Why we built Propel

0:00:10 - Tyler Wells
Today on the Data Chaos Podcast, we're going to have a conversation with Nico Acosta. Nico is the co-founder and CEO of Propel Data. Prior to that, Nico spent nine years at Twilio, where he was a director of product for Programmable Voice and was the founder of Twilio. AI Twilio took Twilio's voice products to 100 million ARR and I had the opportunity to partner with Nico when I shipped my first product to Twilio, the network traversal service. I hope you enjoy the conversation we had on the why for building Propel and how the customer centric experiences he had at Twilio led him to the journey he's on today. Enjoy the listen. 

0:00:48 - Tyler Wells
Nico welcome. 

0:00:49 - Nico Acosta
Thank you, Tyler. Thank you for having me on the show. 

0:00:52 - Tyler Wells
Absolutely. Why don't you tell me a little bit about yourself and your background? 

0:00:55 - Nico Acosta
Of course, I'm Nico Acosta, co-founder here at Propel. Prior to Propel, I was nine years at Twilio. I was an early employee at Twilio. I joined as a third product manager at Twilio. At the time it was a 60, 70-person company and I was there for just under nine years. It's a phenomenal ride. Got the chance to work on all sorts of developer tools, API platforms and mostly help developers integrate communications into their apps. 

0:01:26 - Tyler Wells
That's awesome. That's a heck of a run. Then you left almost two years ago, a little over two years ago. 

0:01:33 - Nico Acosta
Yes, December 2020. Awesome. 

0:01:37 - Tyler Wells
What sort of led you to where you're at today here at Propel? 

0:01:41 - Nico Acosta
Yes, I've always been an entrepreneur. Before Twilio, I had started my own company back in Colombia, where I'm from. When I came to Twilio, I remember when I interviewed with Jeff, he very explicitly said I love to bring entrepreneurs to lead products. That's how it felt at Twilio. After nine years, it was time to move on and to continue my entrepreneurial journey. As I reflected on what were some of the hardest problems we tackled building products at Twilio from the very early days to even when we scaled. They were all around servicing data to our customers Servicing, on one hand, to improve the self-service experience and, on the other hand, to empower complex deployment and enterprise deployment. To have visibility into what they were doing. From the very early days, we wanted to surface webhook data to our developers. It was a real challenge. At the time, we were just a web team that had no data engineering capability. It was always that you couldn't do it, you couldn't do it. It was super frustrating. 

0:02:55 - Tyler Wells
Do you think that was technology or do you think it was just internal to the company? 

0:02:59 - Nico Acosta
At the time I think it was a mix. The web team at the time was a team that was a PHP stack 10 years ago. They never worked with data. It was the front end, that was all the experience. It was definitely a capability. Also, the tools were very, very rough at the time. We had a Hadoop-based warehouse that you had to do pick queries. These worlds were opposite sides of the universe.

0:03:36 - Tyler Wells
Yeah very, very different.   That's wild. Now we're customers asking were they asking for this data? They're sending webhooks to Twilio. I could say us, because I was there at Twilio as well. They're sending webhooks to us and at the time were you getting a lot of customer demand asking, hey, what are the results of this? How are these webhooks performing? What did that look like? 

0:03:56 - Nico Acosta
The symptom was really customers weren't seeing the webhook failures because the webhook comes from Twilio to their customers. Is that fails, for example, a certificate failure, or if your app fails, you don't see it on your end Because it's Twilio firing. You need that visibility. One of the things that happens when you don't provide customers with data is that it's immediately your fault, it's immediately the platforms. We started hearing more than customers asking for data. We started hearing hey, twilio, you're not sending me the webhooks or you're dropping webhooks. It's like is it really that we're dropping webhooks or is there something else? It became a problem like we need to visualize and help our developers visualize what's going on. Why is it failing? And then is there a certain spike or is there a consistent failures over time that they need to address? And it ended up like Twilio was sending the webhooks but it was super difficult to troubleshoot And we identified that as one of the main roadblocks of the developer experience. 

