Predicting vs preventing churn

Steve Hazelton


Founder & CEO


Steve Hazelton
Steve Hazelton

Episode Summary

Today on the show we have Steve Hazelton, Founder and CEO of Sturdy.

In this episode, Steve shares his experience with the varying levels of acceptable churn as your company grows and how to decrease it over time.

We then discussed how pricing can be leveraged to increase retention and how you can reduce onboarding fall-off rates in enterprise deals by charging your customers upfront and we wrapped up by discussing the pros and cons of predicting vs preventing churn.

As usual, I'm excited to hear what you think of this episode, and if you have any feedback, I would love to hear from you. You can email me directly on Don't forget to follow us on Twitter.

Mentioned Resources




Acceptable levels of churn as your company grows00:00:00
Leveraging pricing to increase retention00:00:00
Reducing onboarding fall-off rates by charging customers upfront00:00:00
Predicting vs preventing churn00:00:00


[0:29] Andrew Michael

Hey Steve, welcome to the show.

0:36] Steve Hazelton

Hi, Andrew. How are you?

[0:37] Andrew Michael

I'm great, thanks fasting and, uh, thanks for joining. For the listeners, Steve is the co-founder and c e o of Sturdy helping businesses improve their products, processes, relationships, and revenue by leveraging customer conversations. Prior to founding Sturdy, Steve uh, also founded Gravity Technologies, which he bootstrapped to become one of the Fosters growing privately held companies in the Bay Area. He then went on to find Newton Software, which was later acquired by Paycor, where he went on to serve as Pay Core's general Manager. So my first question for you, Steve, is after finding two companies previously with a successful xFi, what was the motivation to get started again with study

[1:17] Steve Hazelton

<laugh>? Yeah, that's a, uh, that's a loaded question. Um, I could say I'm, I'm very bad at retirement, but I think I've was, I've just really been interested over my career in how to improve, how b2b, uh, well, how SAS business is and ultimately B2B SaaS business operate their businesses more effectively. Um, I don't wanna get ahead of myself here, but to me, there's a tremendous, uh, data store that's inside of all of our businesses that we, that our customers want us to use, uh, they desperately want us to use, and that's their feedback. That could be their emails where they say, Hey, can I get a copy of my contract? Um, or it could be a bug report and a ticketing system. It could even be a transcribed phone call. That unstructured, uh, content in almost every company is not data. Um, and it, if, if it is data, it's a human being manually recording that to Salesforce, quote unquote, if it's important. And so I saw that as an opportunity to improve how businesses operate and to be honest, to solve a problem that I witnessed, um, when I was GM of a large company. And we can go into that more, if you'd like to hear about that. I think that's really fascinating to me, is how, is is kind of how churn and revenue retention are different and how they change from a small to a medium to a large company.

[2:52] Andrew Michael

Yeah, I, I think that is definitely something as well, like my perspective on China retention has changed over time and the importance of it at different stages, uh, of a company's growth. Uh, I you hear your perspective as all, like, uh, maybe elaborate on that a little bit more.

[3:08] Steve Hazelton

Yeah. Um, I mean, I used to say that, so our background or my background, I started a, um, an applicant tracking company. So if anybody's ever gone and applied for a job, it's likely that you've applied for a job through, uh, a product called Newton Software. Um, when we started, and we always had very, very, uh, high revenue retention, in fact, positive revenue retention, we, we would get a customer and we'd be somewhere around 117%, which pat ourselves on the back. We had that number before. That number was cool, I guess, or anybody thought about it. We were really one of the first cloud, uh, enterprise SaaS companies really, very, very early on. And we were kind of in the right place at the right time. We started at a hiring software company during a recession, and then when the recession ended, the financial crisis ended.

[3:59] Steve Hazelton

We were one of the few people selling hiring software when everybody was hiring. So I'm not gonna say that I knew what I was doing, but I was definitely in the right place at the right time. Um, I think when you're small, and we, I don't wanna draw an arbitrary line, but it's depending on your business and what you do, 500 customers, a hundred customers, a thousand customers, you have almost a telepathic understanding of what your customers want in your product. Um, what customers are unhappy or happy, which of your customer success reps are, are, are, are competent or doing a good job, which of your sales reps are giving you bad deals? And there's things that, um, you can rapidly change in your business to address those issues and look like you don't become a large company if you have high churn when you're a small company.

[4:43] Steve Hazelton

That's, that's, that's obvious, right? The challenge that I realized is as you grow capturing, um, you know, we used to say, are we growing with our customers or they growing away from us, right? And as you grow and sometimes look like your business will change, and then there's some discussions I'm sure you've had about how when you move from a small company to a medium or a large one, your churn may go up because you're changing your customers. But we were not, we were a small, medium sized business. So I, and I, I'd say churn for me in losing customers was something I hated more than anything in the entire world. I was, I was, uh, obsessed with it. So, um, when we got b we, we were quite successful. We ended up getting bought by the fourth largest payroll provider in the United States. We went from having 2000 customers to having 23,000 customers or something like that.

