Is your company good at customer success and retention? Chances are that you could be better.
For most businesses with a recurring revenue model, customer churn is a very costly affair. Whenever a customer leaves, you lose out on recurring revenue, forgo the opportunity of expansion (cross sell) revenue and have to pay for another round of acquisition costs to cover the loss.
In my personal experience, customer retention is both art and science. Machine learning and other data science techniques can be used to identify customers who are likely to churn, but it is equally important to craft meaningful and delightful interactions throughout the customer lifecycle.
So, what’s required to become a lean, mean retention machine?
In this episode of Leaders of Analytics, I speak to Sami Kaipa to learn the best practices of data-driven customer retention. Sami is an experienced technology executive, serial entrepreneur and start-up advisor. He is co-founder of Tingono, an AI-driven customer retention platform.
Listen to this episode as we discuss:
- Sami's journey as an entrepreneur and corporate technology executive
- The core elements of customer success and retention that every business should master
- A deep dive into the concepts of customer retention, expansion and NRR
- The economics of customer retention and expansion
- How data science and machine learning can help with retention, and much more.
Connect with Sami on LinkedIn: https://www.linkedin.com/in/samkaipa/
Tingono's blog: https://www.tingono.com/blog
Sami Kaipa, welcome to leaders of analytics, it is so good to have you on the show today. And we are going to cover some really, really interesting topics such as your background, of course, what the company thing gonna does your startup and entrepreneurial history, as well as how to create data driven customer success. And I for one, I'm really interested in hearing your knowledgeable insights on those topics. So send me let's kickoff. And let's hear from you a little bit about yourself, your background and what you do.
Yeah, great. Thanks so much, by the way, Jonas for having me on. It's a pleasure to be here. So my background. So by training, I'm actually a computer scientist, in fact, not a data scientist. And I've worked in enterprise software, my whole career. At the same time, the through line in my career has always been data, right. So early on, I worked in a space that at the time was called e ai enterprise application integration. Think of it as kind of the predecessor or the precursor for the modern data stack. So I built skills in data integration, transformation, cleansing those kinds of areas. And then if you fast forward several years, I ended up starting my first data analytics company in 2012. It was called Glimpzit, we get acquired in 2018, by Forrester Research. And after a stint working with, you know, a very talented team at Forrester, again, on a data analytics platform, my co founder, and I decided to go at it again. So we started ting goto earlier this year. So I guess you could say my career in a nutshell has been a mix of data wrangling and data analytics, with the latter half of my career being in a startup or an entrepreneurial setting.
Yeah, so your background is really interesting, because you've sort of got this mix of being inside big corporates and then jumping out of them again, and doing your thing and creating startups that you've put up for sale. So and you actually found that more than one I don't know if you'd call a data company or AI company, and you also act as a board advisor on a good handful of startups. Could you tell us about your push your entrepreneurial journey and what made you follow that path and your career?
Yeah, so you're absolutely right. I've been in and out of small and big companies. I had an early experience my first job right out of university at a small start. It was called Crossroads software, we eventually got acquired by IBM. But I always had fond memories of that experience the people and just the working environment, but most importantly, the impact I was able to make there. Right. So I knew I wanted to get back to that at some point, you know, I took the plunge almost unaware of the implications of being a founder, as opposed to just being an early employee at a startup. And it's one of those things where you don't know what you don't know. But luckily, I made that leap. Because it's been the best career decision I've I could have made. And if you think about the entrepreneurial journey, and what like, made me follow that path, the way I think about the draw of entrepreneurship is in basically what I call time to impact, right, as an example, in a 50,000 person company, you can decide to do something. And then you know, five months later, approvals from budget and approvals from the stakeholders, you can act, right. And in a shrink the size of the company down to 5000 person company, maybe it's you can reduce that to five weeks. In a startup, you can make a decision, and you can act within five minutes. And that's really the big draw of entrepreneurship, acting with urgency and having just a huge impact.
