This week on The Data Stack Show, Eric and Kostas chat with Spenser Skates, Co-Founder and CEO at Amplitude Analytics. During the episode, Spenser shares the journey of Amplitude to IPO. The conversation also covers data infrastructure, product analytics, how cloud data warehousing has impacted the space, and more.
Highlights from this week’s conversation include:
The Data Stack Show is a weekly podcast powered by RudderStack, the CDP for developers. Each week we’ll talk to data engineers, analysts, and data scientists about their experience around building and maintaining data infrastructure, delivering data and data products, and driving better outcomes across their businesses with data.
RudderStack helps businesses make the most out of their customer data while ensuring data privacy and security. To learn more about RudderStack visit rudderstack.com.
Eric Dodds 00:03
Welcome to The Data Stack Show. Each week we explore the world of data by talking to the people shaping its future. You’ll learn about new data technology and trends and how data teams and processes are run at top companies. The Data Stack Show is brought to you by RudderStack, the CDP for developers. You can learn more at RudderStack.com. Welcome to The Data Stack Show, we have an exciting episode. We’re going to talk with Spenser Skates, the co-founder and CEO of Amplitude, arguably the most successful sort of saas product analytics company. You know, having that competitive set, they went public in late 2021. And it was sort of a decade-long run for them and Kostas. I have so many questions. I think one of the big questions that I have is usually about, like, sort of data infrastructure, you know, you and I work at companies that, you know, provide sort of, let’s say, like more core data infrastructure under the hood. Amplitude has a, you know, the bread and butter of the product is an interface, yet there’s a huge amount of infrastructure underneath it, that is a product in its own right. And that we know from personal experience is pretty, pretty stinkin hard to build and maintain. And so I want to ask Spenser about that, right? Because ultimately, the aha moment that I think they want their users to have hides, like, how much is actually happening under the hood? You know, and I imagine that’s an interesting balance, you know, when you’re building a company, so if we can get to it, that’s my burning question. How about you?
Kostas Pardalis 01:41
Oh, I have plenty of burning questions. But I think it’s the first time that we have a founder here who has gone from starting a company to IPO. So I definitely want to share that experience with him from being like, the person who writes the code, the CEO to the founder, someone who successfully has taken like a company public, right. So yeah, I’ll say the like, of course, that we are doing, like you have, like more of like product and technical conversation with him. But like, it’s important to seize the opportunity here and also learn a little bit more about the journey that he has, because it’s a pretty unique and rare one. So that’s definitely something I plan to read more about.
Eric Dodds 02:34
Absolutely. Well, let’s dig in and find out all about it with Spenser. Spenser, welcome to The Data Stack Show. privileged to have you on and so excited to chat. Eric
Spenser Skates 02:47
Kostas fantastic to meet you both really excited to be here. I think it’s an incredibly exciting time in the data space. Generally. We just love being part of it here at amplitude. And so yeah, really excited to talk about perspectives on where we think it’s going and everything else. Absolutely will so much to discuss based on our quick touch base before the show. But first, give us your background. So how, like what led to you, the CO founding the amplitude? Yeah, so I graduated from MIT in 2010. And was convinced that building a technology company was the best way to have an impact on the world and build a career and a whole bunch of other things. And so ended up starting a few unsuccessful companies. One of them, Sona light, was a voice recognition company that allowed you to talk to your phone, very similar to Siri, and send and receive text messages. And one of the things that happened when we were building it was that we wanted to know how customers were using that product, what they liked and didn’t like and what led to long term engagement on the product. And it was obvious to us that you should look at customer data in order to make that determination. It was interesting at the time, we looked at what was out there on the market. And every analytics tool was very marketing centric, and marketing focus was all about, you know what traffic sources are, leading to people to jump onto your web page or where they’re coming from. And very little of it or none of it allows you to stand the journey from a product standpoint, we said, hey, we need to go build this. So we ended up building something in the house. And what’s interesting is that it ended up being much more successful than the product itself. Lots of other companies, we show that to be like holy cow, I want the same exact insights for my product. At the same time, Facebook came out with a very famous study that they found the best predictor of long term engagement on the platform was how many friends you added. So if you added seven or more friends in the first 10 days, you have an 85% chance of sticking around at least two months. And if you didn’t get to that threshold, the chance was less than half that so that was the single biggest determinant of your long term success as a Facebook user and was seen as nothing on the market could answer that same sort of question at the time. So we said, we need to go build this. Yeah. And so that was the start of amplitude in 2012. You know, I started it with my two co-founders, Curtis and Jeffrey, who launched the company in 2014. We’ve been growing like crazy, since we took it to the public markets in 2021. And then continuing to grow today, we work with a lot of the best companies and most forward thinking product lead companies in tech, whether that be Atlassian, Intuit, DoorDash, PayPal square, you know, you look at the list, and we work with them. And so now, but it’s still very early days in the category. And so the question we’re trying to answer and figure out is, how does this whole thing play out over the next decade? And what does the future look like? Sure.
Eric Dodds 05:45
Okay, I have a question. So, when you’re at Sona, Lytx, and you sort of built this, like MVP, product Analytics, you know, product in a sense to, like, try to answer your own questions like, just curiosity, like, how did you build that? Were you just like, you know, lay some sort of interface on top of your production database? Or like, what was the like, sort of MVP that people were like, Wow, that’s awesome. Yeah, so
Spenser Skates 06:11
We tracked a bunch of data through it, believe it or not, we actually use SQLite. That’s a database, that many actors, so it’s the simplest to just set up and start tracking data for sure. And we started to answer all these questions, right, we wanted to know that the key question that we had was, how much does an improvement in the voice accuracy reckon the accuracy of voice recognition lead to more long term engagement turns out massively impactful? It’s like, I think, a 1% increase in voice recognition accuracy would translate to a 1% increase in long term retention. So improvement, very important. And what’s interesting, though, is we’re fighting SQL in order to do this SQL, yeah, really poor at answering what I think of as time series or sequential questions, because you have to do joins. And you’re kind of doing this like, it’s like, you have multiple levels of nesting for every single additional join you want to do. And so about a few minutes to hack it together with this thing. And we’ve looked at our retention curve, and looked at how the accuracy of whether you are successful on your first match changes that is very, very impactful, as we found out and we, we showed a kind of a few charts on this to other people, like, Oh, holy cow, we want this exact same sort of thing, because it’s the exact thing we want it, sir. And so when it came time to actually build it for amplitude, we, you know, we ended up having to rebuild the whole thing. And it was more generalized and interactive works for any use case. And they said questions, and all of that sort of stuff. But it really started by seeing the power of those insights and seeing other people really wanted those exact same things from a product lens.