0:05:03 - Tyler Wells
So what did you guys do? What did that first project or product iteration look like You're getting all the webhooks, you're getting the results. No one can get to it. You've got angry customers. 

0:05:13 - Nico Acosta
Yeah, this was a product that we called the app monitor back in 2014. And we essentially figured out a way to roll up the webhook errors and show it in a dashboard in the console So like you could see the count of errors by type kind of classical analytical data problem. We had to do this piggybacking on some of the billing infrastructure. That was the most advanced data infrastructure that we had at the time because the team couldn't build it themselves. 

But, interestingly enough, when we faced a similar problem in voice, when we were scaling voice and deploying WebRTC at scale, there was a similar challenge where You have thousands and thousands of agents taking and receiving calls from their browsers in their home connection and a lot can go wrong there, and it does. And if you're managing that deployment, if you're the contact center manager, if you're the lead developer on this, you're hearing all these things that call quality is not good enough, that calls are dropping and you need data to be able to manage it. Sometimes it's a version of Chrome that's buggy, or sometimes it's a specific user and specific network, and that's when, obviously, we got together. We started working together in the voice and video business unit to build voice insights. 

0:06:41 - Tyler Wells
So, like your first sort of like I guess like toe in the water when it comes to data is solving. You know, customer solving, customer problems, getting them the information they need so they understand how webhooks are performing or not performing in this case. And then the second one, sort of more informed and probably a deeper problem, is you've got all these customers using Twilio Client at the time and we're gathering a lot of data in the back end and now they're showing up saying, hey, these things are failing. I have no insight to this. We need to do something. I think it was Zendesk, right, that came to you and said, hey, look, we're gonna leave if you guys don't give us this data. 

0:07:19 - Nico Acosta
Yeah, it was very simple. They were like, hey, we need to. They launched a product at the time called Zendesk Voice. That was kind of a contact center functionality inside of Zendesk, and they said, hey, we need to be able to answer support tickets. So it's Zendesk Voice support tickets for our customers, which is a completely reasonable ask. Right? If we have to proxy a support ticket, every support ticket that we get, proxy it back to Twilio, this relationship's not gonna work, this is not gonna scale. So we need the data to be able to answer this. 

0:07:55 - Tyler Wells
So basically at that point Zendesk is writing into support, support is throwing that over the wall to engineering, engineering's running queries in the back end. Some internal tools, some internal tools, and they're giving pretty low-level tools And so they're turning around giving answers back and data back to the customer in this case Zendesk and so you've got this very expensive round trip. 

0:08:18 - Nico Acosta
Yeah, and what I love about those two stories is that in the webhook case we were removing friction for the self-service experience for the individual developer. For the developer that's starting, that's kind of getting their first web hook up and running and operating at the beginning right, although the webhook thing is useful at scale as well. But the Voice Insights was a power user and was like how do we empower our biggest customers to go 10x and what types of things do we need? And that was one of the secret sauces at the time where what they needed was visibility and operational insights into their deployment. 

0:09:03 - Tyler Wells
So what was that? like you sit down for the first time with Zendesk and they showed up with this problem and it's like, Nico, you and the team have got to solve this. And then you come back to them. You're like, okay, here's Voice Insights. How did that sort of change the discussion at the table at that point? 

0:09:18 - Nico Acosta
It completely changed it because then it was, it was very empowering to them to be able to have a to have that data at like their fingertips where, like they could use the Voice Insights tool, which was a set of dashboards for them that were kind of very opinionated on towards like that customer support workflow, to like drill down and answer those specific customer like voice quality issues. It was a game-changer.

0:09:47 - Tyler Wells
So now, like you've gone out there and like basically flipped the script right, Because we had built all of these dashboards internally for years, because we had to support our customers, right, We had to have a deep understanding when things were failing, when things were working, quality, everything else like that, and then we turn around and it's like we're not gonna give you our internal dashboards because obviously they're probably not as pretty, not gonna be. 