And it is hard though, you got to have a certain mentality to succeed in that world. I've tried it myself. And I've found some elements of entrepreneurship, super liberating, and other things very hard. And I was actually watching this YouTube video last night and the guy on the he the name escapes me, the guy and they said, basically, if you want to build your own thing, you got to be willing to just get a big spoon out and eat shit all day. So pardon the language, everyone, but that's I'm just relaying a quote here. No, you're right, which pretty much means there's nothing that enjoyable about it. It's been a bit polished in media, by all the 20-something billionaires that have come out in the last 20 years. Everyone wants to be Mark Zuckerberg or what have you. But it is actually most of the time full of rejection, for running out of money and all the things that are not glamorous at all. So you must have had some sort of traction to that or affinity for punishment, perhaps. But also, there's something there that attracts you to overcome that. Could you tell us about what that is for you?
Well, yeah, and just to comment on what you're saying, You're absolutely right. A common refrain I hear is, in a startup, you get to be your own boss. And that's the worst reason to go and start a company. Because everyone becomes your boss, essentially, your employees are your boss, your investors are your boss, your customers are your boss, right. So you really do have to go through the grind. I think that it really is just people who do well in this is really just persistence. I didn't have anything magical behind me in terms of intellect, or the ability to turn on a dime or anything like that. It really was the sheer will and persistence of me and my co founder that grew that business to where it needed to go. And then again, We willed the acquisition to happen as well by making the right connections with folks at Forrester and some of our other companies that we were working with. And that's what made it happen.
Yeah, so when you wanted to sell the company, you did you sort of you had to go and market the company and sell it just like you sold the product to customers similar way. And it's that kind of what you're suggesting here?
Yeah, absolutely. In fact, it sort of hit us on the head. To start with, we were actually going after our next round of funding. And we had a couple of term sheets. But coming back to some of the things we were saying earlier, fundraising is glamorous in the media, but it's never a fun process. And or the outcomes are really not that favourable, generally for founders. So we had some term sheets for our next round. And we got approached by one of our advisors who was starting an m&a brokerage firm. And he was basically like, Look, I know of three companies who have actually bandied about your name in their corp dev meetings. So you kind of owe it to yourself to see where this would go. And we were floored at that. But we decided, hey, that's, that's a good path to take, just to see what kind of outcome we can get for ourselves and for our early investors. And that's what we did, ultimately.
Yeah. So you have these lucky events that happen, but you got to put the grind in to get there, right? It's not like you build a company, you have your brilliant idea for other no flying saucers or something. And that just you make the product and people just come and buy it that's far from what happens. And now Sami you read it again, because you're now the co founder and CTO of Tingono. Now, could you tell us about this company and what problems you solve for your customers?
Absolutely. I'm excited to and I know we'll talk about this more but there's a financial metric called Net Revenue retention or NRR. Sometimes, people call it net dollar retention , NDR. and it is essentially, it's the measure of a business's ability to retain and expand their existing customer accounts. And I put that in stark contrast to gain new customers, right? It really is the focus on your existing customer accounts. And it's a major concern now of subscription or recurring revenue based companies, right. And so in a nutshell, my new company, it's called Tingono, enables SaaS companies to boost and IRR to boost their net revenue retention. And the way we do that is of course with data, right? So we analyse a company's set of customer data from their their enterprise systems that could be from your CRM, Salesforce, your product analytics system, as well, your ticketing and support systems and so on. And we build what we call use case driven machine learning models to predict opportunities and risks to NRR. Right. And when I say use case driven, I mean, we've physically build models, machine learning models that help you identify all the ways that you can positively impact on our that is, you know, cross selling and upselling opportunities. And then we also build models to help companies effectively manage everything that negatively impacts NRR, that is downgrade and turn risks.
So you have a platform or tool, but it really is applied and bespoke Lea uniquely to each company with its idiosyncrasies in the data and, and structures and processes and so on. So is that fair to say?