[5:40] Steve Hazelton

And the, the churn rates, um, changed dramatically. Like when I, they made me a GM of a business unit. One of my, and I'm sure you've been there too, Andrew, one of my directives was, okay, we gotta get this turned down, right? And I realized when I started that group, I had no data at all that I could rely on to inform what I was doing. Right. And I think a lot of folks who are listening, if you get a new job as VP of customer success and they put you in charge of churn, you likely don't have any data at all at your fingertips to inform you, like, why is this happening? Right? Um, and so why is that? Well, when we're a small company, it's like founders and VPs are doing most of the interactions with our customers. They know how to competently get back to them.

[6:28] Steve Hazelton

They know how to say, this bug is really important, we need to fix it right away. As you scale, the people who are interacting with your customers arguably, um, are not VPs anymore. Well, that's not arguable. There's no way a tw if, uh, uh, a company doing 300 million a year in revenue has VPs talking to every single customer, right? And so the triage of those issues, the collection of that data and the ability, uh, the ability to filter that data up through your organization, Eva, it just evaporates. And so then you go back to that challenge, like, you know, and I'm getting ahead of myself again, but what I think when we think about churn, there are a number of ways to tackle a churn problem. It's engagement, it's lead source. Um, but if you don't have, if your product, if you are not building the features that your customer base wants, if your reps are making your people mad, um, or not providing quality service that is eventually going to lead to churn, you may not notice it immediately. They may be on a year or a two year contract, right? So I'm really interested in kind of solving that problem of what is the data that I can use to inform, um, these decisions that I make? And, and also, how can I use this data? What can I, what is inside of someone's email? What is inside of these support tickets that I can use to improve my business short-term, uh, uh, mid-term and long-term?

[7:57] Andrew Michael

Yeah, it's interesting the, the way you use the word data, uh, in this context as well. So I think a lot of times when people say like, uh, when they think of data, it's numbers and analytics, uh, product itself. And like, this is the first thing. And I think when you said like, you join a new company and typically you don't have a lot of data, I think normally what happens when you join that company, there's a lot of data in the traditional sense. So there's data and analytics and it's telling you what's happening. And you can see in the numbers that churn's not looking good, retention are good, but where the real value lies is the data. That's the unstructured data. It's the, the things you're alluding to, it's the customer feedback, it's the conversations and, and that gives you the why things are are happening.

[8:35] Andrew Michael

And definitely like at the early stage, like you say, like, uh, I see it as well, uh, in my business and yourself, like we are speaking to customers all the time. We're gathering our feedback, we're iterating fast, we're making changes. And you see that immediate change. But as that scales, that as the company scales, that doesn't scale, uh, more often than not. So I definitely see that is a big change. One of my perceptions that also changed a little bit on churn retention, and I like the quote that you meant, you just mentioned like, you don't get to become a big company, uh, if churn is bad when you're a small company. And, uh, that definitely holds true. But I think at the early stages, while there's always going to be a level of churn that I would say like falls into an acceptable level bucket that might not be at that net negative churn, which is like the holy grill everybody wants to get to in, in a sales business.

[9:19] Andrew Michael

But really when you, at early stage, you still really haven't figured out your product market fit. You haven't really got a good niche, so you are gonna be attracting some of the wrong types of customers. Um, but that money's really gonna help you fuel the growth to find those right customers. And I'm interested, like, is this something that you've seen in your career as well as that, like obviously in the, in the early days trend might not look as good as you'd like it to look, but that's a, like a necessary evil, uh, to be able to grow in the early days.

[9:46] Steve Hazelton

Yeah, I, I, I think that's a, one of the toughest challenges of starting a new business. Um, you may have an idea of what your, uh, I c p your ideal customer profile is, and then you go and get a few customers and then for reasons that you didn't anticipate, maybe they don't, they don't have the money that you need in order to grow business, or you find that there's another direction your company can go that might be more profitable. Um, there might be more demand in that direction and you shift, uh, and you're gonna lose those, those customers that you had. Um, I think that some company, I mean, I've seen it kind of both ways. I saw it at Newton, um, where, which was the hiring software company. The, the thing that we got right early, and I, this probably isn't the best place to discuss this here, like the thing that we got right early was the price.

[10:45] Steve Hazelton

Um, and I think sometimes we lose, like how we priced it was a, was a competitive advantage. This is an example. Our competi, when you, when you buy a hiring software, you have recruiters who use your software, you have jobs that people apply to, um, and then you have people who go in and, and interview candidates and hire them, right? And I, I promise I'll get to my point here in a second, but our competitors at the time were charging per job, or they were charging, um, per, you know, user and, and customers hated that. They're like, okay, I'm hiring more people, so now I gotta pay more for a job. So they would, they were disincented to use your software, right? They were like, okay, I'm not gonna post as many jobs. Well, if you're selling hiring software, you want as many jobs in your system as possible, right?

