Rohit: This is Rohit Nallapeta. I’m here to talk about SaaS, PLG in the New World order. I’d like to get Satish Patil, a data scientist, and an entrepreneur, to talk about this whole session.
Okay, folks, let’s get this party started. Like I said, we usually talk about SaaS product-led growth world. Today I thought I’d bring in a good friend, a great collaborator, a data scientist, and an entrepreneur Satish Patil onto the stage. Satish, please introduce yourself. You can unmute.
Satish: Hey, thanks. Great, thanks for the invitation.
Hello everyone. I’m Satish, founder, and CEO of a product company called Mitibase. We are our relationship intelligence platform. We help revenue operation teams in B2B mid-market companies to help discover massive revenue opportunities. which are deeply hidden in their collective relationships, and we work very closely with the ReOps team. In this journey, I’ve done my Ph.D. in data science from the University of Minnesota Twin Cities in the USA and have been a practitioner data scientist for the last 12 years. So that’s a quick intro.
Rohit: Satish, thank you. There’s a lot to unpack with your introduction. I’ll come to your introduction in a second. Thanks for that. So, you said a couple of words, so I just want to unpack this for a couple of minutes. You said relationship management, ReOps teams, hidden collective intelligence, why don’t you explain a little bit of what does this all means? Why is this all-important? Give us a little bit of information.
Satish: Sure, sure, thanks. So, what happens when we say every organization has a relationship capital, right? And when people say hey, we know our relationships at the organization level, people always think about only the traditional CRM, right- where we house customer interactions and the prospects and leads right? But as an organization, do you have a massive network of relationships, which is typically hidden? Which is not really visible. It is a network and the relationship of all your employees, right? Or the network of your clients, your partners, your investors, your supporters, right? Your ex-employees, right? And all this information is not to be tracked. Then what happens to that right? So, we help and we believe in especially in B2B space where a relationship is the key, this hidden relationship intelligence can bring a lot of massive revenue opportunities.
Rohit: So, you pretty much making that even just from the data science perspective, forget anything else, in the B2B world there’s a lot of hidden relationships, and that itself can make a massive opportunity. So yeah, we’ll come back and visit this. Now let me ask you this question. Look, we’re talking about SaaS, product-led growth, and there’s tons and tons of data that we’re generating. Now you, as a data scientist, what is your view about all of this data? How should we deploy and leverage data science to make this complete?
Satish: Yeah, very good question, Rohit. So, if you look at the data part in, especially in the SaaS world, one of the challenges we have is we are not able to, first, aggregate the data because data is in silos. Second, we are not able to track that properly because we have limitations in terms of what we try, and third, how do we make sense of that data right?
So, if you look at, let me give an example for example at the customer acquisition level right or the product usage level right, or the segmentation level? All these different areas in the SaaS world, where there is a lot of opportunities to track the data, right? And make sense of the gap, but currently, very few companies are really having that other playbook ready. Who can do that effectively and efficiently?
Rohit: Got you, so you’re saying there’s a bunch of contextual aspects to this world that is being ignored and there’s a bunch of again, just for the sake of the user’s thing it’s not about tracking you generating all of this data, but it’s not being leveraged. That’s the point of view that you’re talking about.
Satish: That’s absolutely correct, so let me give a very specific example. For example, let’s say the lead qualification as one specific area, right? So, where you see there’s a lot of data which we typically don’t track. And what if we can really use a lot of data at our disposal to identify those limits properly. So, let me give an example. So typically, people have a regular matrix like marketing qualified lead or the sales qualified lead, but in the SaaS world, there’s one aspect called product qualified lead. Where we typically don’t look at the product usage, right? And can I use product usage data as one of the qualifying criteria? So typically, people call it as a PQL product qualified lead and I can use this data and I can build a bunch of machine learning models to really predict whether hey, you know, based on this app usage, is this guy going to convert the free trial to the bait trail, right? So, a very specific example I just gave you right?