That's absolutely correct. Yes. So why did you start this company? Yeah, well, it was purely based on the pain we personally felt, right. So the last two businesses, let's see, Glimpzit from 2012 on and then at Forrester, as well, we ran a business division there. And we directly felt, on one hand, the positive effects of focusing on existing customer growth, but also the pain of flying blind, really, right not having the data and insights we needed to determine which customers were likely to turn, which ones were likely to upsell and so on, right, who was going to expand their spend with us and who was going to actually leave us at Forrester, we ended up actually building a data science practice to help us with some of these metrics. And, in fact, help us with operationalizing it through dashboards and stuff like that. And then we realised, you know, in that process, we interviewed just a tonne of other sort of peer companies that were working on SaaS businesses. And we realised that every other SaaS business we talked to had the same challenge, right from the, from the CRO, all the way down, had this challenge of, hey, I need more data to understand. And I need to not only diagnose what has happened in the in the past with my business, but predict where my customer accounts will go in the future. And so the light bulb went off, and that's when we decided to start to go.
Yeah, it is very common, isn't it that it comes out of your own need. And it's actually one of the best ways to start a company because you really understand on the ground what the problem is. So Sami, you're talking here about SaaS companies. So for listeners, that Software as a Service is a SS, just in case you didn't know. But I think a lot of these metrics and ways of post optimising your business can also work in in more traditional subscription businesses, such as banks, insurance companies, energy companies, telcos and so on, because it kind of is the same thing. I know Netflix, they show you videos, but you're kind of subscribing month and month, but you're also subscribing to a phone subscription, for instance, in a similar way, and that can be upsell and cross sell. So um, I think we can perhaps talk about more broadly, what are the things that companies whether they are this modern SaaS version, or a more traditional business model? That is nevertheless a subscription type business? What are the elements of customer success and retention that every business like that should master? And in that, could you explain the concepts that you talk about in terms of customer retention, expansion and NRR?
Yeah, absolutely. So I think the narrative goes that with the advent of the subscription economy, and just like you said, more broadly, it doesn't even have to just be software based subscriptions. Customer Success has just grown as a discipline as a function in the business. And customer success, through necessity has really rapidly matured, you need to be able to retain your customers in a recurring revenue model. And there's just no question about that. So I think of customer success as evolving from like, sort of gen one to two what we're in now, which is gen two. Gen one is really customer success was establishing kind of the centre of excellence. For what that function would be, right? What's the best practice for establishing a long term relationship with a customer, onboarding them for ensuring that they generate value from your product, and keep them aware of that value, it's not just enough to make sure that they are, you know, to have them generate that value, but to keep them aware of that value, and talk to the right stakeholders to make sure that they're aware of that. And that's in the form of the very common QBR meetings that that customer success, folks have QBR, of course, standing for a quarterly business review. And that sort of encapsulates gen one, then there's gen two, which is really leveraging data to accomplish, what I think of as kind of the two critical things for for this next round of customer success evolution. One is, is using your data to really understand your customer journey. So when do your customers get turned on to your product, and it's really across your entire go to market discipline, right? When your customers get turned on to your product is more like a marketing thing? How do they experience the sales cycle, what goes into the product purchase decision, what factors impact their attainment of value from your product, and all the way through to, you know, the things that we're talking about finally, renewal, staying with you as a, as a business as a company, and actually spending more with you, right? So you as a company, you have to viscerally understand that customer journey. And then you also use data. I mean, this is what we do, right to know is you use that data to effectively build some sort of churn and or expansion prediction models using your historical data. So step one is kind of understanding your customer journey, that's the diagnosing the past part. Step two is really having an effective model and meet it need not be a machine learning model, but just some sort of model for predicting the future of where your customer accounts are gonna go. So I think those are kind of the the basics for customer success. 101, you asked specifically about NRR. And I do want to just put some definitions in the ground on that as well. Because it's just super powerful, right? Especially in a recurring revenue business. You start by basically ignoring your efforts to land new customers for a moment. And it seems kind of bizarre, but you know, grant me that at least to start, right. Just take the cohort of existing customers that have subscriptions with you. Expansion represents all the way you're able to increase their spend with you. So you can cross sell them to other products, you can upsell them, you know, from one tier, let's say like a basic plan to a pro or an enterprise plan. That's expansion. And then the opposite of that is usually referred to broadly as contraction. When a customer downgrades their level of service, then that's contraction. So downgrading is is zero, right? That we call that churn, basically. So these two concepts make up an RR. It's the recurring revenue you get from your existing customers at the beginning of the time period, you add the expansion revenue, you subtract your contraction revenue, you divide it by the same recurring revenue, and that's your, your NRR.
