189 - Behavior Patterns That Predict Conversion - with Keith Zubchevich

SPEAKER_01: Welcome to Product Led Growth Leaders. We've got a great guest for the show today, Keith Zubcevic. He has spent decades building and transforming technology companies with executive roles at Cisco, Riverbed Technology. He's had about six companies of his own and has been at Conviva for 15 years, rising from strategy to the top seat.

He is today president and CEO of Conviva, globally leading operational analytics that are experience-centric, helping the world's largest digital businesses understand what their users are actually experiencing in real time. Keith, welcome to the show. SPEAKER_00: Thank you. I appreciate your having me.

This is uh this is gonna be a lot of fun. By the way, I've been part of very uh uh startups, many startups. SPEAKER_01: This is uh many startups, okay. SPEAKER_00: So not my own, but but joined my my dad used to tell me either be valedictorian or be best friends with one.

So I chose the latter. So I I've I've been fortunate to find a lot of really smart people and be part of companies that were starting out and turned into massive successes. So it's been a great experience. SPEAKER_01: Yeah, I think you're like me a lot of the times, uh in the past 10-15 years, joined a lot of startups, and you get to be a part of a great team, small team.

Everyone's wearing a lot of different hats. SPEAKER_00: Yes, yes, yeah, uh free revenue, right? They're trying to figure out how to build a business and watching it scale. So the good and fortunate I've I've never uh been a part of uh of a failure.

So every company that I was fortunate enough to join had a great ending, great success, great exit. Um, and so my my hope is to continue that streak. SPEAKER_01: Right. Well, let's talk about the problem that you're solving today.

Let's let's steep ourselves and the audience into thinking about the problem that people run into, that the world uh runs into, that we'll be talking about the solution for today. SPEAKER_00: The main thing is, and in digital business, I think it's something that we bring to the forefront that is really not very visible to digital businesses today, and that's consumer behaviors, right? Everybody looks at how people buy in digital products today, and it's really sequential funny funnels or predictable journeys that you try and set and then see how many people followed that journey.

If you think about how personal people's decisions are to buy anything, you know, whether that's a single transaction or very complex high high-end purchases, all of us have a very specific pattern that we go through to buy. And it's very, it's very personal. It's it's unique to us. You and I might buy the exact same product in two very different ways.

So a digital business building a single product or a single way to see the world is missing either one of us or even both of us. And so Conviva's mission and our job is to really understand what is the consumer experience and behaviors, not funnels, predictive. It is really about collecting every session and organizing that into a behavior pattern and assigning things like segment or archetypes so you can group people into some very similar behavior patterns. But that really surfaces an entirely new way to see your consumers, not funnels because you lose actually a lot, but really understanding how people want to buy.

And if you can meet them where they are and you you you treat them and you you cater to how they want to buy, they'll buy. That is a very similar concept in brick and mortar business today that doesn't get replicated in digital businesses enough. So we want to lead that charge of understanding patterns and how to personalize the digital business to meet every consumer where they are and how they prefer to purchase to improve conversion. SPEAKER_01: Right.

No, so Keith, I'm very excited to have this conversation because as a UX person who does research and design, and we're constantly trying to help clients figure out just the best way to put things together, right? And there's the sort of human side where you're, you know, kind of doing the research and you're understanding people. Then then there's a more sort of automated analytics side where you are gathering the exact data. This is how long this person spent on this page.

This is how uh uh percentage of people click on this thing versus that thing. And then here's how the funnel goes down over time and the percentages. And it's so hard to bridge that gap. There are it's so jam-ped.

It's easy to take make a uh, you know, sort of this, you know, big conclusion of like, oh, well, you know, you know, this button, and if more people click the button, you know, that then make the button three times the size that we'll get more clients, right? It's so easy to make those sort of generalizations. So so let's walk, let's take our time walking through the the types of things that we see. SPEAKER_00: Yeah.

So I think the biggest difference between what you just described, which is what people are doing, right? Because that's very easy to collect is you know how many people did this, how many people clicked that button. But the real intelligence comes from not only what they did, but why. And I'll give you an I'll give you a differentiation between the two.

To understand somebody went to the review review page is is a what they did. Oh, they clicked on this and they read this review. But if you understand and you back out and you understand that they went all the way to shopping cart and then went back to the review page and then got to shopping cart and went back to review page, now you're seeing what they're doing for sure, but it exposes a different picture. We call that a research shopper, right?