[11:32] Steve Hazelton

Like, like that. You're like, yeah, of course. Well, that's not what the other thing that people hated was like, look, this hiring manager only hires once a year. Why do I have to pay for her to log into the system for 11 months? Right? So we just said, you're only gonna pay for, for recruiters. So, but we didn't come to that pricing understanding early on. We had a messed up pricing structure as well. So we had some churn initially when we flipped our price. But, um, the point of that is there's a lot of things that in this, in these businesses, like we look at that, um, like are these people logging in or not? But there's also just these things like, is our pricing sensible? And is that, is that leading to retention? It made it very, very difficult by the way, for, um, for our competitors to come and take our customers too, because we had the sensible pricing that, uh, other folks, uh, couldn't really compete with. They hadn't geared their businesses that way. So it's something to think about, I think, when you're thinking about churn.

[12:28] Andrew Michael

Yeah, definitely. It's very, very interesting. And it's something, I actually had a conversation with a few different people in the past. Like one was actually <inaudible> we've had on the show previously, he was also VP of marketing at Typeform. Uh, and at the early days, like when you start thinking about pricing and packaging, it's like really trying to weigh in like what is the main core value driver that the user is extracting from your product? And that's what you want to get them to get the most out of. And then the second thing is, okay, like how can you then layer on different features and things that add value to that service that you can end up charging for and stuff? And like sometimes you might think, oh, like people want us to do X, like let's charge for X. But actually when you actually provide that value and then layer on a pricing model that works in addition to it, you end up getting engaged users who love the product, continue using it, and then stick around for the additional features.

[13:16] Andrew Michael

So whether it's like, um, the X number of jobs, uh, and you pay like over a certain number of opening, I'm not, I'm not sure what you went to in the end of pricing, but I felt that champagne with the at ts as well using, uh, at a, now I said like, we don't have roles really open at the moment, but yet we still need to pay for the service to keep it open. And, uh, it just didn't make sense for us. Uh, so definitely like pricing I think is a huge factor in figuring out like the early stage.

[13:44] Steve Hazelton

I used to say that price was the number one, the most important, um, aspect of your business. Like the way that it's price, I'm probably wrong. Uh, there's nothing that's the most important to your business. It's execution. It's, um, it's product. But if you have, if you have pricing that doesn't make sense, then it doesn't work. And what, you know, look, one of the things that, um, for, you know, business people and executives here, when we first started Newton, we, we went around to venture capital firms and they said, Hey, if you do freemium, um, we'll fund you, but we're only funded freemium stuff. So it's like, you know, first it had to be viral, then it had to be freemium, and now it has to be, uh, peeled product led. Product led.

[14:24] Andrew Michael

Yeah. Yeah.

[14:24] Steve Hazelton

They're all the same thing, really. Um, exactly.

[14:27] Andrew Michael


[14:28] Steve Hazelton

So we bootstrapped the company, we built it all on our own because we just realized our product couldn't, wouldn't work as freemium because anytime you're gonna plug into somebody's website, it's gonna have to get involved. And that freemium and getting it involved don't, don't fly, right? Like if I'm gonna have, um, so that was, that was pretty interesting where we started with this kind of freemium model. We realized that everybody that we were talking to, once we got plugged into their website, they were going to convert 100% of the time. So why not just start charging them Another really interesting thing, a little off task. And, um, one of the things we realized that really, really helped with our retention and, and, and I think there's, there's, there's this kind of, uh, missed or, or kind of dirty little secret of, of enterprise software specifically and B2B software.

[15:20] Steve Hazelton

It's like you have your current customer base, but you'll also have your onboarding customers, right? And a lot of times they don't get counted in churn if they fall out during that process. And I've seen businesses in the past that have actually really quite high no start rates, right? Or falloff rates or whatever you will call it. And um, the, one of the things that we realized, if you just want a simple tactical, and you may not, this may not work for your business or not, but you can test it, is um, we at Newton had reasonably high, um, fall off rates because customers would sign up for a product and then they'd be like, we're not ready to implement for the next 120 days or the next 180 days. And they would stay in our implementation cycle forever. I bet there's some folks listening that are like, oh my gosh, that happens to me all the time.

[16:12] Steve Hazelton

<laugh>, right? So we started to say, Hey, once sales closes the deal, the customer is paying their subscription fee. So once that customer signs, they have to start paying their fee. Now that sounds crazy, right? You're like, how can you, how could somebody possibly, how could you start charging somebody before you've given them the service? Um, what it did is it made sure that when sales closed a deal, they didn't say, oh, just sign the deal now and when you are ready will implement you. And it dramatically improved cash flow. Um, it dramat it completely eliminated no starts, they went from probably five to 6% to I think down to like one out of a thousand, right? It just completely went away cuz they knew they were ready to go. And then we just had an exception in our business where we said, Hey, if a customer really doesn't wanna pay the first month or the first two months while they're rolling it out, they don't have to. But pushing that, again, thinking how we can push that back up into, into the sales team, I I thought that was really interesting. Maybe that can help your listeners as well. Yeah.

[17:20] Andrew Michael

It, it makes sense as well. And I think the other aspect that it adds is that level of urgency. Um, because you put a little bit of pressure on the end customer as well. Like now, no, I'm paying for the service. I better go ahead and get it rolled out and, uh, make things start happening. Uh, it reminds me a little bit of a, like a previous episode with Janna Bato where we were discussing, like in their case, what they had previously was they had a free trial and there was like, first the trial was 30 days I think. And then they said, well, what if we cut that in half and they made it 15 days and, uh, in like doubled conversions, uh, because of like, that increased like, ah, now only 15 days to try this out. Then they haled it again and conversions went up again.