Eric Dodds 07:51
Yeah, that’s super interesting. It is. Yeah, I mean, I think it says a lot about your motivation to answer those questions. Because that’s some pretty gutsy sequel, and like a lot of work on what is probably ultimately, like, a fairly brittle system that has all these insights. But that’s some commitment
Spenser Skates 08:07
might actually just be a side note, my challenge for anyone listening is if you think SQL can be the long term answer on this stuff, just try building a basic funnel out of SQL. Just try doing that. Let me know how it goes. Yeah, you know, and then get that. But yeah, SQL is great for a whole bunch of things of transactional data. But when it comes to looking, mapping out user journeys, there’s a whole bunch where it falls down pretty quick. Yeah, it gets pretty complex. Okay. i We have so many topics to get to, but So, congrats on the amplitude IPO,
Eric Dodds 08:42
you know, decade of, you know, passionate work. I’m interested to know, as you reflect on that experience, you know, what would you tell the Spenser Skates of 2012? You know, when he first started the company, like, you know, looking a decade back, what would you tell that version of Spenser skates sort of knowing what you know, now,
Spenser Skates 09:07
What’s interesting, wouldn’t be that much different for sure. We could have saved some time. In terms of moving on, we probably could have moved on from Sonlight faster than we could have. One of the things in the early days that I didn’t do well was asking for money, and ended up wasting a lot of time talking to people, customers who would or use it for free, but not actually be interested in paying or not valued enough to pay us. We probably wasted about a year on that before I actually started before I felt confident enough in our product to ask for money, I would have done it maybe three or six months into that would have been the change. But other than that, you know, I wouldn’t have changed that much in the really early days. I think one of the big ones is that I figured out that I should go ask for money for the product. We ended up accelerating like crazy. So we launched in 2014. And right away, I would basically not spend time on a bunch of customers. As who didn’t want to pay anyway, and you know, you’re arguing about $50 a month with two customers who are willing to pay you 1000s of dollars a month for the product. And it was like that was such a great forcing function to get us to spend time on problems that mattered when customers that mattered and accelerated, really focused and accelerated our development tremendously. And so, that was a big kind of moment in 2014, where I switched full time from building and being the product person to being the first salesperson. So that was a really big change. And then, you know, after that, I think we hired executive team, we kind of grew it out, I think, I definitely would have pushed myself to do that earlier and figure it out earlier, hiring executives, there’s so much knowledge about organization building, and how to build how to do that, well, that’s out there that you don’t have as a first if you’re a first time founder, CEO.
Kostas Pardalis 10:55
And it’s much
Spenser Skates 10:59
more valuable if you can bring that in from the outside to leverage and then grow the company. Now, that leads to all sorts of other problems, like how do you make sure to maintain the culture and set of values you want? You know, execs come with all sorts of problems, like needing to be highly paid. And I’d sent him as differently and all of that here, it can be, it is a real accelerant there is, versus if you’re trying to figure out how to build an organization yourself. The other comment I’d make is that, you know, at the rate a lot of these early companies are growing, where you’re doubling or tripling every year, that’s a very unnatural thing for an organization to go through. Like if you’re growing maybe 10% Every year, that’s more natural. That’s how most groups of people, you know, self organize, and change over time. But if you’re double, triple, that’s like, you know, no, it’s a different company every three to six months, exactly. New Company. And so you have to get help to set that up for success. So I would have pushed myself to do that earlier. You know, but I think the fact that we did make those changes and adjustments after not too long, first, in terms of me focusing from being an engineer to being a salesperson, incredibly important, and then from hiring outside help, and executives who had done and seen different parts of the journey before, you know, both of those were, those are the biggest thing I’d say, you know, engineering centric founders get wrong is appreciation for salespeople lack of appreciation for what great executives can bring in, you know, they try to reinvent it and do it the wrong way. They’re, like, doing the Google Hey, how we’re not going to have managers, turns out managers actually serve a really important function, someone needs to make the call, who is not on the team? Yep. And, you know, same thing with sales. It’s like, one of the really interesting things to me about sales was that, you know, you think you have this image in your head of the sleazy used car salesman, that’s a lot of your experience. But great salespeople, it’s all about matching up opportunities where there’s a problem and pain that your company can solve in the market, and aligning to organizations on that. Now they come with their own set of, you know, differences in that, like, the way they do work is very different. They do work by customers, you know, they’re very incentivized by cash compensation. So it’s a very, you know, different sort of mentality. But if you can do that, well, then you can build a great business around. I’ve seen too many great engineers that viewed building beautiful products that don’t have enough of an appreciation of the importance of that to the success of the company and ended up failing.
