars per video or your dollars per whatever it is. That metric for your service is one of the most important churns and retention indicators and we’ll give you amazing insight into the value that customers are getting.
Mark Stiving: Nice. So you could almost start that as a dollar per value and whatever that value is or however that customer measures the value they get from our product.
Carl Gold: Yeah, exactly. Only the act might not be the direct value, but you want the act to be as close as you can get.
Mark Stiving: Yeah. Nice. Carl, thank you so much for your time today. If anybody wants to contact you, how can they do that?
Carl Gold: My website, which is fightchurnwithdata.com and that has information about the book. You know my social media contact also @Carl24K on Twitter.
Mark Stiving: All right. Thank you so much. Episode 67 all finished. Let’s say my favorite part of my, how to say everything was my favorite part of today’s podcast. I absolutely love it. I absolutely loved it. It was so much fun geeking out and thinking about the models, so what was your favorite part? Please let us know in the comments or wherever you downloaded and listened while you’re at it, would you please give us a five-star review? They’re very valuable to us. Don’t forget we have that free community at community.championsofvalue.com where you’ll see all of the things that we publish and have great conversations there. If you have questions or comments about the podcast or about pricing in general, feel free to email me, firstname.lastname@example.org. Now, go make an impact!
Mark is a pricing expert who helps companies understand value, how to create it, communicate it and capture it. He has a PhD from U.C. Berkeley and an MBA from Santa Clara University, plus 25+ years pricing experience. As an educator, speaker and coach, Mark applies innovative, value-based pricing strategies to guide growth and increase profits for large and small companies.