[18:03] Andrew Michael

Uh, and then what they slowly started to do afterwards was like, okay, now let me use like the days now to incentivize them to take action. So it was like, if you add your credit card, you get an extra seven days. If you do X action, you get another three days on the trial. Like, and then they use that to like sort of gamify the onboarding experience, which ended up being like a super, super successful. But I think what I got that takeaway was really like, that created this urgency for both sides. Like, okay, now I need to actually get this set up and take advantage of it.

[18:31] Steve Hazelton

I think that what that example you gave is, is really brilliant, right? Because it's also psychological, like if I have a month to figure out how to use a product, then I'll probably never figure out how to use it. If I have seven days, then I'm gonna sign up and I'm gonna be like, well, I better log in tomorrow and start putting some data in here, start getting some processes in place. So that sense of urgency, I'm incentive start working on it sooner instead of procrastinating it. And then at day 21 where I haven't done anything with a product, I go, well, you know, I never really used it, which is the death of all conversion, right? Well, I never really got any value out of it forcing somebody to get on the product sooner rather than later. That kind of goes to our point, we realized you can't get any value out of, uh, our product until you get it plugged into your website. So that that's, there's just no point in even giving it away free until it's, it's plugged up to the website. Cause they can't, there's nothing you can use.

[19:24] Andrew Michael

Yeah, it's weird that you say like the IT teams, cause I think Hot Jaw was similar in the sense that for a large percentage of the customers they required, um, to have somebody like technical to be able to put the code on the system. But I think what they were lucky, I think in the early days is they spent a lot of times we spent a lot of time in terms of onboarding and finding ways to make that easier and that that set up work. So I think you can make it work, but I also see in your scenario where it's, it's more like the product was more for technical folks, so you were like likely that the technical person knew what they were doing with. I think in HR you don't have that same, um, like technical expertise. They can just jump on the website and add something.

[20:05] Andrew Michael

Whereas like in marketing, people generally spend a lot of time on the website and so forth. But I wanna put you on the spot as well a little bit because before, uh, we started chatting at the beginning of the show and, uh, you brought up something and you said you might not mention this on the air, but like, I, I think it's worth talking about. And, uh, we're talking a little bit briefly about the concept of, um, predicting churn versus preventing churn. And I would like you to just like share your thoughts, uh, on this as one and, uh, we can dive into some of the points deeper. Uh

[20:37] Steve Hazelton

Oh. Okay. So, um, where do I start here? So I, I think, um, a good ex a a good anecdote, well, to explain what, uh, uh, to give some context to why we're talking about this. I guess what's, what's sturdy, uh, kind of core value proposition is, is that we take that feedback and we turn that we find the themes and topics in that. So in your emails, which are a black hole in your, in your, um, in your tickets, whatever, we already talked about that a little bit. So as we were going out and getting prospects and, and talking to our first potential customers, one of them, one of the people I talked to said, I want AI to tell me <laugh>, I want AI to tell me when one of my customers goes and does a, uh, a demo with one of our competitors.

[21:26] Steve Hazelton

Um, it's tempting to tell somebody that that can happen. That's magic, right? Like that's, it's surveillance, but it's also like we have to understand that, you know, our company actually used to be called sturdy ai and now we're We took the AI out of it because we have to be honest with what AI can actually do, uh, for our businesses and what it can't do, it can't do magic. Um, I'm sorry, like it, you, we could argue about whether magic actually exists or not, but, um, AI's good to weigh, oversimplify it and don't flame me AI people who are on this, but AI generally is good at things that people are not good at, and people are good at things that AI's not good at, right? So AI people are really good at identifying like one or two things and saying this thing is like that.

[22:11] Steve Hazelton

I mean, AI's really good at taking a lot of stuff and finding patterns and themes in it. So one of the things our, what we would call language models that we've built that use machine learning and, and, and unsupervised clustering and artificial intelligences, it's good at identifying, like if someone says, I want a copy of my contract, or I want, when is my renewal date? Or I wanna cancel, or This is a bugger features, very, very good at identifying those themes and topics like it's been trained on millions and millions of pieces a day. Um, so we can help prevent churn. We can say this customer and I, I was listening to one of your previous episodes and I love what this, uh, he, he, um, g was his name, right? And he said, executive change is like the strongest, um, indicator of, of churn.

[23:03] Steve Hazelton

There is. And it is like, we know this in our data as well. Like if you, if if your buyer or one of the executives at, you know, let's say you're selling finance software and someone says, Hey, our new CFO O needs a login, your chance of that customer churning is in, in the next 180 days is like over 50%. Very, very high. Now, if you have 'em on a year contract, then their chance of him churn a year is over 50. But work with me on that, right? It's very, very high. So we can help you prevent churn or an AI system that is detecting this. There might be others out there, or if, if you're manually harvesting this data, like, hey, success team, anytime someone requests a copy of a contract, I wanna know. Or anytime you, you hear about executive change, however you get this data.