0:10:11 - Nico Acosta
You know, maybe don't look at, yeah harder to use. 

0:10:14 - Tyler Wells
They're kind of like you're just slapping together in DataDog or whatever you're using at the time, and now you've got an actual data product. You're turning around and delivering to them and empowering them with that data, like you said. 

0:10:26 - Nico Acosta
Yeah, absolutely, and the interesting part about that experience was that, if you look at what we did, we ended up building the data team at the time was more like of the warehouse, like starting the first versions of Redshift (Tyler Wells - where data goes to die), and we needed to move fast. We needed to get this problem solved, like we were one of the big teams that were the big revenue engines of the company and we ended up staffing a team to build this and that team grew significantly. 

0:11:04 - Tyler Wells
What do you think it was at first? Was it like what? five to 10 engineers Were fully staffed team to kind of deliver the first iterations? 

0:11:11 - Nico Acosta
Yeah, the first iterations that we put like in front of customers in production like called it a beta. It was around five to six engineers, five to seven engineers, and then it grew to 20 something, as obviously the data grew, the feature set grew and the business grew. One of the things that we were unsure, especially for this use case, was how, how willing customers were, the willingness to pay for something like this, and we came up with a model where you have some insights for free and then you had an advanced paid version and it sold really, really well. 

0:11:51 - Tyler Wells
Didn’t it set the record for the fastest product to a million-dollar ARR, or something like that? 

0:11:56 - Nico Acosta
Yeah, it's sold really really well. And it was very clear to us that it was a big problem, that we were hitting something that customers needed, but then kind of this model was not unique to voice quality or voice or even communications. When we started, when we shipped this, all the other teams at Twilio were like, wow, I want that, I want that, I want that. So the messaging team was like, hey, I want messaging insights. And the IoT team was like, hey, I want IoT insights. And, more interestingly, our customers then started saying like, hey, how can I get this in front of my customers? Because I don't want my customers sending me support tickets either. I want to empower them to answer their own questions. So we had customers like Zen Desk, like Salesforce and all their saying like, hey, we want to incorporate this functionality into our product, And obviously we had built it in a very lean way. 

0:12:59 - Tyler Wells
But it was also still API it was API first right, So like there was always the API. Yeah, okay. I mean like all things at Twilio right. We started with the API, of course. In this case, because you want to deliver a dashboard, customer-facing dashboard, you had to put some user interface on it. But I very much remember customers saying hey, this is great, I've got this dashboard, I can use it. amazing. But I really want to. I want to hit the API and pull that data into my internal tools. 

0:13:25 - Nico Acosta
And some did, and that was when we realized that customer-facing analytics was just so much more than dashboards. We had customers hit the API and we also had use cases for internal use cases for that data. So our routing engine at the time they consumed the voice quality insights data to make routing decisions And like that needed to be with all the things that a customer-facing analytics product needs to be, highly available, and support high concurrency. It needs to respond super, super fast, needs to be reliable, and so we realized that there's some API-driven use cases, there is some visualization use cases and then there's kind of some data sharing use cases where customers really want just the raw data and to massages and to do whatever they want with it. 

0:14:23 - Tyler Wells
And so, basically, they could do all of that. So the first iterations of this were I've got the dashboards, I've got the APIs, you've got machine-to-machine use cases. You sort of run the gamut at this point. Yeah, so how does this start to influence your next move? Obviously, like. You now need the groundwork and the foundation throughout a couple of projects, big projects inside of Twilio, big customer-facing projects And let's see what point did you leave voice? 

0:14:52 - Nico Acosta
I Left voice I think it was early 2018 to start the AI and ML business unit. 