Eric Dodds 13:29
Yeah, yeah, super interesting. Was that hard for you? Like being so focused on the product or less some of that stuff on sort of handoff that responsibility? Or like, did you stay involved for a while, because that’s, uh, you know, especially in the early days, like, you know, that’s a huge, you know, sort of, you know, part of you goes into the product in many ways. Yeah,
Spenser Skates 13:50
I was lucky in that I had two really strong co-founders Curtis and Jeffrey. And what would happen is, we meet weekly to talk through different things on the engineering or product side, and I will push on different areas. It’s like, hey, if we consider doing the segmentation report, like this way, or have we considered processing data this way? And they would say, Oh, we already thought that through, you know, and it’s bad for these reasons. And we came up with a better solution after like five or six times that it’s like, Okay, you guys clearly got it. Like, I’m not gonna have more time on this. Obviously, you know, we, you know, I’d have ways to keep up to speed with what was going on on the product side. And you know, we’d have weekly check ins and sinks but 90% of the time was then like, I’m just going to focus on customers and sales and driving that.
Kostas Pardalis 14:42
Suppose I have a question for you. You mentioned that you went from the beginning, like, going into sales are trollee like you had to do that’s right. Tell us a little bit more about his experience. How you experienced his transition. and similar to us like a few things that you’ve learned that you’ve found valuable. And maybe also surprising, especially for the people or like engineers out there who have never thought of like, you know, going out there and becoming like trying to sell something. So that would be amazing, the hardest part was reorienting,
Spenser Skates 15:25
instead of going deep on a particular problem, because that’s as an engineer, if you go deep, you understand the first principles, you can build up to a model that solves a particular problem. Instead, what you’re really doing is you’re figuring out how to navigate uncertainty in a world. And you’re figuring out where there actually is pain on the other side. And it’s a very, very different mindset. Like, one of the mistakes I see a lot of engineers doing is they’ll go into, and they’ll say, hey, I want to learn this, let me what’s the best book on sales, because that’s how they learned how to be a great programmer, they read it, you know, the types of stuff turns out, that’s actually a terrible way to learn sales, the best way you have to learn sales by going and doing it, going and spending time meeting with customers, meeting them face to face, meeting them on site, getting to know them asking about their problems. The other big thing that was really high leverage for me is I got a coach, this guy, Mitch mirando, who would come once a week, spend time reviewing what I was doing on the sales side, and then give me lots of feedback and coaching. And so for example, one of the questions he always asks me is I talk about a customer. And he’d always ask me, Hey, what’s the pain here? And I’d be like, well, they wanted to make some sort of sequel report. It’s not working. And I’m like, Yeah, but what’s the business pain? Like, that’s not a business pain, Spenser. And so I go back, and you know, ask the customer about that. And then I come back to him, you’d be like, what’s the pain? I’m still not hearing like a business pain. And so after a few cycles of this in every single meeting, I would think, okay, like Mitch is gonna ask me, “What’s the pain?” So let me actually spend some time asking the customer about their business problems and what the pain and pain is. And that helps you get much, much clearer as to whether, because it’s a two sided thing, there’s what your product is. And so many engineers are in love with what their product is, and how great it is, how many features and functionality it has. There it has, but it doesn’t matter unless it can solve what it is a business is looking to do. And so really getting clear and understanding of that, as the kind of golden thread to follow with a customer and from that flows, how you should spend your time which customers are a good fit for what you do, how you pitch it back to them and position what you do, what you end up building on the product and engineering side, and everything else. And so that really you know that that was huge. And so yeah, I think the biggest thing is learning by doing and then getting coaching and feedback from experts on it, that accelerates your learning so much faster than you could do on your own. So many engineers will spend lots of time on problems that don’t matter. And the way you figure out which problems matter is by asking and talking with customers, and the way you figure out how to ask and talk with customers. And the right way to narrow in on the ones that are important is by getting help getting coaching on what it means to be a great salesperson. I can
Kostas Pardalis 18:18
totally relate to that. Because I felt like I had a similar experience. And because like it was myself writing code that they decided to start the company and then suddenly I had to go and sell. And it is like a transformational, let’s say experience. And I would say like one of the things that I still like to recognize a lot when I talk like with ng C’s like, you use like a war, you talk about like navigating uncertainty. And I think that’s like, if someone asked me to describe what’s the difference between the salesperson, the engineer, or like I would say, the ligaments and so uncertainty. A salesperson loves uncertainty. It’s like where did the fried right leg come from? Totally Yeah. And it’s like these transitions that you have to do like, regarding uncertainty, which is like, super. I mean, it can really boost your growth and personal growth outside now of like, you know, being successful in whatever you do, like, personally, I think like it’s an experience for Cavazos, right? That’s
Spenser Skates 19:24
one of the metaphors that I really like is that imagine? There’s 100 customers out there, all of which potentially, you have commerce you could have a conversation with and solve your pain. What you’re trying to do is figure out those 100 black boxes, which are the few that really desperately need what you need the most. And that’s your job as a salesperson, and that’s what the art of sales is all about is trying to figure that.