It's not just that they went to the review page, it's that you have to understand the totality of the behavior to understand, okay, this is how this person wants to buy. And we will label that automatically by servicing that pattern and saying this is the pattern of a research shopper. You don't get that by just saying, well, they clicked the review page. Everybody can click a review page.

It doesn't mean we're all research shoppers. It's the overarching behavior pattern of a loop, how many times they clicked it, where they went to the review page. So, for example, they get all the way to shopping cart and they go back to review, that signals hesitancy, right? Now all of a sudden, review page has multiple meanings of what of what they did that is unlocked through understanding why they're doing that.

And when you understand why someone's doing it, again, you can cater to them by meeting them where they are and helping them feel comfortable in very various ways, whether that's a marketing campaign or if it's an agent or you're building a product that has the ability to give them easy access to what they want to find so that they build their confidence up, right? Those are the decisions you make off of behaviors, not just people clicking buttons. People clicking buttons tells you nothing.

SPEAKER_01: Yeah. Yeah. Okay. So now we're getting to a really, really core uh challenge.

When you're in the position of you're trying to optimize, make an experience better, um, so you can help the business, and you're looking at the data. So, first of all, there's this sort of problem of detail and volume with the data. You get a ton of data, it's a massive amount, especially the more you're collecting. People say big data.

Well, the more you're collecting, the more problematic it is to try to interpret this. So it sounds like you're saying you're boiling it up to another more meaningful layer of abstraction. SPEAKER_00: Yes, that's a great point. Absolutely.

We we collect every event within the session and we don't just dump it into a database that requires you to ask. So the inefficiency of today versus what Canviva is trying to drive in the market is we have a saying that today you're overinvested for marginal gain, which means you have a ton of data, you have a ton of tools, you have a ton of people, you have a data science, and you're asking a lot of questions, but you're asking questions that are coming from your own head, and you have to understand what you need to ask.

Canviva's position is you start from the bottom up. We collect all of these behavior patterns, we compute them, we create them, and we put them into an analytics engine that in real time surfaces trends in behavior patterns. Because the data should speak for itself. You shouldn't have to ask questions.

And that's the biggest methodology shift that we have today versus traditional product methodologies, is we don't expect you to see a metric change and then you got to go figure out what happened. We surface data that actually says this is a behavior pattern, this happened to interrupt that behavior pattern, by the way. And if you fix this or you do this, you will move these people to conversion. Then we will have a dollar amount, we'll have everything there.

But that's a bottoms-up method methodology versus your point, which is all this data, and I just try and come from the top and ask a bunch of questions and hire a bunch of data scientists to try and figure out what I don't know. And it's the oxymoron of I have to ask the question of what I don't know. Now, if I already know why I'm asking the question. So it really is about letting the data speak for itself and understanding the behavior patterns.

And then once you understand what is happening in the behavior pattern, I can act on that because it's showing me what consumers want to do. Not just showing me what they do, but it's showing me what they want to do and what potentially got in the way. SPEAKER_01: So uh is your solution a software solution, a consulting solution, or something in between? SPEAKER_00: It's a both.

So we provide the analytics platform, the data collection. So we collect it, we we sit today in about 8 billion devices. So you're right, this is a big data problem. So my my my founding team, my group of validatorians in this company uh is uh all academics on big data.

My co-founder, Jan Stoika, actually was the founder of Databricks. So that gives you a perspective on Conviva. We're not a product analytics event simple tech tool. We're actually a big data company because to your point, when you collect everybody's session and you compute everybody's behavior pattern with rich metadata, that's a massive big data problem.

And we do do that all in real time. So that's why I said we sit in about 8 billion devices today. Most of your streaming media apps, we measure all of your video experience and we're telling those publishers exactly what you're experiencing and what's in the way of you watching more. Whether that's a bad recommendation or if it's a bad quality issue, or if it's something about the content you just may not like, all of that data is in the sessions that we distill up to the publishers to say, if you make this change, people will watch longer.

It's the same with consumer behavior patterns. We can definitively tell you that if you do these things, it will unlock that consumer moving forward because we know their intent, we know what they wanted to do. And all we're saying is if you can understand what they wanted to do and see what got in the way, remove that and I will get them to conversion versus general metrics and just understanding, okay, people did if I fix that, I'll get more. That's a very loose connection to conversion versus us.