It's like very data heavy, very data-driven. I would face the same problem is like, hey, we want to do insights on like the speech recognition stuff that we're doing, the NLP stuff that we're doing, and, as a small team, was like, oh God, do I have to staff this? And then that's where, like, I wish we had a platform, like an analytics API platform, that we could just hit and get the data that we need in a performant way that we can use in our products. And so that's where the bug started. like, this should really be a platform. I love what we built in Voice Insights and love what we built with the Debugger, but there's an opportunity for a platform so that every product development team could just build this without having to staff up on data engineering, which should be one of the yeah, one of the most expensive, one of the most expensive, one of the hardest roles to hire And in voice, we had the fortune to work with some phenomenal, phenomenal data engineers, but that's not always the case. 

0:16:08 - Tyler Wells
No, no, I mean that's super hard. I remember trying to staff that And when we were hiring for that, I mean you're competing with so many other companies out there and you're sort of competing with this super value to build this super value add product and trying to staff up, for that is when you're already trying to build the product right, you've got to hire for the AI team and then you've got to turn around and say, okay, I've got this AI platform now that can do the voice rec and speech detail, all these really cool things. That just generates more data. Now, just more data. And so now that I've got more data, I've got to hire more people. 

0:16:43 - Nico Acosta
Okay, Yeah, and then come to the end of my tenure at Twilio wanted to take some time off and decompress from nine years, and as the weeks passed, it became increasingly clear that I thought about like what would have made my life easier, what would have made my team's life easier, what, how could I've gotten to build more of what I wanted at Twilio? And it was very clear that having customer-facing analytics platform and that that problem needed to be solved at the platform level, not at each individual team level. At the time, like Twilio had like 100 or so product teams, so it's just not feasible for that to be solved all around the company. So that became very.

0:17:34 - Tyler Wells
Well, it's feasible, but it's very expensive, right, I mean?

0:17:36 - Tyler Wells
and like you've got to hire for that. So it's like, okay, you want to build more, but you're build more in this case is I don't want to go build more customer-facing analytics, I want to build more speech rec. I want to build more AI. I want to build more roadmap. 

0:17:50 - Nico Acosta
And then sometimes the world of customer-facing analytics is a spectrum right. Sometimes you just want, hey, I want to show a little chart here for, like the trending price of an airline ticket right before the checkout, flow right.

Yeah, it's all data, it's all analytical data, and that will help me convert more. Say, it's just one chart, right, that's it. So sometimes you just need to build one chart and do an experiment, or ship that chart and move on with your roadmap, and sometimes you have a whole like analytics product line that you're envisioning or anything in between, right, exactly. 

0:18:34 - Tyler Wells
So here you are, you're kind of like you now left Twilio. 

Where you're in the process of leaving Twilio, You've got that sort of that bug in the back of your mind of like the entrepreneurial bug again And tell me, how do you start validating this idea right, because I mean I know you're gonna do a lot of diligence, you're gonna spend a lot of time, you're not just gonna go jump on the first thing that kind of catches your mind and start building. Tell me a little bit about that process, of how you validated this and really sort of solidified your thinking. 

0:19:01 - Nico Acosta
Yeah. So from my experience, I had clear this was a challenge that we faced at Twilio. So my biggest question is, like, is this a challenge all our companies have? So I had the nugget from all the Twilio customers that said that they wanted to build insights into their products. So I started talking to customers. I started talking to potential customers, right And I started talking to companies all our founders that I knew and I found that it was always, always a frustration. There was always analytics, it was this thing that just, the can that kept being kicked down the road, that it was always punted. And as companies start to scale like customers start asking for it. Well, not asking, it sounded me like they demanded it. They demanded yeah. So I talked to a bunch of companies like this is really a big problem. When you look around, you see, like, take a look, like all the 99% of the SaaS products out there, they're analytics kind of suck right. 

0:20:09 - Tyler Wells
Well, yeah, and how are they solving it? So like, if you look the SaaS products today that are trying to deliver that are not doing it, Twilio style, right. So Twilio style or like. 

0:20:17 - Nico Acosta
LinkedIn. all the folks built these teams And LinkedIn huge team they built a bunch of Pino right Right, Like they built super hardcore technologies there to solve the LinkedIn insights problem. 

0:20:29 - Tyler Wells
So the companies that can't make that investment don't have the capital to make that investment. How are they trying to solve that problem? 