Kostas Pardalis 19:50
Yeah, yep. 100% 100% So, okay, you let’s, I mean, we can talk about the stuff like for hours I think like it’s it’s all Amazing to have like someone who has been through like so much transformation to be honest, like from writing the code, you know, reaching the point like to IPO. But let’s, let’s talk a little bit more about like the technology, it’s and, and the product. And let’s start with talking about product leaks. And like the difference between product analytics or analytics, error rates, like analytics, are nothing new. Like since we’d have like, computers probably like even before that, like people were using numbers like to try and figure out how to run their business or like do things in zero, the eyes like, like, we’d have, like, since forever. But why do we reach the point in which your opinion that product analytics became like a thing of their own? It’s important enough to justify a company going into an IPO. Right? And so tell us a little bit about that. Because that’s like, also like very, I think, like, super interesting, I
Spenser Skates 20:54
i think what’s happened is that, because of the rise of online, as online used to just be a marketing channel, where you just market to your users, you have a web page, you collect leads, but the actual transaction, now you have full fledged applications across the industry. And this is happening across in every industry, whether you’re looking at b2b, the whole rise of SAS, rather, you look at media companies, now media companies, the main their growth channel, you look at Disney, number one growth channels, Disney plus everything else about their business is actually shrinking. Parks is, you know, Parks is static, you know, the distribution channels for movies are shrinking. It’s like, the digital and online is the future growth, you look at retail, Walmart, you know, they realize that online is the future of their business. So every single industry you go across, it’s like the online product is the growth channel. And so because of the growth channel, that means you want to invest in it and control it. And so how do you do that? Well, you need to be able to track it first and foremost. And that’s where product analytics and product data comes in. I think the other part of it is that technologically, it wasn’t really possible to track how people used a product beforehand, you go back to the days of licensed software, where you buy something and install it locally, you buy some box software, and you install it locally on your machine can actually see how users using it in real time. But because all these application workflows have migrated to the cloud, all of a sudden, it’s possible to see them. And then the last thing I’d say on it is that the whole so this has led to the rise of you know what’s called product led growth movement. So because you see that there are some companies that do it very successfully, companies like Facebook and Netflix and Atlassian, and HubSpot, and square and PayPal, you know, tons and tons of companies that do this really well. And so everyone else is like, dang, I need to figure out how to do this exact same thing using this exact same methodology. And that’s where the need for it arises because they realize that if they don’t figure out how to make their product, a distribution channel for the business and control and make it successful, that they’re going to be left behind by companies that do. And so I think that’s what’s led to the rise of product analytics and product data being a standalone category. And that’s what we are now. I want to be really clear, we’re tiny, our revenue guidance for last year was 230 million. If you look at that, compared to the other big vertical SAS players, you look at what’s the equivalent on the sales side Salesforce, right. So you know, they do sales analytics. And you know, that is a company that does 10s of billions in business every year, you look at the marketing side, Adobe, you know, that’s a company that does, you know, similar order of magnitude of revenue, you know, amplitude, we’re only, you know, 230 million in revenue. And so we’re so early in the space relative to, to where the long term is. And so I think this still has quite a few ways to play out.
Eric Dodds 23:58
I interject a question for both. Spenser, for you and for Kostas. So it was interesting hearing you talk about sort of the distinctions between, you know, analytics and product analytics, but in many ways, like, if you think about, like marketing analytics, and you think about, like, a marketing website as a product, that distinction is really not, in many ways is a real distinction. Yeah, it’s not a real distinction, right? Like, is there a distinction between marketing analytics and product? I mean, obviously, they’re like tools that are like, you know, like Adobe analytics or, you know, just generally used for, like, marketing and attribution. But like, in reality, there’s not, at least from my perspective, like there isn’t a distinction, but what say you
Spenser Skates 24:40
totally, completely agree it’s the same, it’s going to converge the same thing. From a customer, you don’t think Am I on the marketing website? Or am I on the product part of the website, you know, it’s all part of the soundex that’s it, you get an email or push notification it’s all integrated from your standpoint, you’re not thinking am I interfacing with a company’s marketing team or We’re the product team. And so I think long term, the tools will end up converging. And so that’s why you saw amplitude. Last year, we ended up launching a whole ton of features that allow you to do marketing analysis, because people had the same sort of questions they want to view, the integrated journey where you see someone landing, how does someone go from landing on the page to ending up being a great long term customer? And where did those customers come from? And so we ended up developing Elia. Now, I think, historically, product and marketing teams have been quite separate. And so that’s why these two things have emerged separately, because first you had digital marketing and the whole explosion of the marketing tech stack online, because online was a marketing channel initially, and then you have marketing sort of questions that you’re trying to answer. So where are my leads coming from? What’s converting? How much time are they spending on paid? What are my most popular pages and pieces of content? Do you have product quest type questions being answered? Like, what’s How do people get stuck in my onboarding funnel? What’s the long term engagement rate of my customers? What causes customers to come back over time, or what causes them to churn out of my business over time? And so you get product type questions. Now? Ultimately, again, from a customer standpoint, that whole journey looks the exact same, those questions are actually very tightly related. And so I think over the long term, you’re gonna get convergence
Kostas Pardalis 26:23
of those two spaces. Yeah, I mean, I don’t know if you agree. There’s like, huge overlap between the two. They’re also like some, like, maybe a little bit like fundamental differences, like, let’s say, we, like there is a very important I think, difference with product, you always have almost like the test phase behind the data. Right? Do you have like a customer who have
Spenser Skates 26:50
a login? Yeah, you’re generating absence or profile, use some sort of, you know, recurring engagement with this company versus just being anonymous. Yeah.
Kostas Pardalis 27:00
And probably like a bit of a better signal to noise ratio, because like the marketing you also have all these unknown animals. Like draftee lets you have like to work with like, much more noise away and Bobby’s muskie. There are some different questions that you need to answer there. But I think like, fundamentally, in terms of like the technology, or like, even like, the way that you interact with the TikTok, like WYSIWYG UI SOAP, like, too much, like almost too much of his like, you know, like doing, right? So
Spenser Skates 27:35
Normally, a lot of times, you actually want those questions to be related. So you want to acquire a bunch of users, but you don’t care just about Sannomiya. Living paid you care are these long term great customers for our business, spend a lot, because if they aren’t willing to spend more on them, I just don’t care about someone hitting my landing page, I care about how much revenue I’m going to derive over their entire lifetime. And so they’re actually very closely related. Yeah,
Eric Dodds 27:58
I mean, part of the reason I ask is because, like, you sort of being in the world of marketing, and especially working in the world of SAS, like, among my peers, like the people who I really respect to you, and it’s like, okay, you really know how to, like, figure out how to build something for scale, like, are generally just using product analytics to like, look at, you know, acquisition to like retention, essentially. Right? Because like, that’s actually how you make a true, data driven decision. Anyway, sorry. Sorry to interject here, Costas. But, you know, if I had to ask
Kostas Pardalis 28:29
mother doll, that was like a great question. And actually, specifically, if you have like insides of like your guns, like, outside, obviously, like, we’re thinking about product analytics. And I would assume that like the dominant persona using amplitudes, our product managers are like people, but maybe that also like other people that are doing Have you seen, like marketing or like using aid or like other people outside the organization actually using amplitude?