Yeah, and it is it is a very powerful concept, not in the least, because for a lot of the subscription type businesses, the marginal cost of product delivery is not necessarily zero, but but very low, close to zero, right? So there's very little cost in having someone continue to use your software. But there's a lot of onboarding costs and getting it on getting customer set up and getting them to use it and all that. So this is also part of the equation that how do you avoid that? And how do you sell to the ones where you've already achieved that and gotten over that hump? Now, Sami, tell us about how, with some specifics, how can data science and machine learning help this course? How do you see that progressing? And could you give us perhaps some examples of how you use it in day to day to to help increase revenue and reduce churn? Yeah,
so believe it or not, that's actually the simplest part. Right? So once you've, once you've nailed down the business use case, which really does take some thinking, right? The data sciences part that just sort of follows, right. And we know that when you're actually setting up things like data science, Analysis and Modelling, you really have to get down to the level of detail of defining your goals, defining your terms, and so on and so forth. So, you know, we actively work with customers on asking a question that seemingly is very obvious, but we ask them, like, define churn for us. And you think oh, well, that's when a customer leaves you but it can be defined in so many very specific ways. Does that mean at the end of a renewal, they leave you can you have mid-cycle churn is you know, a customer who spent 80% less than they did last year could that can be considered churn and so on and so forth. So really, you have to get down to different Finding these business metrics first, but then the data science becomes a simple part. And the way to think about it is, you know, if past occurrences of churn and expansion are indicators of the future, then really, this becomes a classic supervised learning problem, ultimately, right, we have a significant set of tabular data, ultimately, right account behaviours over time, have they turned Have they gone up in spend, and so on, and then affiliated in time in a time series with that we have other business signals that we're hoping can explain those those account behaviours, right. And in, you know, I'll come back to classic SaaS and software as a service. Those are usually things like logins to your platform, or time spent in a session on your platform or tasks completed in the app, but they need not be product related. They can also be sales related, like pipeline, pipeline progression for a particular opportunity or deal. But they can also be on the other side of the go to market, Journey support, tickets logged and things of that nature, right. So any number of these factors can be involved. And you know, as we know, in machine learning, we call these features, we call these features of the the model. And ultimately, we're trying to figure out whether they have an impact on churn or cross sell or upsell. And just to give you the last bit of detail, we run a time series analysis to be able to predict future opportunities and risks. So like I said, it's really a classic data science problem that can be brought to the space.
It is a classic data science problem, there's sort of two parts to it, in my mind, send me you reminding me having worked on a similar problem some years ago, where it was also a subscription type business where clients, we had a very particular moment where there will be high churn, because customers would be on a fixed contract for some years. And then they would after though she is move on to a month on month variable contracts. And the pricing would also typically change at that point, because there was sort of a reset point of sorts, right, that had a fixed contract with a fixed price for a fixed length. And same issue, we found that certain login activity and account activity prior to that date indicated that some customers were more likely to shop around and others and so on. One challenge we had was, was not identifying who they might be, but actually coming up with a compelling, offer a reason to stay that was also viable and profitable for us, as the company selling, selling the product. So I assume here that part of your thing is identifying but the other part is sort of how do you tailor the right solution to prevent the tournament, and there can be, of course, pricing the features in a product, but also just the thing that we talked about earlier is how do we just not upset our customers throughout their journey? How do we avoid having them lock support tickets? or what have you? Is that part of your sort of sphere of approaches to the problem, I suppose?