0:20:36 - Nico Acosta
So there's well, a couple of ways that you go about it. One, obviously, is to build your own. You stand up your own analytics infrastructure. That's kind of like the LinkedIn, Uber, Twilio kind of companies of that generation did that. The other approach is doing embedded analytics, which is slapping an embedded looker that queries your Redshift or your Snowflake and embedding that in your product. So a lot of companies have done that in the last 10 years And, like there's a Gartner report in embedded analytics, it's nothing new. 

0:21:15 - Tyler Wells
Okay. But so you see that and like there's obviously a solution there today and it's embedded analytics or build your own. It's kind of the two, I would say, sort of like competing ways of solving this. 

0:21:29 - Nico Acosta
Yeah. 

0:21:30 - Tyler Wells
But what's the feedback you're getting? I mean because you still went out and started Propel, obviously. So, as you're again doing your research, kind of, what are you hearing? 

0:21:39 - Nico Acosta
The first thing you do is put your product manager hat on right. There's just no way in hell I wouldn't embed Looker into my product. Never did it at Twilio. That would just not be acceptable. It's a clunky experience. It doesn't feel native. It's inelegant; it's just slow because it depends on the underlying data warehouse And essentially the resources that you don't control right. And then ultimately, you're kind of building out, painting yourself into a corner, building out tech debt the flexibility like that assumes. Like the only thing that you do is build a dashboard, build that, and sometimes you want to do that, sometimes you don't, sometimes you just want to include a number in the checkout flow, sometimes you want to hit the API to do personalization, sometimes you want to send an email, sometimes whatever it is, so that the embedded analytics approach ends up kind of leading you in a very tricky spot. 

0:22:49 - Tyler Wells
Sounds painful, sounds very painful. I'm sure the front-end folks and the designers love it, right, So it's like they just get this iframe essentially right and it's like here it is. This is what you get. And now I've got to ship that into my product, where I've spent all this time and money and effort to design this beautiful product, And then I get this thing that's like, not mine. 

0:23:10 - Nico Acosta
Yeah, I mean there are multiple problems. An iframe is the wrong contract between a product and the data stack. If you think about like kind of production-grade software engineering practices of like testing and source control, all that stuff, it's like you're embedding this kind of iframe that you can't do anything with it, right? So it kind of goes against all the good engineering practices. That said, it's impossible for it to feel fully native, right? 

0:23:49 - Tyler Wells
Yeah, you also said it's the wrong contract. What do you mean? Can you expand on that? What do you mean by it's the wrong contract? 

0:23:54 - Nico Acosta
A contract or an API, like one system, defines its interface to talk to another system, right, and that's the contract. Like I can hit these fields, or I can, there's these resources, these resources have these properties, right, in an iframe world. That is very muddy, right? If you're hitting who knows what table in the data warehouse and a column gets dropped, they'll break your iframe, right, and they won't realize that until you yell right? 

0:24:23 - Tyler Wells
Well, you hope you realize it because you've got testing in place, but you're probably going to hear about it when your customer goes to load that up and it's like, oh shit, this is broken. 

0:24:37 - Nico Acosta
In an API world, it's very clear what is a breaking change and what is not right And, like you, have a pretty clear. Each part has to respect their API and their contract. So you hit that. 

0:24:49 - Tyler Wells
You hit that testability like right on the head. then you've got like amazing, yeah, these embedded dashboards. 

0:24:53 - Nico Acosta
That's why they keep breaking right As the organization goes, like somebody changed the table, nobody realized. Or you decided to change the table's name like nobody realized. 

0:25:04 - Tyler Wells
And you're at this point that's serviced in the product. Got it, got it. Now. That makes a ton of sense. So here you are. You've got this background right. So you've got this amazing background. You've seen it at scale. You've seen what works, you've seen what doesn't work. You've gone out and you've sort of shopped the idea around. You've done a bunch of validation. You've seen what's in the market today. You've seen what is lacking, what people are demanding, what they really want, or you're starting to formulate ideas around that. Now what, what happens from here?