[23:48] Steve Hazelton

But if you're acting on that data, um, you can prevent churn. You could go and give that executive a demo, you could turn them into a buyer, um, and then an advocate eventually, right? But if you just say, yeah, here's a login, you're dead, right? We can well, you're, you're gonna lose that customer. I think the prediction part is where we get a little bit over our skis in this industry and, and I'm talking about the customer intelligence industry, like one, and and again, this is controver. I mean this is, this is a little bit controversial as my take, um, but we can use machine learning technologies to identify themes and topics in your customer feedback, right? But your business is, and, and that's a common theme, but how you run your business and, and why your customers churn, um, are not probably similar to the cu the customer sitting next to you.

[24:45] Steve Hazelton

The, the way customers communicate with you would be similar, but the way you interact with your customers, what causes their churn is different. And so when someone says, we're gonna predict your churn, I think that, and this is what you were trying to get me to, to say, Andrew, and I'm really dancing around it. I, oh, I, I think that we have to be honest that that's not today really what you should be buying and what people should be selling you. Um, anything. And here's, unless you're huge and here's why. Um, I could take, we could take throw number out there, 10,000 churned customers and put them into a model and take all of your data from your entire business and run it through a model. And we would come out with something that would say, um, this customer is now more likely to churn than this customer.

[25:36] Steve Hazelton

We, that would be predictive. The problem is, if you're in a B2B SaaS business, which is our primary and B2B SaaS and B2B services are our primary customer base, if you have 10,000 churns <laugh>, you've, you're already in trouble, right? So that's my point. We can help. Yeah. I don't want you to get to 10,000 churns so that our model works. Let me put it that way. I would rather you not get to 10,000 churns. I'd rather you be able to adapt to your customers more effectively before you get there. That would be my point. Um, I think if you want me to stop there, I think you can get to the other controversial thing that you're, you're probably gonna put me on the spot on is like, and, and please, customer success, success folks, sit down for this. Um, you're not gonna like this, I think, but, um, I don't, I don't like NPS score.

[26:23] Steve Hazelton

Um, I think it's a, I think it's somewhat of a data point that can inform how we run our businesses as a C E O. And I would say my background is a product C e o. I'm a, I designed our first 50 screens of software. I designed sturdy software and now we have a designer. But, um, um, I came from politics way back when and I've worked in political polling and straw polling, which is what an NPS score is fundamentally flawed. Full stop. Yeah. It just doesn't, you get people who are really passionate about saying they hate you, are really passionate about saying they love you, but if you wanna make sure you're growing with your customers and not away from them, those would be the worst people to listen to. So my, my dream someday, I guess really if I want to talk about what I see as the dream for revenue retention and how I think what we're doing can help that.

[27:19] Steve Hazelton

And I hope there are other companies that, that come along and have the same vision and we can compete against them. I, I don't mind competitors. Um, I wanna go into a board meeting and I want someone to say, and I know I, we can go through some numbers here if you're interested, Andrea, cuz there's some interesting data that I have from these conversations. But I want someone to say 5% of our customers in the last 90 days expressed unhappiness, right? That's, that's a hundred thousand dollars in ar. And my goal as a CS leader is the next board meeting we have, it's gonna be 4%, it's gonna be 80,000, right? Bang, that's a number, right? Like that is data that you can say like, uh, in a meeting you can say, this is my goal. I'm gonna, and how am I going to act on that, right?

[28:03] Steve Hazelton

We get this a lot with what we're doing is like people will say, I love this data that you're capturing, but what do I do with it? So here's an example. I'm gonna take unhappiness down from, uh, we, you know, we have 5% of our customer base is unhappy, has expressed unhappiness in 90 days. Now that may not be like, look, look, look, we know that that's not always the end of the world. They might be, but the first thing we're gonna do is we're gonna identify which bug or feature is generating the most hap uh, unhappiness, right? That's our product does that, by the way, using AI <laugh>, there's a plug, but we're gonna identify that. And once we identify that we're gonna work with the engineering team to get that fixed engineering team's gonna go, wow, you actually have data about this stuff now, right?

[28:51] Steve Hazelton

Instead of, hey, this one customer's really annoyed, right? Uh, the second thing we're gonna do is we're gonna go and find our reps and we have customers that are doing this today. It's really powerful. We're gonna, we're gonna go and talk to our reps and say, Hey, you are, you are generating more unhappiness than us like you seem to be, and how can we help you? Uh, what are the issues around that? And it may not be their fault, it may be you have 20 customers who are just grumpy, but how are we gonna, how are we gonna do that? And I, I think that's just one example of what this data can do. But I think it's, I think to me, I've sat in board meetings before and they kind of go like this, right? The first hour is sales, the second hour is uh, um, engineering, and then there's like 15 minutes about churn and retention, right?

[29:43] Steve Hazelton

They're like, well, what are we gonna do here? And you're, and you're, and there's, you're, I could be, this is my experience, it might be different, but there's very little like, here is our plan and here is the data that backs up the execution of that strategy. And I think that's what we can help. And it's not, you're not gonna just use us. There's also gonna be data coming from like a Pando that's gonna inform if people are logging in. There's gonna be engagement data. But taking this all together and creating a, a, a holistic, um, few of your customer to drive that I'm really excited about.