Yeah, absolutely. And I love this question, by the way, because it's really the right question to be asking, let's say we've effectively built a churn prediction model, and we can give you a risk probability score for each of your customer accounts. And let's say we've predicted that a particular you know, one particular customer is most at risk. Right? What do you do with that? Right, the relevant next question is, is exactly what you're talking about? Is the why, right? Why are we predicting a customer is likely to turn? What are the factors behind that? Right. And hopefully, those are leading indicators, right? Those are leading indicators that signal turn to the model. It's crucial that we answer this why question. So every churn prediction model has to have an element of, you know, I'm gonna throw out an ML term, but of explainability, right. That's the way to really identify the leading indicators for churn and examine them for each customer, for each customer, but also for each customer cohort you care about, since they aren't all the same, right? So let's say once you explain the reasons behind churn, that's when you take the action, and hopefully, you can tie those things together, right? activity, the customer is doing this activity. How do we tie that to what is the next best action for us to prevent churn or to promote an opportunity for upsell or something like that? I'll start by saying in some cases, you might find that letting a particular customer churn is actually the best action you can take. Right? They're a bad fit, and it would take far too many of your customer success and maybe even engineering resources to retain a customer like that and it's just not worth it. Right. So that that's something that you should be aware of as well, but in a lot of other cases, knowing ahead of time if a customer is at risk, and why right gives the CSM team a huge leg up in the pursuit of retention, common activities would include delivering something like a value impact assessment, making sure your customer understands the value of your products, engaging more customer stakeholders, this is like a tactic making sure you have a good relationship with not just sort of the key person but more stakeholders at that company. But then I think you mentioned it is some of the the low level tactics are like offering free trials of new features to provide more value to the customer. And in fact, we talked about some of the pinnacle of kind of this data analysis would be to eventually tie actions that CSMs have taken based on the knowledge that some of these companies are going to turn and what the risk factors are. And then tying that back into the model itself. So we can actually just output things like these are the actions you should take. So ultimately, I'll end by saying that the explainability piece is huge, why a customer is at risk drives the actions or a set of actions that the CSM should take.
One thing that comes to mind when you say, Oh, that was from again, this sample that I talked about, when we presented the problem and potential solution that we could use data science to pick out who was going to churn. The first initial response from the business or sort of approach or reaction was great, let's give people a discount so we can retain them. And that is, of course, potentially useful, and potentially very dangerous. So what we did was we took a sample, a cohort, sample cohort of churning customers that were coming up for renewal, and we took from that a holdout sample and a test sample. And we tried different discount offers that were at different amounts, but also given to communicate it to the customer in different ways. So either via letter or phone call, either proactively or reactively. Right. So we might, we could either wait for the customer to call us and then we would give it, the account will be flagged or we will proactively reach out and say, Hey, we know you're coming up for renewal, here's a great offer. And what we found was that often the customers that were looking around, were savvy enough to know the market well, and therefore the discounts we had to give were sort of pretty steep. And what that meant in reality was that we were giving discounts proactively to people that will change but also a bunch that wouldn't churn and we had to give steep discounts. So once we compared the two cohorts after the event, we could see that we had retained a lot more. But we had also given steep discounts to people who wouldn't have turned anyway. And not at least not at that point in time. Again, this is a long, this is a long thing, we don't know if they would have been trained a year later. But But time being in that short window of maybe six months, we actually gave away more discounts than we would have lost in revenue from from the cohort churning. So this is a typical scenario where discounting is fraught with risk. So it's one of those that actually, you might want to avoid.