Rohit: Correct, correct. I mean, I think this is the same problem where, when I’m building Glance, we’re talking about product-led growth. It needs is one of those nuanced, what do you call, data input that hides between all of this. But then the SaaS world is also the PLG world where the product-led growth where the product has to do the heavy lifting is also the world as a world of data silos, right? There’s a lot of challenges. Where do you see data science playing a big role here? Because I know look Satish, a lot of people over here know the systems, do the thing. So, what we’re trying to do is. Where can we leverage it? Is it the volume of data that is important for data science, is the quality? Is it the nuance in algorithmic thinking? Those are the things I was trying to know.
Satish: Sure, so there are, I mean the lot what you said, I’ll just try to bucketize the answer. So, one is the data required? Yes, right? But the context and the relevance and the timing are the keys, right? Because when you, let’s say you are on day zero, you may not have all the data required because your funnel is small, you don’t have a product usage, you don’t know how do you use that, right?
So, there is timing. There is a context. There is a maturity of the data is needed. But let me give the stage. The first is at the acquisition level and I just mentioned data science can certainly be used. How are we typically acquiring data over the top of the funnel? So, where is this concept called product qualified lead is important.
The second level is the lead qualification where Glance, you are solving a very interesting problem where how can I use different metadata right in terms of the product usage, in terms of the how old he’s using this platform. You know the location of this guy, the industries from all these different disconnected, disparate data, right? And how do I make sense of that using data science to really qualify this person so that my marketing folks can really put more effort into that right?
Another aspect is cohort identification, right? The lead identification is at the individual level but look at the cohort identification. Can I find out a bunch of buckets of all those customers or the free users? Or the tribe users who have signed up? And these cohorts can be based on, as I said, location or you know the industry or the functionality or the acquisition channels where they come from and all those buckets, I can track the behavior right? So, do we have the data to create those buckets? That is one data science aspect of it, and once I have created that bucket, can I or do I have the ability to track the pattern or the behavior right? How long has it been there on the differentiations, whether he’s really bouncing off from my platform? Is he really using that platform? And once I’ve created those cohorts then I know the behavior of the cohort. Then I can add all those propensity models, right?
So, is this cohort good enough so that by and large 80%, 90% of those cohort users are ready for upsell? Are they ready for, from free trial to paid trial, right? Another aspect, so one is the acquisition, then the lead identification. I just mentioned the cohort. And by the way, Glance is doing a fantastic job even during that cohort, when I see your platform.
The fourth is the churn. One aspect is we have to reduce the churn. So, once I know or I know he’s a paid user or he’s a part of a cohort, which is a paid user; can I use data science to really predict that he’s going to churn or not, or whether he’s going to renew or not? Or whether he’s going to have any upsell possibility or not? Then I can do some sort of a recommendation based, you know, a treatment on them. Treatment can be, giving them a bunch of offers, lending them more support so that I can reduce that churn, right? So that’s so these are three-four examples I gave you. Sorry, it was the wrong answer but.
Rohit: No, I think there’s no wrong answer. I think you took us through the entire cycle of where all the data sciences so contextually applying. But here’s the thought, Satish. Now, what happens? Do we have to build models? Look, I also want to talk a little bit about the technology that supports all of this. But let’s hit your lead qualification part first and then let’s talk about the graph databases and stuff like that, right?
What ends up happening is in the ICP qualification. Sometimes the ICP qualification becomes a challenge. Can data science help us? The second thing is, let’s talk a little bit about the first outreach. This is where your relationship intelligence comes in, right? Where sales guy needs to pick his Rolodex to call some people, but he ends up hitting all of his thousand people Rolodex and then now he has to build an inside sales team or an outbound team and then kind of call people. So, talk a little bit about the ICP plus where a relationship management tool like you guys will talk more about.
Satish: Sure, so one of the key aspects in the SaaS world is qualify, qualify, qualify, right? Everyone says that the better qualification, the better outcome in terms of the acquisition, in terms of conversion from free to the paid and upsell, and the retention and the referral, right?
So, one of the challenges of the ICPs is during the sign-up process or during the acquisition process. I don’t get data. When I say ‘I’ it is the sense of SaaS companies. So typically, we sign up with a quick name and an email and that’s all right? So, for the SaaS company, because we also have to maintain our user experience. So, the challenge of maintaining the user experience is I’m not able to get that lot of ICP data. So that’s where like you people like Glance or the company like Glance can enrich that my prospects or sorry the user sign up.