[30:15] Andrew Michael

Yeah. So there's quite a lot to unpack and there's not much time left. The one thing I wanted to double click on a little bit was, and I you say they're controversial, some of the points, but I, I tend to agree with pretty much everything you said, um, as well in the terms of like, when it comes to using predictive models to predict churn, I think it's very, very difficult. Cause as you say, like if you get to that stage where you have enough data to be able to get to the level you can accurately predict, uh, churn, it means you have a very, very big churn problem already. So you're most likely never gonna get to that point where you have, uh, the data that you need to be able to, uh, to train a system. And I think when it comes to prediction as well, like the worst place to start really is focusing on churn itself.

[30:58] Andrew Michael

Like I think a much better way is to focus on what success looks like. And this is always the same thing when it comes to churn retention. I think like the first place people go to is like, let's do an exit survey and let's see why people are churning. But really like the, the focus should be on like, what are the successful customers doing? What are they doing that others are not? And how can we get more people to be doing those actions? So I think like the models that tend to work are the ones that actually focus more on the success metrics rather than, uh, what are the, the metrics that aren't happening or, or aren't there. So there there is, there's some validity in making that, uh, side of work, but it's not focusing on like predicting churn itself. It's more focused on like predicting success, uh, and what's unsuccessful. I'll stop. That's great. I,

[31:41] Steve Hazelton

Uh, I'm gonna steal that. Um, I, I, but I think that's a really good point. Um, we oftentimes focused on reducing the nega, the negatives, but are we looking at the positives? What's driving the success? That's great. Um, which is another kind of anecdote. I mean, I think by and large we're pretty most larger companies, success and support organizations do a pretty poor job of capturing the success side. So we have all of these processes about, you know, if there's a bug, go to Jira. If there's a, if there's an outage, go here. If there's a security thing, go here. What do we have if a customer says, I really love that feature.

[32:28] Andrew Michael

Yeah, but

[32:29] Steve Hazelton

What process do we have to cap? Like, look, look marketing, anybody in marketing is like, yeah, that, I'd really love to know that every, every single time that happens, right? Um, but what's driving that? Because that, that's a referenceable customer. What, what, what led to that event? And how do we capture that? How do we look into that? That's,

[32:46] Andrew Michael

Yeah, and I think that's like a problem as well that you mentioned with NPSs because I think nps you only really capture the extremes. So people are only gonna really respond to an NPS server either if they're extremely happy or if they're extremely pissed off and they really don't like your service. Like most users just don't have the time to really sit and, uh, give the time of day. And then second of all is like typically probably like your best users are gonna be the users are the busiest. They really don't have the time to be responding to surveys and to the thing. And I really like what you said though, as well in the sense that these conversations still happen. Like these messages are still being passed, but they're not being passed through NPSs and we just don't really measure that effectively. Uh, to be able to like structure that data and be able to deliver it in metric and like it, it click to my head is all a little bit in terms of how you, uh, would go about this from your perspective as a product from study is just being able to give structure to these unstructured conversations that happen in feedback and so forth.

[33:41] Andrew Michael

And then being able to put a monetary value again because you know, the user that you know, who they've been speaking to, you know, what their contract value size is. Uh, and definitely I see that as a much more powerful metric rather than saying, okay, this month the NPSs was this, and these were the main things that people mentioned. The reason for this n p s score, like you're only really getting a small subset of users that are at extreme levels and they don't really account for the general population of your app, but the conversations that are happening all the time, either between customers or through support or in feedback, um, rare have like a much richer data set than just like a standard once off, uh, every 90 days N ps surveys that goes out to your customers, whatever. It's, so I really, really like that. It's like where, where do you go from there?

[34:25] Steve Hazelton

Um, where do we go from there from a product standpoint? Um, there's you, there's just so much that we can do with this. I mean, so just to give you, to give you an idea, the first, the first thing that we do and we kind of gloss over it, is we have to ingest all of this data from all of these data sources. So, and we go direct to the box and I don't want to make a big pitch about, about sturdy, but yeah, so you have email that's this black hole. We ingest your customer, the emails that go to and from your customers. We have some trick stuff that we do. I I, I think we might be one of the only companies that can get data straight from inboxes from about customers. By the way, you don't need a wit, you don't need a plugin.

[35:06] Steve Hazelton

But we clean that up, we organize that. And so if you've ever been in a company where you're like, why do I need to get four people in a room in order to know what's going on for my best customer? Cuz you gotta get the person who, who is managing Zend desk. You gotta get the, the emails from the CS person. So that's all in one view. Um, the thing that we're shipping now is what we call our privacy clean rooms. So as this data is coming in, we're cleaning out all, um, if you carry pseudonym all of the names and the accounts and in the unstructured data and in the structured data, and then if there's any s p I like, if for some reason, um, and this comes up, if someone puts a social security number in an email, we'll destroy it, um, in our system and then we'll let you know.