I agree with you. If I could just comment, I'd say you're absolutely right. There's many ways of being proactive. You don't want to land yourself into a landmine by any means. But what you wanted. I mean, that's why I mentioned this thing about like, understanding your customer journey like this, really understanding it. So being proactive within the framework of your customer journey, and your customer playbook. And just being disciplined to follow that is really the thing that will set you apart in terms of being proactive. You know, I mentioned earlier that if a customer is a bad fit, if you give them a discount, they're still a bad fit, right? So it's something that you want to avoid and just be disciplined with your playbook.
Yeah, and I think that last bit, there is a great call out. And it really highlights the fact that if we take a step back, getting this right, it's not something you do in a few months, it's typically a long journey for some companies multi year, and you shouldn't do it, when the tides out and you're standing there naked on the beach, all of a sudden, you need to, you need to start this while surfing skirt, and you've got a lot of revenue coming in. And you don't have a churn problem. Because when you have churn problems that that's when panic ensues. And you do need to make things easy for me to sit here and say, be proactive, but it's nevertheless, it's true that it's because what I'm hearing you say Sami is you got to really be very meticulous with the detail and group clients and customers up into really small cohorts, often and give them sometimes bespoke treatment. And then, of course, can be more or less profitable, depending on what type of business it is. If it's Netflix, and everyone's paying 10 $15 a month, it's different to a SaaS Customer, that is an enterprise where you're paying $2 million a year, they can get a lot more love, individual app, but again, very different scenario. And on that Sami, could you explain to us sort of the typical economics of customer retention and expansion, as in? How valuable are new versus existing customers? And I know this is a very hard question to answer, because I've just given an example of two completely complete opposites with the Netflix versus large enterprise customer example. But nevertheless, there's sort of some finance, economic, or financial dynamics in this that are common or similar across the SaaS space or recurring revenue space. Could you talk to us about that? And how people should think about that?
Yeah, absolutely. And, you know, I think you and I are both data, data driven folks, and analytical by nature. So I think pointing back to the map really helps us to get started. And then we can sort of layer on top of an analysis. All the math tells us that the best way to attain amazing growth for your recurring revenue business is to focus on your existing customer base. And the bottom line is pretty simple as is, you mentioned it earlier, the CAC, right the cost of acquiring a new customer is just multiples, higher than retaining your existing customer base. So let's understand this at the level of the map, right? It's rather expensive to acquire a new customer, there's a goldmine in terms of revenue growth, when you can keep existing customers happy. And when you can quickly expand their spend with you as well. So I'm going to come back to this one financial metric. But I mentioned earlier that the health of a recurring revenue business is measured and NRR. Why is is NRR. So important, just take this scenario, right? You aren't adding any new customers, if you have an NRR of 100%. And you don't add any new customers, your revenue stays the same. That in and of itself is an achievement, right? If you have an NRR of 110%, you double your revenue in eight years, again, without adding a single new customer. It basically means you know, and then we can go even further to that. An NRR of 120% means you don't double your revenue in five years. Again, not even adding a single new customer, it basically means that you could fire your entire sales staff, and you'll still grow your revenue with a positive NRR. And that's really the kind of the profound impact that focusing on existing customers really has.
Yeah, so do you have any success stories that describe this example? Or maybe also the opposite of great product, but they didn't create the proper revenue stream?
Yeah, absolutely. The success stories are a couple of obvious ones. I think that I wouldn't be the first to mention but snowflake, right, a data warehousing and analytics company. It had an nr of 158% at the time of their IPO just tremendous. Right. And
so break break that down for us. How does that occur? Do you sort of as an as an outsider looking in, what have they done right in that process to achieve that.