0:25:36 - Nico Acosta
The third thing is also what's changed since those early days of like, when had the Hadoop based warehouse, and what's changed? right And the big thing are like data stacks have evolved quite a bit, especially cloud data warehouses like Snowflake, BigQuery, And the barriers of entry for these warehouses are now super low. With a credit card, you can open your stuff like account, Like 2%. Companies can have a data warehouse that 10 years ago you needed an entire team to have right. 

0:26:14 - Tyler Wells
That's like a credit card. You get this massive sophistication. 

0:26:17 - Nico Acosta
Yeah, exactly, that's a fundamental shift, right. But then you get all your data into Snowflake. You do whatever you need to do, you transform it, but then, when you're going to expose it again, you have to build a bunch of stuff. At the minimum, you have to build a service that queries the data, that exposes it to an API, that caches it, that adds authentication, that adds a security layer, and then you're starting to kind of rebuild that, and that's the thing that changed. Now there are thousands and thousands of companies on cloud data warehouses, which has advanced their data stack significantly, which means that there is even more value to use that data in customer-facing use cases. 

0:27:05 - Tyler Wells
But if you take a couple steps back, you just kind of said something interesting, which is like the companies swipe their credit card, they've got their data going into Snowflake And then the important bit is now they’ve got to go build a team to get it out of Snowflake, to serve it to their customers, to get the analytics built, to have an API, to have securities, to think about multi-tenancy. 

0:27:24 - Nico Acosta
I mean that's a lot of work. It's a lot of work. And if you think about what are the differentiating capabilities that a company should have right? One is how they model their data. That's you're gonna need domain knowledge to do that right. How to model their data. Whether they do it, it's no flake with DBT. Whatever tool they use, you need that expertise. Those are like fundamental assets that a company has. On the other hand, the other capability is understanding their customers right. And what do their customers need from the data, right? Is it an API? Is it a dashboard? If it's a dashboard, what charts? How do you want it to slice and dice? How do you want to aggregate or know how? And that experience and knowing what to service the customers, how to visualize it, that's another core competence. Everything in the middle is known value add right. Everything in the middle can be outsourced, right, and that's what we're doing at Propel. 

0:28:32 - Tyler Wells
All right, so you've got all this ammo now and you're sort of you're ready to go, You're chomping at that proverbial bit of entrepreneurship. Tell me how you raised and what you went from that, how that sort of evolved. 

0:28:45 - Nico Acosta
Yeah. So I started talking to early stage investors probably talk to 30, 40 different investors and an interesting thing happened is that they've all heard that their portfolio companies were frustrated with analytics. So the story very quickly resonated and, despite being early, we did a seed round with Matrix fairly quickly and started building The found Matrix to be a great partner and partner that gets the data space really really well. It's a developer space. It takes a special type of VC to understand a developer go to market where you're empowering builders that then will build on top of the platform, and Matrix definitely had that experience. 

0:29:37 - Tyler Wells
I mean with the view over there one of the partners, Pat, coming from Twilio.  He's got that deep developer experience, which is great, and then tell me a little bit about. You've got a bunch of angels as well. Who are some of those folks that you can talk about? 

0:29:51 - Nico Acosta
Obviously So. We've got a phenomenal group of angels obviously a lot of the Twilio mafia, from co-founders to former CTO and pretty much all the product and engineering leads at the time. But it's not just Twilio folks. We've got some really, really good angels in the data space, like co-founder of Databricks, co-founder of FounderDB, of LabelDocs, and, yeah, really, really happy with the group that's supporting us. 

0:30:28 - Tyler Wells
Yeah they're pretty amazing.  I have to agree with you. 

And so here you are, you raise money. 

‍0:30:32 - Tyler Wells
You're getting ready to raise, you're getting ready to close on that check And you're ready to build. Where are you at now? What's happening with the company and how are things going? 