Spenser Skates 28:59
Absolutely. That’s actually one of the biggest learnings we had from last year was that product data is not only useful to product managers, it’s actually useful to lots of people in organization, anyone who touches the product experience also cares about product data. So firstly, just within the product development team, you’re talking about engineers, designers, folks like that, that also need to leverage and use the data. And then when you start to get outside of it, marketers actually use this too, I think we have a good percentage of our customer base, that is marketers and also using amplitude to supplement what they do. See, we’re facing teams, that’s been the most one of the most interesting ones to meet customer facing teams will figure out like a support team wants to know, what did someone do before they ran into a support incident, and they’ll use amplitude for that. Or they’ll use it to figure out which customers are at risk of churn because they’re not really engaged with the product, or what sort of features are the highest paying customers using? And so you start to map it out and it’s like almost every single function across a company needs access to product data. What makes product managers in particular special is that they’re at the top. They’re the ones with these questions first, and they’re at the Tempest. They’re the ones that want to start to figure out, Okay, I just shipped this feature, what sort of impact has it had? Since we launched it? How many people are using it? Is it leading to more engagement and retention?
Kostas Pardalis 30:19
100%. And just because we were talking about sales at the beginning, have you also seen like, especially because of the product lens, kind of growth like Parag and granola soft? Have you also seen salespeople like getting value out of this data? And we’ll do the same by like, accessing, like directing something like copy here, but by I don’t know, like some of these insights going back to Salesforce, and oh,
Spenser Skates 30:45
completely. We use it a bunch here ourselves, it’s actually a huge initiative for this year, is to get everyone on sales and customer success trained on it, so that they can use it. And a lot of the team already uses it, but making sure we’re we’re expecting everyone to do that so that they can find more opportunities and more within their existing customer base. Yeah,
Kostas Pardalis 31:04
yeah. Yeah, it makes a little sense. All right. You mentioned that, like, he started like that to get money. 2012 rights. That’s right. But it’s almost like we live in years now. So there’s a lot of change that happened today. That’s right. And one of the like, the biggest, let’s say, things that have changed is like Cloud Data Warehouse, like people have much more access to, let’s say, tools, that’s traditionally, you would see like the enterprise HVAC system that now does the Cloud Data Warehouse chains, let’s say the dynamics between amplitude as a product and product, mix with the market out there, if you have stomach and like once you’re welcome using happening other.
Spenser Skates 31:57
What’s really interesting is that the rise of cloud data warehouses, if you look at the largest use case for it, is to be the repository for what your customers are doing in your product and all the behavioral data. And so it’s actually quite the rise of cloud data rest is really interesting, because it’s been like kind of a parallel rise, you know, obviously, you have a vertically specific end to end integrated tool, like an amplitude, where we’re specifically targeting product data and all of that, but then you have Cloud Data Warehouse, which allows you to say, Okay, let me put that data and all the other data you have into a single place. And so that has become massive. What’s interesting, and talking to a lot of the data leaders is the number one challenge they have is how do you unlock data out of that data warehouse and get it in the hands of end users and aggregate all this data across your business? It’s a huge mess, like, you know, schema, there’s no consistent schema naming convention. There’s tables all over the place. You know, events aren’t tied to consistently all sorts of issues like that. Question is, how do you then operationalize it and start to leverage it day to day in your business?
Kostas Pardalis 33:02
And how do you see amplitude working together with a data warehouse? How does it happen today, and like, if you can serve a little bit of your vision, like for the future around Google to be like, great, totally.
Spenser Skates 33:15
So I think it’s going to play out similar to the previous generation, in which you have for more complex, more sophisticated analysis that requires data from multiple teams, you’re going to do have an analyst or data scientist, do that analysis customer on top of the cloud data warehouse, and that’s going to be your kind of system of record. In addition to that, you also probably have a tool that end users use day to day to answer the questions. And so they want to know how many people are using this feature today, what’s my daily active user count, and they’re not going to go into a data warehouse and write a sequel to. And that’s where an amplitude comes in. And so just like your sales team uses Salesforce, so your marketing team uses Adobe, you know, your product team is going to use amplitude. In terms of overlap. I think one of the really interesting things you might say it’s like, okay, well, you have the same datasets in both. So don’t you need to work together? Absolutely. The number one integration we have as a company today is with Snowflake, we have hundreds of customers that are both Snowflake users and amplitude users at the same time. And so they’ll often take data and amplitude and then send it to their Snowflake data warehouse, so that they can keep a copy of it there and cross reference it with other pieces of database and data in that database. And so while the product team is still using it, day to day, you end up having, you know, also that data for more sophisticated analysis and your data warehouse. The other thing I’d say with cloud data warehouses and in amplitude is that I think a lot of these data leaders come in with this vision of handling to go build out the stack by itself where I’m going to put all this data warehouse or put like a bi thing on top. And one of the things you realize is that the workflows per function are so customized and specific that generalized BI tools don’t really cut it. So it’s not like if you think of the analogy, the sales side, it’s not like you have sales managers running their forecasts out of a data warehouse that makes no sense not to use Tableau, or Looker to do that they’re gonna use Salesforce. And so I think it’s the same way with product teams. One of the biggest places that we see need for is actually when the central data team gets too overwhelmed with questions to be able to keep up, because there’s a never ending list of questions and things you want to know. And so you need something to self-serve those hands users like that. The reason we were pulled into Atlassian was because the product team, one of their mobile product teams, constantly had all these questions they wanted to get answered by the central team, and they couldn’t get answers to that. And so they ended up deciding to say, hey, we need to buy something that just does it. They used amplitude. Other teams saw what they were doing, so they decided to adopt. And then finally the central data team, they were very forward thinking. And so they were thinking, how can we have an endless backlog of questions? How can we actually self-serve all of these customers and get them to answer things without having to come to us? And then they decided to standardize on amplitude, the funny thing, we call it in the among data leaders, we call it like data, breadlines, it’s like we’re rationing data, because there’s not enough analysts and people who can write SQL, and who understand the data model to go around. And so you end up having to have a constant shortage of data and data insights, when, ideally, the entire company is running off data all the time. I remember talking to one data leader at Airbnb a few years ago, and their aspiration was they were gonna hire a data scientist for every single product manager. And I’m
Kostas Pardalis 36:43
like, yeah, like, you guys are gonna. That is, that’s just the
Spenser Skates 36:49
craziest thing. I mean, and this is Airbnb, and they still weren’t able to do it. And this Airbnb, one of the, you know, one of the most successful Silicon Valley companies in recent history, they still weren’t able to attract enough and pay for enough folks in the data science side, you know, and so they ended up having to go to third party tools. And so I think a lot of people don’t realize what a bottleneck, that thing is, for the rest of their business.