We have a highly, you know, a lot stronger connection to conversion rate because we we actually measure what people want and what they want to do, and then we give that intelligence back to the cons uh to the e-commerce company. SPEAKER_01: Let's give the audience an idea of what it's like to work with Conviva. Are they are they primarily operating within a software that you've diagnosed them and set up and they're primarily in a dashboard, or are they mostly consulting with you with software tools coming along with it?

SPEAKER_00: Yeah. So we do the integration on with the with the business. So we'll integrate into the product website or apps, and then we immediately start auto-collecting. So there is no pre-configuration, there's no labeling, there's no tagging.

We have an auto-collect capability because we do all the compute on the platform. That's where we become a big data company. So we don't have to protect by limiting the number of events. Our mission is to collect everything because we have the DNA and we have the horsepower to compute everything.

So we open it up, we collect everything from the client side, and we start to build and compute these patterns. And we start to then provide the analytics that gives the customer the ability to, it surfaces automatically. Like I said, it's not a question of you have to ask and you have to go figure out what happened. Our dashboards, once we do the integration and we help the customer set up, literally starts identifying high value patterns that if you make changes, will improve conversion.

And we also show, by the way, I have a simple, you know, I grew up in Arkansas, so I have a simple business philosophy of do more of what works, less of what doesn't. We tend to lose that in business. You know, we have we overcomplicate it to keep it simple. Just do more of what works, less of what doesn't.

So we show high value segments, we show high value patterns and what got in the way of those high value, high conversion rate patterns. So again, it's very highly probabilistic that if you make that change, they'll convert. But we also show here's bouncers, here's things that you should ignore. Now, in our world, in the product world, in some cases, bouncers will trigger a product decision because, oh wow, they left, right?

We need to fix that. But it but if it was a low value customer that was never going to buy, you're investing in solving a problem that is never gonna convert versus solving problems that highly probable will convert. That's the difference in our methodology. So we surface both.

We show the bouncers, we show the things you shouldn't spend time on, we show the things that you could spend time on, and if you did, it would probably improve conversion. And then we have a section of behavior patterns that if you do this, you will absolutely convert. These are your core customers, these are the high intent, highly converting, and something got in the way. That that that is the that that's the methodology that gives you know the e-commerce teams the ability to react and prioritize.

SPEAKER_01: Ladies and gentlemen, we are speaking with uh Keith Zubchevich of Conviva, uh Conviva, I'm sorry, conviva.ai, um C-O-N-V-I-V-A.ai. Um and so uh how do you grapple with the problem of correctness, Keith?

This is something that I'm personally very curious about because AI tooling, you know, we build our software just like we did 20 years ago, but now we have this awesome opportunity to streamline certain aspects of what the software does, you know, categorization and some other things behind the scenes, right? And, you know, we want it to be as correct as possible. Now I see some software tools nowadays that jump too fast and too far with the AI, where it's just so assumption ridden and it's this overly confident but good looking sort of stuff versus um some that are more modest with it.

I'm just curious about your approach to you know how you think about the data integrity, where taking all of these little bits of information at the bottom about usage data, and then we're boiling it up um to things that are much more meaningful and usable. So, how do you think about that and approach it? SPEAKER_00: Yeah, no, that's a great point because we're really leaning into agente. So our belief is when you look at the number of behavior patterns we can generate for any digital business, and we're talking thousands, because you think about it, we're all personal in our decision making.

So there's a there's literally probably millions of decision patterns that people will go through. But we we distill those into archetypes or segments automatically, by the way, not predicted segments that you try and jam square pegs into round holes. We literally take all the patterns and we surface them based on the pattern and the and the segment and the dimension into groups. So now I can act on larger numbers.

So I really can look at hey, if I do this, this is the number of people that had this issue. If I do that, I will convert this number. So there's a direct correlation between data and decision. But the reality for us is that we're a technology platform first.

And so this is a great question with agentic because our view is agentic should operate on good data to begin with. And so when we talk about thousands of patterns, see the view for us in agentic is that the agent should be pulling from all of these patterns, not creating them, not trying to organize them, because that's where the data accuracy becomes a problem, is AI is not a necessarily an accurate compute system. It's a it's an aggregation system. So the sorry for French, but shit in, shit out.

If you have bad data, your agent's not gonna make it any better. It's gonna actually take the data that that it that it has access to and make a decision. And it could be totally wrong. So we start with compute the data, we put the data into our platform, and then agents or agentic actually access those patterns.