[35:50] Steve Hazelton

Like someone's sending around social security numbers, which would be really bad. And it does happen, by the way. Um, cuz I, I know when it happens. But, so we, there there's this concept of like, look, if you wanted to hire, um, and I'm, I'm getting a little in the weeds here, but if you wanted to hire a consulting firm to help you with churn or you wanted them to do some data analysis, you can give them clean, um, private privately clean data without even your customer names in it if you don't want it. Yeah. Um, we then continue to create business models and how do those business mo or the language models, sorry, and how do those language models interact with each other? Like, um, for example, we probably, you probably know that about depending on who you are, 25 to 50% of your tickets have a product informing, uh, theme in them.

[36:34] Steve Hazelton

Bug report feature requests, or how do I do this? Those will be the big three, right? In email by the way, it's about 15%, which is crazy, right? Because how many times in your business do you go, where are all the how-to messages in my emails? But they're coming through again in B2B SaaS, it wouldn't be coming through in b2c, I don't think. Um, and then at the end of that, we have a bot. We have a a a, an automations engine, some folks would call it robotic process automation, where we're, we're making the systems that you already use, hopefully better. So like if you have a C S P, we can kick off, we can detect, we might detect like this person requested a copy of their contract and they just had an executive change, five alarm fire pump that into, um, Gainsight and kick off a a playbook.

[37:18] Steve Hazelton

So that's other stuff we can do. Um, there's ever more, the, the most, the thing that we're really working hard on is how do we, how do we allow our users to work with this data more effectively? Like creating a dashboard where you can log in and say, okay, our product led growth group, what percentage of, is there a trend in feature requests here versus enterprise? All sorts of ways of, uh, of, of slicing that data and, and, and, and crunching it up. And then we do have quite large customers, so we have to create this private goes back to their privacy clean room. There's folks in marketing who want to use our data. There's folks in product who want to use our data. How do you inter how, what level of permission do you get when you log into that system? So I, we could go on and on about that, but that's what we're, um, I I really think what I would like to see someday is, well, I know it's happening now.

[38:22] Steve Hazelton

I know that one of our customers today is incorporating our dashboards into their board meetings. I mean that's, it makes a lot of sense, right? It's data that you never knew how to use before. We also have to work out as a company here, like we don't even know what we don't know yet, right? Like is there data inside of this content in this unstructured content that if it's synergized with usage data, if it's synergized with payment data, um, are there gonna be other insights that we're able to, to derive from that? Real simple. We never even think, we didn't even think of this earlier, but I i I would challenge the folks that are listening right now, go into your business and say, how many of our top 100 customers or whatever metric you have for the really important, because in my kind of understanding, it seems like a lot of businesses, like 20% or or or 50% of their revenue or 80% of their revenue comes from like 20% of their customers. Pretty common. I think we've seen of that 20%, how many of them have not, you have not interacted with in the last 90 days at all? Like you haven't sent them an email, you haven't, like they haven't contacted you, they're just doing stuff and you don't know what's going on with them. Um, go and go and track that down. It'll be surprising to you by the way. Yeah,

[39:49] Andrew Michael

We could, we could go on forever, but I see we up on time. I want, I have two questions ask every guest, but I want rapid fire answers, uh, from you on this. So the one I think you already answered and to some degree, but we can get a quick repeat is hypothetical scenario. You join a new company channel, retention's not doing great, you get put in charge to turn things around in 90 days. Uh, the CO says we need, uh, to see some action fast. What do you do? The trick is you're not gonna tell me I'm gonna speak to customers, figure out their pain points and do that. You're just gonna pick a tactic, uh, that you've seen work previously at the company and run with that blindly, hopefully hoping it works with this audience and this customer base that you have. What do you do?

[40:30] Steve Hazelton

Um, I'm gonna, I'm gonna answer it with two things. The two, uh, and I'm gonna limit it to only two. I'm only gonna, the first one is I'm gonna tell you I can't solve your turn problem in 90 days. Like I just can't <laugh> the second one. Like I've actually seen someone get fired unfairly because they said that, don't do that. You can't. The second thing I think from a, from a business to business assassin services perspective, the first thing that I would do is I would make a manager level person do an exit interview of every turn. Now that sounds weird after what I said just now, every turn, like they have to get it. You don't let the person who was the account manager or the customer success rep do that survey. You make their boss do that survey and you make them do it or you do it yourself. And hopefully it's not in such a high volume that you're like, oh, that's impossible. Hopefully it's in the tens or the fifties or the hundreds and not the thousands. But um, that's the very first thing I would do that's gonna illuminate the, i I that would be it if I didn't, if I wasn't gonna go and buy software, if I wasn't gonna go do anything else, if I just was gonna immediately implement a process that would go back to the everybody we lost in the,

[41:41] Andrew Michael

It's sort of, it's sort of a cop out, but I'll let you have it cuz you're still going to speak to customers and figuring out what the problem is. But, uh, we, we short on time so I'm not gonna press you today, but I agree with you like in 90 days, like it's, this is sort of a trick question, that's why mostly it just leads to like short-term fixes, but uh, channel retention is, uh, it's a long-term problem. Oh, you,

[41:59] Steve Hazelton

You said you I wasn't going to be allowed to talk to customers.

[42:01] Andrew Michael

Yeah, you won't be allowed to.