Kostas Pardalis 37:12
100%. All right, one last question from me, and then I’ll give them back to to Eric, got to get started with us, like something is happening in the industry today that you find really exciting, not so only so But together with amplitude like that, something that we feel like, it’s going to be really exciting, how Americans is going to grow around that based on like, the trends that are happening out there, right, like, what’s what excites you? So much?
Spenser Skates 37:48
I think one of the biggest things is, how do you bring this way of leveraging data from what I think of as the sophisticated innovative product that leads growth companies to the rest of the market? I think, still a lot of innovations that go there. Eric, one of the points you made earlier, was that one of the reasons Google Analytics is so successful and ubiquitous is because it has a bunch of charts that work out of the box for marketing analytics. And I think
Kostas Pardalis 38:19
Spenser Skates 38:21
needs to go through that same transformation, it’s still not there yet, use us, if you some are competitors, it’s still like, you have to build a bunch of reports, when you first sign up, there’s not a ton of stuff that works right away. If you don’t already have knowledge of adults, think about it, use the data. And so I think this industry, we’re in the kind of Crossing the Chasm moment where you’re going from the early innovators, the early adopters, to the early majority and early majority, it’s like, hey, you know, don’t really know how to do this, you have to teach me how to do this. And so I still think there’s a long ways to go and doing that, you know, we’re going to be doing a launch on the innovation side this year, where we’re looking at coming out with out of the box reporting and a whole bunch on the UI side, to make it easy, but I’m really curious to see what other innovations come out in this space, because that’s the number one requests we get from customers is, what are the best practices? How do I do this? Yeah. And we’ve playbooks and you know, we do training sessions and all that sort of stuff. But I think there’s going to need to be more breakthroughs on the product side to make all this stuff work out of the box, like you don’t be an engineer, you don’t have to be a data scientist, in order to start to understand how to leverage this in your job day to day. And so I think this industry is going to have to go through that moment. Because obviously, folks like us, we’re very familiar with data data stack, both the self built one, you know, that has the horizontal tools like snowflakes and RudderStack. And you know, other companies, as well as the vertically integrated ones like us,
Kostas Pardalis 39:47
but it’s a there’s, there’s still
Spenser Skates 39:51
a lot more ways to go. One of the other questions I love asking data leaders. I’ll ask them, what percentage are you in the way of realizing your vision? They always say, oh, I’m only 5%. And only 10% of the way there, no one ever said, ever, like, oh, yeah, I’m done building my data stack. And we’re good. And we just need to do this. I need more resources, more people like, yeah, so all this capability built off, that tells me the vast majority of the innovation in space is ahead of us. And so I think the question that we all have to face is like, how can we make it work in a simple enough way for the early majority user, so that they start driving and getting a bunch of value out of this without having to be, you know, learn engineering, or learn SQL or be be as much experts and data as we are? Sure, well, let’s, so I want to use the last part of your time to actually dig into some more technical stuff. And I think this is a great jumping off point. And so I love the vision of, you know, product analytics, sort of becoming as turnkey as say, like a Google, you know, Google Analytics,
Eric Dodds 40:54
although which this is a whole other podcast episode, but we should talk about some time the, like, the challenge of dealing with, you know, 20, year old, you know, methodologies and taxonomies, and the Google ecosystem, because it’s amazing how pervasive that’s become, and how painful it’s become, especially compared with a tool like amplitude. When you think about delivering that level of sort out of the box, you know, sort of like, value, let’s say to like an end user who’s logging the platform. There are certainly UI elements to that. But then there are also probably some, like instrumentation elements, right? Because, you know, ultimately, amplitude sort of like, you know, delivering insights at the end of a pipeline of data, right. So you have events that are coming in, they’re being processed, and then you can build reports, or you can produce reports or whatever. And the initial like instrumentation of that in whatever products you’re tracking, right. So like, let’s say you have a mobile app that has a significant impact, right? That’s actually the raw material that sort of, you know, drives the insights. How do you think about providing that sort of out of the box experience in terms of like, where, in that pipeline? You’re trying to impact that? Is it like super early, you know, to, you know, I guess, like maybe in the auto tagging type sense, where you’re like, trying to, like, control the schema, like very early on so that it’s easier to produce those? Or is it really more of a UI like reporting challenge,
Spenser Skates 42:28
I think the whole thing, the whole pipeline needs to be improved. So first, instrumentation is the number one blocker to someone successfully adopting product analytics today. And so if you look at
Kostas Pardalis 42:42
what how Google Analytics
Spenser Skates 42:43
does a good job of it, they auto tag pages for you. And so I think you’re going to need to do something similar, because right now you end up having to tag each single event at a time. And with hundreds or 1000s of events per charter to get started in a simple way. Now, we have ways to do it, where it’s like, hey, let’s start with the top five events that go from there. I think it still remains the number one block, it’s hard to know, like, very hard, very hard problems. And so there’s a lot I mean, there’s autotrac ways to do that. Those have major downsides, which is why we and there’s gonna be a bunch of no two ways that we’re going to be trying to tackle this year. So we’re just to see how much progress we can make for sure, in the UX, you know, needs to be simple enough that someone non tactile that you can use. And then I think the other big part is, how do you get a reasonable set of dashboards default out of the box, so that someone, okay, this makes sense. I get it
Eric Dodds 43:35
when they’re using it for the first time. And so I think all of those end to end pipeline is going to be to improve to make it successfully adopted by the early majority. Yep. Yeah. Super interesting. Okay, digging into that pipeline, one of the things that’s always fascinated me about amplitude is that you can sort of derive these incredible insights and your I remember my first experience using amplitude, and you know, you sort of like, drill into an event and like, made some sort of discovery on an error that was happening somewhere. I was like, Oh, my goodness, this is like a huge problem in this mobile app. And like, wow, this is awesome. You know, it’s just like, the ability to drill down is super cool. But that kind of obfuscates the amount of actual data, like engineering, pipeline complexity in terms of the build. You find SDKs very difficult, you know, like to actually build a really robust SDK, you know, fault tolerance, like there are all these sorts of things. And so how could you think about that, you know, because really, like, you’re delivering an insight to an end user in the UI. But the product that’s generating that is actually like, phenomenally complicated, like data pipeline software in and of itself, right, like it’s almost like it’s a completely separate product.