Most companies, and this is what people need to understand, is you see all of these tech companies move to agentic. If if the data wasn't accurate and it wasn't, it wasn't specific, it wasn't directly correlated to something before, agents not going to solve that. And so everybody sort of hides, you know, I speak to a lot of investors, and they call it AI wash because all of these sort of modest technology companies, and I call them modest because they have very low tech, they have very low number of events, they're sort of general in what they're reporting, that then say, oh, agents are the answer.

Now you'll be able to find that all the things you want to know through deploying an agent front end on this. But it's just front-ending the old decisions that you could have made before. It's not creating new things, it's not like it can organize, structure, and clean the data that then surfaces it to you in a better way. Whatever was collected, computed, and stored before is the limiting factor of an agent.

And so the question that every consumer, every digital business needs to ask is what's your core IP? What does the data look like? Without an agent, could I have gotten that answer without an agent giving it to me? If the answer is no, and I look at your old dashboards and I look at your old capabilities and it wasn't there before, the agent's not going to invent it.

It's not the end-all be-all for fixing bad technology and bad data. So that's why we consider ourselves a big data company that enables agents to be smarter on consumer behavior patterns, not be the end-all be-all of consumer behavior patterns, and it's it does everything for you. That's a that's a flawed premise. And it's unfortunately being propagated by tech companies who were trying to hide bad technology.

SPEAKER_01: Oh, absolutely. And so, as a big data company that's clearly doing something new and important based on this premise of, you know, humans have our nuanced behaviors, and you just can't over-summarize just based on overly simplistic things. It's not going to work. In creating this kind of a product, I think it's you know, probably just very timely.

Um, how big do you see the role of UX for you guys as the creators of the product and user experience? SPEAKER_00: Critical. I I think there's two things. One is personalization of the UX.

I think, you know, again, we just talked about if everything is in a monolithic level and everybody else's and every of your behavior patterns are personal, then you have a you have a conflict. But I think you know, people are looking to agentic webs, to agentic applications that are personalizing the website, personalizing the UI to a person. Provide those behavior patterns to that agent and personalize the UI. That's a great outcome, by the way.

That's a great outcome. The second wave is deploying agents in UI, right? So you have some level of of click-through data, you have some level of of collection, and then it flips to an agent to say, hey, I collected all this. I see that you did all these things.

Now let me help you take this to purchase or let me help you add on to this. Let me answer some questions if you have questions. So in some cases, it would be a hybrid. And then long term, the UI becomes an agentic interface, right?

And so this it's sort of a migratory approach to getting to personalization and high levels of consumer experience. SPEAKER_01: Yeah, and part of the reason I ask that is a lot of folks don't realize how much of just the experience the UX is embedded in everything, even the LLMs that we use every day, the fact that you can talk to it like a human and you don't have to learn any kind of syntax. And you know, let's say you're working on a document or a deck and it pops up on the side, and now you can read it and talk to your LLM at the same time, that's all a user experience.

And that gap is closing rapidly. SPEAKER_00: And context is key there because we we we talk about context in two forms. In the example of you have click-through data that then flips to an agent, for example, we see in our customers, if you don't have the collection of the web and application behavior data that you then give to the agent as the person flips to the agent, the agent starts off with, Hi, how are you? What do you want?

That that's a waste of time because the consumer just went through a ton of things that you should, they think you should already know, but the agent is a separate thing that then starts them cold. We take the context data from the previous before the conversation starts. We provide that to the agent. So when you hit the agent, the agent says, Ah, hi Thomas, I saw that you were looking for this, I saw that you went to the review page, I saw that you got to the purchase point, I saw you saw shipping and then stop.

Is it a shipping cost issue? Now it's an agent that's providing benefit to the consumer. We see in our customers that if you don't provide that context, it's three minutes of wasted resetting of context, which means you're wasting consumers three. I mean, three minutes.

If you and I sat quiet for three minutes, that's a long time. And every business that starts an agent with high and doesn't know what happened before the conversation started is already in the experience hole. Because now I'm frustrated because you don't know what I just did. I just spent this time, and I'm gonna spend three minutes re-explaining myself.

Guess what happens? I'm either leaving then and we see a high abandonment rate right then, or I'm gonna do this because I need to get it done, but I'm never coming back because you don't have a good experience. So it wasn't that the agent wasn't accurate, wasn't that it didn't ask a question, it's that it didn't have the context to make my experience optimized. It didn't know enough for me to understand that you hear me, you see me, and you're here to help me.