Yeah. So I mean, first of all, I wouldn't call it luck, I would call it that they have the stickiness of Sticky, sticky products, right. So basically, what an NRR of 158% essentially means is, you basically don't have a term problem, you don't have a systemic term problem at all. I think they had something like 3500 enterprise customer accounts and something like 23 instances of customers leaving, right so that is very much negligible churn. And on the upside, they had a set of products and, and kind of a pricing model for those products that lead folks to adopt product A, that snowflakes old, and easily be able to adopt product B because they could see the value overlaying product a product C, product D, and so on and so forth. But also just move up in terms of spend with the usage as well, right? Because they have a usage model for for pricing. So that's what leads to basically NRR of that astronomical number, right. And at the time of IPO, I think they were considered the biggest software, you know, IPO of all time. So that was that was one example. Another example, I'll just mention real quick, because it's been in the news recently is the design firm called figma. Right. And they were acquired by Adobe, recently, something like a $20 billion valuation. Their NRR was 150%, at the time of that acquisition, and I think the takeaway from both of these examples is the point that NRR is a direct measure nowadays, not just for recurring revenue, strength and recurring revenue growth, but also it's a proxy for company valuation. Right. And you'll hear that from m&a brokers and corp dev teams nowadays as well.
Yes, so some of these so called Stupid valuations that you get in the market of really, really, really high share prices and market caps, they are often tied to this multiple of NRR. And just seeing okay, if they add x 1000, or X million customers a year and those accounts broaden in our era, why then all of a sudden, we can see lots and lots and lots of money, falling out the back that in 10 years, really is something that companies should spend time and energy on getting right whether you are a SaaS company or a more traditional subscription type business. So let me let's get back to gonna for a moment. So your your talk about the platform being based on auto ml, could you talk to us about what that means, specifically, and why you've chosen to build a solution or focus on a problem here, that is sort of relatively narrow scope, both in terms of the problem you're trying to attack, but also the use of group? Suppose.
Yeah, and ultimately, I want to convince you that a relatively narrow scope is actually the key to to auto ml, especially, sort of democratising that. But let me take a step back, because, first of all, we took a conscious decision to create custom or what we're calling micro tailored, machine learning models for each and every company we work with. The other approach would have been to create a generic model that works for every SaaS company or every subscription based company. And clearly, we took the harder path, right? So but despite having taken that harder path, we just felt it was the right approach. I mean, every company is unique. The reasons or the features and models speak right that determine why a customer decides to turn for let's take, you know, a company like Slack as an example are different than the reasons why, you know, Zoom customers churn, right. And they're, they're specific down to product usage and product features. So the model really needs to reflect the uniqueness of these businesses. And that's why we chose these, these custom, the custom model approach. And once we made that determination, we had to figure out how to scale the custom model approach. So that we're not hiring, we're not scaling our business with data science, data scientist hires, ultimately, right. And so that's where auto ml comes into play. We're effectively only building AutoML as the underpinnings of our analytics platform so we can achieve scale for these models. What we're not in the business of is building a general purpose like auto ML platform as a toolkit for you know, the customers data scientists who can play around with the auto ml capabilities, that's just not the business that we're, we're in our flavour of, of auto ml should be essentially, you know, opaque to our customers. Because our end users are not data scientists, right? They are sales ops or sales account managers or CSMs. So in fact, I'll come back to what I said earlier. So in fact, a relatively narrow scope in terms of the problem scope is and the use case is a good thing for us. Because we're bringing our models directly to the line of business, there are a lot of variables that can go into building these models. And if you just narrow the scope down to hey, we need to predict the likelihood of churn and describe the factors, that is something that can very effectively be delivered through an auto ML platform.
Hi, there, dear listener, I just want to quickly let you know that I have recently published a book with six other authors, called demystifying AI for the enterprise, a playbook for digital transformation. If you'd like to learn more about the book, then head over to www dot leaders of analytics.com/ai. Now back to the show. Yeah, and as an outsider, having spent a lot of my career on churn problems in various organisations, so and outside of the thing going on, of course, here, but from a data and analytics point of view, and insider to the problem, I think it makes a lot of sense, because it is a somewhat generic problem in almost every industry, and it is very time consuming. But it is also bespoke, as you say to the organisation itself. So if I paraphrase what you say, the audit platform is for you to be able to scale scale your model production, at that model production is always unique, because it pertains to the features or this the situational context of that organisation. So send me that makes me think that what really is critical at the end of the day is input data, right? The data that the company has on it customers and their behaviour? And how do you work with clients to ensure that this customer data is available and at the right quality to use for you and your solution?