Spenser Skates 44:59
Yeah, Now I’ll tell you it’s even more complicated than that. Because if you look at product data, first, the surface area is enormous. An average product has 1000 different touch points. So now it is different, you can’t even hold that stuff in your head. And so how are you going to navigate and know what’s going on there in the right way? So that’s a huge challenge already. And so how do you even categorize make sense of all of that? And then on top of that, it’s like, every single user navigates those 1000 data points in a unique way to play and users, you have a million different paths. And so how the hell are you going to synthesize and analyze that in a coherent and meaningful way? And so I think there’s really hard problems around both of those issues. Now, to your point, Eric, I think one of the things in the product side is we think a lot about how do you make it easy, because there’s, you have to abstract away and make reasonable guesses on and how to present that complexity because the vast majority of people like that’s where 99% of the value lies, because you don’t you can’t have the average person, your org needed to be trained on that complexity. It’s not going to work. Yeah. But you can get value out of product data, absolutely. Whether there’s someone on the support side, trying to figure out what went wrong, whether there’s someone in marketing training, or what campaigns are working, whether it’s someone in design, trying to figure out, hey, what’s broken about this interaction, like there’s so many people in the company that need to leverage this data? And so and so the hard part from a product standpoint is how do you abstract away that complexity. So like, one of the things that we did, that I’m very proud of in the early days was, we actually made charts interactive. So you could click on a data point and see the list of users that made that data point and then what they were doing before that data point. And that’s been an incredibly valuable tool, because it allows you to kind of break down and drill down on a specific data point or user or set of users in a really effective way that’s intuitive without exposing you to that complexity, all the way up front. And so I think there’s going to be a lot more innovations on the product and an interface side there, that that drives goodness. So I think we’re still in the really early days, like you have to imagine you’d see, have you seen all this AI stuff come out recently, where you’re taking this incredibly complex thing language, and then figuring out how to extract it and automate it and perform it at a high level in the same way, people, you know, people do day to day, right. And so I think the need to happen on this on product data, and behavioral data in general, where there’s going to need to be a whole bunch of automation abstraction, that makes it simple. And I think we’re still in the very, very early days of that. Now, I wish I had the answer to all of that. We’re still trying to figure out in a lot of different ways, like one of the things we’re looking at is like, how do you map your data? So if you have 1000 different data points, how do you map to those 1000 data points in a reasonable way that you can understand them? For example? Yeah. So anyway, I’m really excited to see what sort of innovations happen over the next few years, because every innovation is going to double the number of people that this data is accessible to.
Eric Dodds 48:10
Yeah. And so one last question, as we sort of wind up here. Maybe I always say that, and then it ends up being two or three. But does the data model become more opinionated as you move towards as you move in that direction? Right? Because you think about I mean, there is a data model with an amplitude, right. But in many ways, at least as a user, my experience is that you’re modeling the data in a way that sort of gives me a canvas that allows me to sort of build my own models on top of that model, if you will, right. So funnels or cohorts or whatever, right. And so it’s sort of like a blank canvas, there is an underlying data model that as you move in that direction, it stands to reason that you would actually become more opinionated in certain ways.
Spenser Skates 49:00
Yeah, you have to, you have to, there’s no way you have to make some set of assumptions of what you care about. And what you want to look at what the you have to, for example, just let’s take events,
Kostas Pardalis 49:08
if you have 1000 events,
Spenser Skates 49:12
you can’t show those 1000s of events to an end user, and expect them to have to pick between them, you’re gonna have to make some assumptions for them about which ones are important and which ones aren’t. So from just from that standpoint, you have to start to opinions, what sort of charts and analyzes you present out of the box, same date, you know, if you want to have a default set of dashboards, someone doesn’t have to think about which charts they want to create, you’re gonna have to provide that for them out of the box. And so I think, real art? Well, if you look at what Google Analytics did, with as a marketing leader, there’s typically a standard set of questions you try to ask with my traffic coming from what content is popular. How long are people spending on site? And so you Google Analytics, you can drive those with products, it’s actually quite custom. And you need right what makes it challenging is you need a forcing function to align everyone towards a kind of similar point of view. Because if you ask people really simple questions like, hey, what do you consider a DA, you, every single business give you a different answer on that question. Question is, how do you simplify that, so that you get some sort of standardization and a reasonable set of answers. And so anyway, that’s where I think the exciting part is in the space and every innovation that we that that gets developed is going to have a massive impact in terms of the number of people who are going to data,
Eric Dodds 50:28
Do you envision that as sort of an industry standard? I was gonna say, almost open source, but I don’t think that’s the right word, especially using Google Analytics as the example right? You know, sort of like Google Analytics, you sort of have like, uniques, you have sessions and some of those standardized things, right. But it doesn’t really extend to the point of, you know, sort of daily active user, other things like that. Do you envision some of those things sort of being established? Or even just do you envision amplitude playing a role in sort of, like, industry wide saying, like, yeah, not everyone’s going to adopt the same exact definition, but like, we’re sort of establishing this as like, the benchmark for these sort of core product flavored metrics.