0:30:42 - Nico Acosta
Yeah, so that's when we start working together, start building the team, And that's when we earlier we reconnected again, and so you joined me in early March 2021. And since you joined, we were like OK, let's first do a POC, will this even work? 

0:31:02 - Tyler Wells
Yeah. 

0:31:04 - Nico Acosta
Is this concept of having an analytics API platform that caches the data, that serves an API and that you can build even feasible? Can it handle large-scale data? Will it be fast enough? 

0:31:22 - Tyler Wells
Yeah, we had a lot of questions but not a lot of answers. 

0:31:24 - Nico Acosta
Yeah, a lot of questions. So we started answering some of those questions. Even before going to more customers, we started answering some of those questions. And the thing with data is that everything works at a small scale. I think everything is fast at a small scale. It's when you put in load, when you put in terabytes, that the stuff gets complicated and you need to have the right architecture from the beginning. So if building API platforms is hard, building data API platforms is harder because of that, and so we did a lot of experimentation and that built in an experimentation culture inside of Propel.   Everything's an experiment, right. 

Everything is an experiment and double down on what works, yep, and we experiment a lot. We experiment a lot and these not always lead to shipping new products. But start informing us what's possible, what's not, and guide us through. 

0:32:31 - Tyler Wells
And so here we are. It's 2023, still pretty early. How big is the company and what are you looking forward to this year? 

0:32:40 - Nico Acosta
So we're 10 people. We've got a phenomenal team of engineering, design, product and sales, so we have a very well-rounded team. We brought some of the best talent from Twilio over. I'm super, super happy, but we also have some talent from other places. It's phenomenal. We're globally distributed, so we're between the West Coast and Berlin, pretty much. What am I excited about? 2023? 2023 is going to be a great year, like. We're starting onboarding some of the most interesting customers. So we started working with Courier last year. They've now fully deployed their analytics in production. It's their analytics on their notifications API. They also, as an API, they had a pretty significant amount of data. So, operating at a large scale, we were able to validate that value at a large scale, which is amazing, and they're growing very, very healthy And this is the year where we bring in more and more customers. 

0:33:44 - Tyler Wells
Yeah, to me the growth is going to be, I’m looking forward to it. Let's kind of drill into Courier just a little bit. They were the first customer, our first customer. What do you think like in your opinion? Obviously, we've heard from Seth (CTO), but what do you think it was that made them take the leap and start building on top of Propel? 

0:34:03 - Nico Acosta
Yeah. so I think you could boil it down into three things. First, they had the need. They were friendly and everything, but they had the need. They've always wanted to ship analytics and it was something that their customers were asking for. but they didn't get around. But they had the need and they had the urgency. Second, Seth, a very seasoned CTO. he had done this in the past. He knew how to do it. He'd built analytics teams before. he knew what it would take and he was very clear like I don't want to build this again. I don't want to spend five of my engineering headcount on this, but furthermore, I don't want to be carrying the pager for this stack. I don't want to get paged that the analytics stack is down because he's worked through operationalizing analytics, he's worked through data incidents and he's like I don't want to do it again, that's a big, big task. 

And third, we really connected on the philosophy of the product, on the API approach, where for them it's really important to be an integrated experience to fully control the look and feel. They can put a front-end engineer on it right but they're not gonna sacrifice their customer experience to save, like some front-end engineering time. So they built exactly the dashboard that they wanted using Propel’s API, so that the combination of those three of like the need and urgency, the experience that said, had have you built this before? And kind of like the alignment on the philosophy was what drove and made careers such as successful Propel deployment. 

0:36:02 - Tyler Wells
And how long do you think it took them from the concept of what it is they want to build? Cause I remember them showing up with a very good idea of what they wanted from a front-end perspective. Right, they started with that sort of customer-facing design even before, kind of like yeah, they had data over here and everything else like that.

But they started with that customer-facing design to influence heavily the data that they needed to collect, the way they needed to transform it. So once they showed up with that design, how long did it take them to get to production and ship it? 