Yeah, that's a great question. You know, I'll say right off the bat, that we're not solving the garbage in garbage out problem, I think that's, that's not going to be solved by by an application like this. But I think most of what you're talking about is going to be teased out in the sales process. The companies we work with need to have at least two years of account and product usage data, we need to see enough instances of expansion, enough instances of churn to be able to train a model to identify these occurrences. And I think we do employ some clever techniques. You know, it's an auto ML platform. So we need to wrap these into automatically doing these activities. But we identify things like low volume or sparse data scenarios. And we do what we can, you know, some of the techniques that we use are like tabular, Ganz, generative adversarial networks, and the like, right. But again, there's no magic bullet to these sorts of things. I guess what I'll say is, while there is still a range of data maturity amongst, you know, some of these SaaS companies, on one extreme, you know, there's a company that hasn't even adopted best of breed, go to market systems like Salesforce or HubSpot for CRM, they are becoming increasingly rare. I mean, I think I'll put that out there as kind of an axiom these days, we're finding a tonne of, you know, even on those kind of later stage, small companies, but definitely on the medium and larger sized companies that have fully embrace not only these enterprise systems, but the modern data stack, right. So they have a robust data warehouse, they have instituted ETL processes that push data and pull data from your enterprise systems to the warehouse and vice versa. And obviously, we can do a lot more with that type of company than than the former.
Yeah, that makes sense. So there is a minimum level of maturity required to sort of have the ticket to play. That makes total sense. So many, this has been really, really interesting. I have a couple of questions left for you, before we wrap it up. The first one, which I always ask of any guests on the show is to pay it forward. So who would you like to see as the next guest on leaders of analytics and why?
Yeah, that's a great question. I thought about this. Whether we can actually achieve getting these folks on the podcast is going to be a question, but I think it would be awesome to have someone from the open AI Institute. I'm sure you're very familiar with this institute, run by Sam Altman or Greg Brockman. But really someone who's on the innovation side and is is fully aware of what's happening on on the innovation side, and that would be interesting to just hear them talk tech. There's so much buzz around GPT three, obviously, but I'm sure they have a perspective on What's the future of data science and machine learning in particular? In fact, they're they're probably already working on it. So they could just tell us. So I think, you know, I'm fascinated by future telling or prognosticating, right, and I'm sure the listeners of leaders in analytics would be as well.
But let's see what I can do. Sometimes I surprise myself with what I'm able to convince people to do. So I will definitely be chasing up some of the people from open AI. I think that's a great suggestion. Sami. Lastly, where can people find out more about you and get a hold of your content?
Yeah, well, I independently blog on LinkedIn, when I find the time mostly it's about startups and entrepreneurship. But nowadays, I'm mostly 100% focused on blogging through Tango now, so you can go to, you know, very simply Tingono.com, tingono.com/blog. And you can check out my content. And also, just find me on LinkedIn and reach out and I'm happy to talk to people directly.
Yeah, and listeners, I really do recommend you go and check out Tingono's blog, there's a lot of content on there that describes in written format, how to go about thinking about that recurring revenue and customer success and the metrics of scaling your business using your existing customer base. And the more you read about this, the more it'll sort of define itself in your head. And if nothing else, you will get lots of funny memes. And they are the things on that website, because it's their content, it's a bit of fun to read as well. So you should go and do that. It's not boring university type texts. It's it's quite engaged. So well done to us, me and the colleagues who are producing that content. Sami Kaipa. Thank you so much for being on leaders analytics. I really enjoyed this conversation. And I wish you all the best with synchronous success and the success of your customers. And let's hope you bring down churn rates all over the world with this product.
That is absolutely the goal. And it was really a pleasure talking to you as well. Jonas. So thank you very much for this opportunity.