You're not just automating a process. SPEAKER_01: Yeah, yeah. And so, and returning to something you said earlier about being a big data company, I want to make sure that I kind of have the um the segmentation right on y'all. It you're not trying to be in the same space as sort of like a full story or pendo or the product data, or or or might that be kind of roadmap, but today you're solving a more surgically precise problem.

SPEAKER_00: Yeah, we're picking up where they leave off, right? All of you know, you look at you look at all the traditional product analytics companies, they they all do sort of funnel creation, you know, predefined uh journeys, and then and then they try and collect, oh, you had some friction here, we think it probably affected this. So there's some high level of of impact within the product and heat maps and session replay. There's a ton of things that you can do with those tools.

We sort of pick up where they left off in understanding how people want to buy, which is a larger, more richer data set. Now you could run them in tandem. So some some of our customers replace those product analytics tools because they sort of realize, well, shit, that's a very low, you know, sort of low, like I said in the start, it's it's high investment, low marginal gains, um, you know, versus man, let's invest in consumer behavior patterns because it exposes what we call precise decisions, right?

I'm making decisions that I can actually move people through their behavior pattern, but it also sets the stage for agentic. So the difference between us and all product analytics is we have a we have a significant improvement in conversion because we have a direct connection to how people want to buy and you could solve how they want to buy and personalize their experience, but it is the foundation for agentic into the future. Funnels and agents will never cross, right? Because again, an agent is a conversation, it's not a prescriptive sequence of things.

Right. I could ask a question, you could ask a question, we go two totally different ways. The agent has to be agile enough to respond to what we want and be able to move us through how we want to buy versus you know, sequence. The agent's like, well, I'm gonna do this and then this and then this and then this.

Right. That doesn't work in agentics. So the patterns are are beneficial today and set the foundation for the future. SPEAKER_01: So you guys are going in to try to really do a good job at that piece first, and then whoever knows, you know, where you evolve from there, you know.

And I really love products that really try to, you know, think of themselves as part of an ecosystem where you are one specific, you know, plant or animal in that ecosystem that does a really good job at that. And then you can just sort of link with the larger structure. SPEAKER_00: Um that's our hedgehog concept is simple. It's it's understanding consumer behavior patterns to personalize a consumer's experience to improve conversion, right?

Yes. That's that is a that's a very specific methodology. We're very good at building consumer scale behavior patterns and reporting on those and trending on those in real time. And then again, providing that to an agent so that you can improve personalization to a what we call hyper-personalization, almost a personal.

SPEAKER_01: How do people get started? This uh conviva.ai, C O N V I V A. Do they start by talking to you guys or downloading something?

SPEAKER_00: No, they they they come to the website, they go to the website. You know, we we do free POCs. So if people want to because that's where data comes to life. So I'll say the difference between us and all product analytics companies, the the traditional product analytics is you know, once you see a funnel and product analytics, and then you see consume the behavior patterns, you and you literally see all of how your consumers want to buy, it's an it's an unbelievable experience.

So we do a free POC to actually show digital businesses here's your consumer behavior patterns. And by the way, the average behavior pattern of a larger e-commerce company is over 50 steps. Put that in perspective. People are building six and seven-step funnels, but the average consumer behavior pattern is over 50.

We see some consumer behavior patterns in our large e-com companies of 150 steps, right? This behavior patterns are not simple sequential steps. We want it to be, right? Because that's how we organize our thoughts, and that's how product organizes their thoughts.

But if you really think how we buy, we'll have you know multiple steps we go through to buy anything, and that all gets lost in a funnel. So when we talk about bringing the data to life, really exposing how people want to buy your product is an incredibly enlightening step. And we love it because the light bulb comes on, and then all of a sudden they're like, wow, if I did this, I'd move that. If I made a marketing decision here, I'd I'd convert much, much faster.

So there's a ton of things that unlock and become immediately visible when you really see how people want to buy your product. Not what they buy, but how they buy. SPEAKER_01: Wow, super, super interesting stuff and exciting things. And you guys are illuminating the nuances and details of that world.

Um, very exciting. Everyone, please check it out. Uh conviva.ai.

Um, anything else, Keith? SPEAKER_00: No, I appreciate your time. I I love this topic. We're on the forefront of an amazing revolution with AgenTic, and we're just getting started.

So it's a fun time to be in product. It's a fun time to be in tech. SPEAKER_01: Oh, heck yeah. Everybody check it out.

Keith, thanks for being with us today. Thanks for having me.