0:36:31 - Nico Acosta
Yeah, I'm glad you mentioned it because it's kind of like the pool approach like you think about, like the manufacturing. It's the pool approach versus here's what we need and works backwards from it. Right, here's what the customer experience would look like and work backwards from it. And when you do that, you can really clarify your data requirements. So they knew exactly what data to get into their data warehouse, they knew exactly what to model it and once they had that, we did a hackathon. That was a day in San Francisco where we built the first POC in like an afternoon. It was. So got data flowing to their console in an afternoon and then obviously, tidying things up, cleaning it up and taking everything for production to a couple of weeks, and they were live. 

0:37:24 - Tyler Wells
Yeah, i kind of remember being on site when we did that hackathon. I think it was the very first time they were actually able to visualize that data, the first time they had put the time and effort to visualize that data. And they visualized it through, obviously, through Propel, through Propel and our APIs. And I remember them looking at that time series, the bar chart graph, and they're like wait a second, kind of scratching their heads going yeah, Okay, I think there's something wrong here. Wait a second. What happened? 

0:37:51 - Nico Acosta
We were already starting to discover insights just from seeing the data. 

0:37:55 - Tyler Wells
Yeah, and it was kind of like that first time and they kind of had to go back and realize that some of their transforms were a little bit inaccurate and then they turned it around and, of course, change the design and now they got this very pretty design that they have. So it's been exciting to watch it evolve over time as well. It's been amazing As a user customer. yeah. Well, Niko, I mean I think it's an amazing story of how you've gotten here, how we've gotten here, but I mean, more importantly, how you kind of started with Webhooks and started ideating, started getting the idea and that's sort of like the, the sort of, I guess I would say like benchmark almost, of like background, of sort of like experience from Webhooks to Voice, to AI, and then went out and tested everything. and now here you are, we're running Propel, you're running Propel, we've got great financing and we've got a great year ahead of us. So definitely appreciate you telling the story. I think it's a fascinating one and obviously happy to be on this journey with you as well. 

0:38:56 - Nico Acosta
So Yeah, I think it's very, very exciting, very. we've got a great team, we've got a great opportunity ahead of us. The world data is right now, probably one of the most underutilized resources that software companies have. The next 10 years, we're going to be able to do so much more, and it's probably the most important decision that startups make is how they're going to use their data, what types of products they're going to build with it, and, ultimately, that's where, like, the true differentiation is going to come, because nowadays, everybody can build a web app, everybody can build a mobile app, but not everyone will have the data and can build with data. So that's what's going to lead to kind of next generation of companies. 

0:39:47 - Tyler Wells
Excellent. So one last question before I let you go and we have to get back to our always busy day. If there's folks out there, if there's developers out there right now, developers, product managers, whoever that's faced with that same situation they're thinking themselves like my customers really need this data, like I've got to get this data in front of my customers. I need to get that customer-facing analytics. What would your advice be to them? now that you've done this so many times and now you built this company, what would your advice be to them? 

0:40:21 - Nico Acosta
Yeah, that's a really good one. So the first one is really start focusing on the customers. Really start focusing on what are the minimum requirements that you can get out, build, put an API like Propel in place and don't wait until your data stack is perfect. It will never be perfect. I see companies waiting to rearchitect their data platform and spend years and years and years. It will never be perfect. The best thing you can do for your team is clarifying the data requirements, making the trade off and simplifying them, and you do that by understanding your customer, having an API on top of it and then abstracting so that you can evolve your data stack kind of with more freedom. So that would be my recommendation Really drill down, go at it from the customer's perspective backwards. Put an API. Don't wait for data to be for like the rearchitecture of data to be perfect. Be very clear with the requirements. Narrow them down and you can ship something. 

0:41:25 - Tyler Wells
Beautiful. I think that's a great thing to end on. So everyone listening it's take that data chaos and get to that data clarity, because it is chaotic, but you can bring clarity to your customers and empower them. So that's a great thing to do through data. Well, thank you, Nico. Thank you. Have a great rest of your day. Appreciate it.