Spenser Skates 51:13
Yeah, I think it’s gonna, for sure, it’s gonna need to be something bigger than just amplitude and what we drive. I mean, people have asked us for those standards, and we’ve developed a few like we’ve written, we have a product Analytics Playbook on our website, we have retention playbook, we have a bunch of other pieces of things. It’s like, you know, here’s how you should think about each of these metrics. But the question is, you know, what sort of outside forcing function is like, coming through the ecosystem? And start pushing people to do the same thing? That I don’t know the answer to? You know, I think it could be, but that will be really important to the long term success, because that enables all sorts of shortcuts and fast ways of getting to the data and being clear about what’s going on.
Eric Dodds 51:54
Yeah, for sure. Okay, well, we’re close to done here. One more question. And this is getting really practical . You know, you have built an incredibly successful product analytics company, but also have an engineering background and have probably thought more than most people about product analytics. For our listeners who maybe haven’t used product analytics very much like, let’s say, after this episode, they go, you know, wire up the amplitude, like, what would you encourage them as their first sort of, like, couple of reports to explore? Like, where’s the best place to start with product analytics, especially considering you know, it’s not necessarily your Google Analytics, we just get these reports like, where
Spenser Skates 52:32
should they start, I would start off by picking five things, five events in your product that you want to understand. So just, it could be Hey, someone used this feature today, or someone completed a checkout or someone hit the landing, like just pick five, start with that start by instrumenting, that that would be the most important thing to do. From there, you can get all sorts of insights on how people are rotating over time. Are they coming back to the app? Are they getting stuck? And then you can start to narrow in on a deeper and deeper question. So, a very typical one is the onboarding funnel. A very important thing is how people get set up and started with your product for the first time. And so I think,
Kostas Pardalis 53:16
start off starting off with,
Spenser Skates 53:18
to not feel you have to instrument the entire application right away. That’s one of the biggest mistakes, because that can be a very significant effort. Yeah, talking 1000s of events, start off with five, I’d recommend five and you want to get 10 Great, but you know, just so that you can keep track of and that will help you really start to get an understanding of the shape of the usage of your product. And you can build on it and build your and that a lot naturally lead to more questions, which will lead you to insert more stuff, which will
Kostas Pardalis 53:46
lead to even more even more
Spenser Skates 53:50
insights from your app. And it’s it’s great virtuous cycle that you build on. My biggest advice is, don’t be overly ambitious with your tracking from the start, you know, instead let it build organically over time.
Eric Dodds 54:04
Well, Spenser, thishas been such a wonderful time, an hour has flown by it feels like we just just hit a record. So thank you for being so generous with your time.
Spenser Skates 54:12
Eric costs us fantastic in meeting you. Thank you everyone for listening, really appreciate it. And yeah, check out amplitude. If you haven’t seen it, we actually have a very generous free plan that you can just get started with right away. And yeah, hope to talk to you guys again soon.
Eric Dodds 54:30
Absolutely. You know, Kostas, I think one of the things that I appreciate most about that episode is there was sort of an underlying sense of humbleness that Spenser brought to the table. Taking a you know, he said, I tried to start multiple companies, they were failures. He starts amplitude, in many ways, sort of a almost a classic, you know, founder story where, you know, you’re trying to build a company and you end up solving a problem that isn’t directly related to like what you’re trying to sell that company that ends up becoming, you know, what you really focus on, they took it to market. Over a decade they went public. And his willingness to explain how small amplitude is in comparison to companies like Salesforce, and Adobe, I think, speaks a lot to why, or at least in part, why they have been very successful. Because if that’s their attitude, you know, in many ways you almost hear him talking. It’s like, we haven’t even scratched the surface, right? We have so much more to learn so much more to build. We’re so small compared to the established players. And I just think that’s an attitude that I’m going to take with me and really think about over the next several weeks, because I think that’s rare, right? To see that. And it was really encouraging, and I think it says a lot about the future of amplitude.
Kostas Pardalis 55:52
Oh, 100%. And I would add to what you just said that the excitement to seal hearts, you know, like, most people that when they think about like their printers, super lakes, I think a company they think of like a survey where you starts, and hopefully you grow it, and at some point you x sheets, and we were another we will take the major exit being by going public, and then you’re done. Yeah, you cross the finish line AIDS. And then you talk with this guy, and he’s like, Oh, we’re just starting. Yeah. Like we have just scratched the surface in this leg. So what’s more that we can do here it likes, so much more like to build. That’s, together with what you said about how humble he is, something I’d like I’ll definitely keep and think about and reflect on Ethan’s shared experiences like how much of these cycles were problems.
Eric Dodds 56:56
Indeed. Well, thank you for joining us on The Data Stack Show. Always a pleasure to talk with brilliant people like Spenser skates. Many more on the horizon this spring, so definitely stay tuned, subscribe. If you haven’t, we’ll notify you of new episodes. Also, we have the live meet and greets data Council Austin and march you definitely don’t want to miss. Go to Data sec show.com to sign up and register and you can meet Costas and I in person. And we’ll do some live recordings. Until then we will catch you on the next one. We hope you enjoyed this episode of The Data Stack Show. Be sure to subscribe on your favorite podcast app to get notified about new episodes every week. We’d also love your feedback. You can email me, Eric Dodds, at email@example.com. That’s E-R-I-C at datastackshow.com. The show is brought to you by RudderStack, the CDP for developers. Learn how to build a CDP on your data warehouse at RudderStack.com.