[42:03] Steve Hazelton

I misunderstood that. Um, no worries,

[42:05] Andrew Michael

No worries. Uh, what do you know today about channel retention that you wishing knew when you got started with your career?

[42:14] Steve Hazelton

Hmm. That's a really good question.

[42:23] Steve Hazelton

Um, I think, uh, uh, well I'll give you a little bit of a cop out answer on this one too, but, um, when I ran a, a small company relative to the, in the larger scheme of things, right? If you're doing under 20 million in revenue, people consider you a small company. So I ran a couple small companies for years. I was excited to get bought by a large company cuz I was gonna learn how much better they were at all of this stuff. They'd be better at managing churn, they'd be better at reporting, they'd be better at all this stuff. Um, you're, you're, yeah, I was totally wrong. They're, they're, it's counterintuitive, but I I think the, I don't wanna annoy anybody on this and then call anybody out, but I'm just saying in my experience, yeah, you get worse at it, the bigger you get. And that's something that I wish I would've known right after we got Bob.

[43:26] Andrew Michael


[43:27] Steve Hazelton

That like, it, it's, it's harder the bigger you get. It's not easier and it does and you're not better at it. I've

[43:33] Andrew Michael

Heard, I've heard similar stories as well from like ex-colleagues that were at Hot Jar that left and went to much bigger companies afterwards and sort of like they said, the level is nowhere near where we were at, uh, at those early days. And like the level of sophistication, like they were shocked. They said like, I didn't, I had the same like there in their minds I had the same thoughts as you that there would go to place that had everything figured out and had all the answers and yeah.

[43:57] Steve Hazelton

The level of sophistication. Yeah. Um, interesting. You're probably more sophisticated now when you're a, a smaller company than you will be when you're a large company. You're more sophisticated at dealing with it. Yeah. Yeah. Um,

[44:11] Andrew Michael

I think it also makes a little bit of sense because maybe at the earlier stage you obviously, like in terms of hiring the, the bars probably a lot higher because you have less resources and you really need to make things happen. But then more so it just becomes like the ownership thing. It's like, uh, the larger you get there becomes like less ownership over the problem in a way. So it just becomes sort of so many different people doing a similar thing that there's no real, like, I don't know how bring everything together. I, in my mind at least this is how like I see.

[44:42] Steve Hazelton

No, I think that someone answers your first question that you, that I like hopped out on like from a tactic standpoint. Like if someone said, come on board right now, and the first tactic I would implement would be like, every time someone requests a copy of the contract, every time someone asks about a renewal date, I don't care what they say. Like, when do I renew? We got a new executive, which we used to call a new sheriff in town that is getting forwarded to me every time, right? Like full stop. The thing that I wish I knew is that when you get in a big company that doesn't work anymore. Like it'll work for a period of time and then people will, you, you'll be able to get it to work for like three weeks or a month or two months and then eventually over time it will degrade and it will stop happening.

[45:25] Steve Hazelton

Um, so you'll be able to get these, these really like manual processes in place, which it really, if you think about it, that's a manual, a human being manually recognizing something, manually collecting the data, classifying it for me, and then automatically rooting it to me. It's like they're doing a robot, they're doing what a robot could do. And a lot of companies we're wasting money, making our people do robotic stuff. Yeah. Um, that would be the first thing I would do. And I think the understanding is that will fail at a big company eventually. It, it will never, you'll never be able to pull that off at scale

[45:58] Andrew Michael

For sure.

[45:59] Steve Hazelton

I would, no, I would never be able to pull that off at scale. Other people might, might be able to.

[46:04] Andrew Michael

Yeah. I see. We are up on time though, so it, I think we are continue discussing forever on this topic. Uh, is there any final thoughts you wanna leave the listeners with before we finish off today?

[46:16] Steve Hazelton

Um, I guess right now keep your head up. The, the markets, um, certainly interesting. Uh, and the folks, my experience, I've run managed two businesses through a recession, crash and the financial crisis. And if you make it out the other side, you're gonna do fantastically well. So just, uh, keep your head up and survive. It will hone if you make it through to the other side, you'll know your customer base better than anybody. Your customers will be there for you and you'll, you'll be ready to go when everybody else gave up. That would be my only thing to say, I guess.

[46:52] Andrew Michael

Nice. Uh, yes sir. Thanks for joining today Steve. And for the listeners, like anything that we mentioned today, we'll make sure to leave links to in the show notes. Uh, really, really appreciate your time and wish your best of luck now going through weathering this, uh, next storm, uh, that we're all in at the moment.

[47:07] Steve Hazelton

Yeah, thank you. I think it's a very topical discussion for what we're, what we're in for right now. So it's, it's perfect timing.

[47:15] Andrew Michael

Absolutely. Thanks Steve. Cheers.

[47:19] Steve Hazelton



Steve Hazelton
Steve Hazelton

The show

My name is Andrew Michael and I started CHURN.FM, as I was tired of hearing stories about some magical silver bullet that solved churn for company X.

In this podcast, you will hear from founders and subscription economy pros working in product, marketing, customer success, support, and operations roles across different stages of company growth, who are taking a systematic approach to increase retention and engagement within their organizations.


Listen To Next