So, if we can get more data about the user who is adding a very little bit of information during signup, his profile, his role, his age, his location, his industry, you know what product he is making, what his different aspect of it is. If I am able to enrich along with my sign-up process, then I’m able to really qualify that person very well. Now the challenge of the early very early SaaS companies and I’m sure you found the same problem. Some of your customers might complain, hey, you know we don’t have a lot of data. How do I qualify? That’s where you know a Glance platform can add that enrichment or we can also do a transfer learning. Sorry for the technical term, but basically you can really leverage what other companies with similar cohorts have done in the other companies, in the other SaaS companies, in the other ICPs and we can really leverage that knowledge from the other company’s SaaS ICP to our company SaaS ICP. This is how we can enrich the user’s data or we can leverage the knowledge from the other SaaS companies who have similar ICPs like mine. Now, that’s the one aspect of it.
The other is that typically when the customer journey starts in SaaS, the first problem is discoverability, right? And I need to buy an email list and the two ways in which we can build the email list, typically. I purchased the license of this GTM Platform, purchase this data, or I will organically generate the audience, using a very nice landing page and the creatives. But they both have their limitations in terms of time.
So, what do we do at Mitibase is because you have emails of thousands of people which you have already been talking to using your email and calendar. So, if I look at your C-level folks and the investors and the VPs and the marketers, if I look at their email, they already have been talking to, you know, thousands of people on a day-to-day basis. Unfortunately, that data is not visible.
So, what if I can convert their Rolodexes as a list, right? As a marketing list and then I can apply my ICP filter then this list is becoming a very powerful marketing campaign list and you know why, Rohit? Because all these people are coming from your email and calendar, so there is a recall value right. There is a familiarity principle, so your click-through rate, your open rate, and everything you know go very, very high. So that’s one aspect of how you can use relationship intelligence.
Rohit: Super. You talk about the recall value, how Rolodex access can be mined with some context, and all these things. So how do we go into models? I mean, it’s an expensive thing, right? I can’t keep on building models. The cost of technology or data science is going to be expensive, right?
Satish: So, when you look at data science, there are a couple of myths people have. So, when I look at data science, data science doesn’t have to be always a model-driven right. See at the end of the day with what we’re trying to do is, as a SaaS world, I need to qualify, I need to make sure that I am talking to the right folks who have the right fit. Now at the bottom of that is the pattern recognition. Now, the pattern can be heuristic, to begin with, because if I’m in an early-stage SaaS company I may not have a lot of data. I don’t need a very massive complex neural network model. At the end of the day what I need is a pattern. If I am able to figure out that pattern, even with the rudimentary models, simpler models. And that’s why I love the Occam razor mental models. Whether the simplest solution is the best solution, right?
So, where can you start with the simple models which are less expensive, which you can build with less data, and where you don’t have to have a lot of bandwidth or very complex technology capability required to build that model, right? So, you can start with the simple model, heuristic model, and sometimes I love hybrid models where I don’t need everything automated on day one, right? So that’s the way I tell other SaaS companies whatever data you have if we can figure it out some pattern even with the heuristic method that can do start or day zero or day one model. And then as you break your progress as you get more data as your customer base increases, your data size increases. Of course, then you can add more complex models and do the heavier lifting, but it doesn’t need to be on day zero. That’s what I’m trying to say.
Rohit: Got you. I will invite a good friend of mine and a colleague Sanjay Malhotra here. He had a couple of questions Sanjay please introduce yourself a little bit about Tech Smarter.
Sanjay: Right, OK yeah, sorry. You know technology. Can you guys hear me OK now?
Rohit: Sounds good.
Sanjay: Yeah OK. All right, yes so. Anyway, first of all, thank you both. This is great stuff, great stuff. So, I’m listening to this on a holistic level, right? And getting my background. So, I’ve been in sales and business development in software and services for about 20 years. And so, when I look at what you guys are talking about and selling signals and buying signals and everything else. The questions that bubble the mind to me first and I’ll sort of parse this into a three-part question. The first part is, how do you actually figure out if you’re picking up the right signals. What do I mean by that is there are a lot of false positives that you can get, right. It’s one thing to see in an email thread participation and everything else. But what I’ve come to learn right in my experience, is what looks like great from the get-go isn’t necessarily great from the get-go, and the flip of it is some things that seem like man, it’s not going to work in end up being this most solid deals. So how do we deal with it? I won’t call it a signal-to-noise ratio, but something around that, how do you get knowledge about these signals seems to be more positive. Is there a ratio? Is there a way of looking at it and is there an algorithm that gives you more confidence about this set of dynamics and data, versus another set?
Satish: Go ahead with it, Rohit.
Rohit: That’s a very interesting question he posed, so I was going to say how do you separate the noise from the signals.
Satish: So, Sanjay you asked a very nice question. I think this is really a nirvana question for the sales and business development guy. So, the way I looked at it is based on the way I have been talking to customers is two things. So first is that we also have to not look at, only the digital signals all the time because digital signals have their own limitation. Because they are user-driven. Sometimes they are not updated. So typically, what we get signals is typically measured on LinkedIn sales navigator or from anywhere else they are typically user-driven.
So that is the one aspect of it. So, one way to avoid or improve the ratio of signal to noise is to look at the connectedness. You don’t look at those signals in isolation, right. So, and that’s why in data science, the big aspect of data science, which has been completely ignored in the last 20 years, is connectedness.
Everybody looked at patterns and trends, you know, and the behavior. But can I make sense of 20 things that this impacts and that have a relationship with this? So, one aspect to improve is you look at and that’s what the graph analytics and the graph databases play a lot of roles is can I look at signals and the noise in the data in a more connectedness fashion. And then you see the impact of each other and you can figure it out. Hey, you know what? This signal doesn’t make sense because it has less connectedness with other folks or other concepts other that’s one aspect of it.
The second what we add is a relationship. And I’m talking about the relationship which is not digital is not you know that I’m connected to Sanjay Malhotra on LinkedIn, right? So, where I have the real relationship which you can derive from email, calendar, LinkedIn, CRM, publicly available information, the interactions, so when I add this layer on top of each other, the connectedness, let’s a strong recommendation. So, if Rohit recommends something to you Sanjay, there is less noise, I believe. So, there’s a high likelihood that there is a definite signal and less noise. So, the connectedness and on top of that or real relationship intelligence plays out or removes a lot of the noise and really gives you a real, valuable step in front of you, but that’s the way I looked at it.
Sanjay: Yeah no. Absolutely good. Great answer, great answer. Very right because of all right. So, I know we’re coming down like a couple of minutes left. A couple of things. First of all. I would love to and Rohit can facilitate this. I would love to get a live demo of this and see how it works.
One of the things. That you were talking about so. And this is key right? So, you take connectedness, the connectedness has degrees of strength. And those degrees of strength are hey, what is your personal relationship? In the real world, we talk about personal relationships, and that’s the degree of connectedness and that becomes much more powerful, right? If you’re trying to do this algorithmically, you’re looking at saying OK because Rohit and I have known each other for 10 plus years, because Rohit and I interact frequently because Rohit and I have XYZ. Those are algorithmically like other algorithmic aspects. Therefore, if Rohit’s buying signal or Rohit’s contributable aspect is more powerful than the one that you and I have right now because we’ve just met.
Satish: Right, that’s absolutely correct. So, relevancy, relationship strength, and also who? For example, if I see a recommendation or a relationship that was hidden from my investor’s Rolodex right. So, it has a lot of contexts. Also, context, recency, relevancy, and strength, play a big role, I agree.
Sanjay: Sure, sure, yeah it makes absolute sense that. And those are the soft things that are a little bit more difficult to figure out. Well, actually I shouldn’t say that. Let me back up, those are data points, right? That makes a lot of sense that actually going to inform the algorithm and then there are soft data points, right? What type of relationship is this and those things are not things that we can necessarily quantify.
Sanjay: Do you have a prototype or a live demo that you can share with us?
Rohit: Yes, and we can talk.