Kevin in the Wild / Gastinterview

Building 75+ Startups: Why Most Software Fails (w/ Kevin Riedl)

Aufgenommen auf Englisch 31:54

Aufnahme veröffentlicht

Transkript lesen
Transkript

Aus dem Audio mit Whisper transkribiert und mithilfe von KI mit den automatischen Plattformuntertiteln abgeglichen. Unverständliche Stellen stehen in eckigen Klammern. Originalsprache, ohne Sprecherzuordnung; Fehler bleiben möglich und die Aufnahme ist maßgeblich.

Intro & Kevin Riedl’s journey to founding Wavect

Diese Stelle ansehen · 0:00

0:00 Like everyone learns at a different pace, right? Some people learn it after the first product and some people don’t learn it after a hundred. I see that with very successful founders and CEOs that we’re working with, they don’t understand the basics. I see it really every week, people come with these sky high expectations.

0:19 There are many ways, but I think AI changed the process substantially. It shortened it. Sometimes maybe the idea is great, but it’s only something usable to yourself because for small projects that you only use yourself any money is a lot of money. Welcome to Inside the Stack, everyone.

0:50 I still remain your host Oluwole and I have here with me Kevin Riedl, who is a founder, the founder of Wavect. Wavect is a software company that has shipped like tons of products. I feel like they’ve shipped even more products than what I’ve built since my, since I started my tech career. I want to believe that it’s over 50, 50 products, 60, 70.

1:19 75, like, is what we publicly claim. Yeah. 75. Well, honestly, that’s a lot. And I’m sure there is a whole lot of things that we would get from Kevin. So what I want you guys to realize is Kevin is very handsome technically. And this conversation would sit right at the intersection of software engineering, product thinking, client work, and definitely building a company.

1:44 So for everyone out there who’s into tech and who’s actually thinking of building a company, Kevin is here to share his experiences, to share some of his thoughts. And we’ll talk about how his thinking has changed after shipping so many products. We’ll talk about how he approaches client requests, the tradeoffs behind technical decisions, you know, what happens when systems need to scale and how a lean engineering team can sometimes achieve far more than what you might expect.

2:18 So Kevin, nice to have you on the show. Thanks for having me. Really a pleasure. And thanks for the very nice intro as well. Hey, you’re welcome. So could you tell everyone, our viewers, a little bit about yourself, your journey in tech, what got you started in tech? Sure, sure. I will try to make it short to not bore anyone, but I basically was one of those kids that started building websites in, I think, 11 years, 12 years old.

2:53 And once that got to a certain point, I started actually trying to get into systems, if you know what I mean. Everything was legal, but I tried to find ways. And it was just always a challenge. And I just loved that. And along that time was when I joined my first identity-based startup, which then got me hooked into building my own startup.

3:27 And then one stone after another, like one brick after another was laid. And yeah, like lots of failures along the way, tons of learnings. And yeah, over the last years that there were investments into startups, there were another few projects that we have started. And yeah, since 2018, 2022, depending on how you want to frame it, we are basically building software projects for international companies, mostly US, Dubai, Singapore, you name it, like basically the innovation hotspots.

Operating an AI-native software agency

Diese Stelle ansehen · 4:04

4:11 And yeah, here we are. Yeah, that’s nice. All right. So can you just give us a brief about Wavect? What do you do? How? What’s your process like? Sure, sure. Like Wavect in general is just a hardcore AI native software agency, basically. So we really bank on the bet that mid-sized agencies will be in trouble through the recent AI changes.

4:42 Because you do have enough cost for your employees that you have certain pressure, but you also don’t have the benefits yet of a 50, 100 person plus agency. So that’s kind of our bet. So we said, okay, how can we structure our company in a way to basically outperform all these at least mid-sized agencies and deliver better and faster than most of them as much as we can.

5:14 And yeah, also about the company, we have been teaching kids around the world from Thailand to Germany, Austria, online about software development and AI for many years now. In summary, we have educated or let’s say conquered in certain ways, like crossed ways with over 10,000 kids. That’s at least our internal estimation.

5:42 So take it with a grain of salt. And yeah, from a like, if I understand your question correctly, the process of building software obviously is a long one and you can go into. Yeah, yeah, I get I get it. So I think maybe during the course of the interview or the session, we will definitely get into one or two of how it actually works.

6:13 All right. So my very first question is, you’ve been involved in building about, let’s say, give or take 75 products, give or take 75 products. So what do you understand about building software today that you simply couldn’t have understood? Maybe like after you’ve built your first project, your first two products, your first three products.

6:37 You know, it’s a significant milestone 75. Building five products is not the same as building 20 products. Building 20 products is not the same as building 50 products.

The #1 lesson from 75+ products: What not to build

Diese Stelle ansehen · 6:45

6:48 Building 50 products is not the same as building 75 products. There is a lot of learning along the way. So can you just share with us some of those nuggets of knowledge? What were those things that you did not understand then, but that you understood, but that you understand now? Sure. That’s a great question.

7:05 The main thing I would say is like everyone learns at a different pace, right? Some people learn it after the first product and some people don’t learn it after 100. I see that with very successful founders and CEOs that we’re working with, they don’t understand the basics and oftentimes not all of them, obviously.

7:27 And the main issue is not how to build, but what not to build and how to make sure you build the right thing, like the 100 year old story, sorry I’m exaggerating here, about product market fit. So making sure you build something people actually want and how you get there, right? So obviously things have changed with AI a little bit because it was, it is easier now to get something you can demo and use as some kind of an MVP, which is awesome and great.

8:07 But the core principle of, hey, make sure you don’t waste time, tokens, money, resources on the perfect dream product, as many people try to build, still holds up because I see it really every week. People come with these sky-high expectations. The product should do everything before they even on the market, even sold the product for a single dollar, anything like that.

8:37 And that’s why most software products actually fail in my opinion, and that’s definitely the core learning. I made that countless of times myself. I still sometimes tend to do it myself because it’s just so appealing, so so cool to think, okay, I’m building that because I think it’s cool. I think it makes sense.

8:59 But that doesn’t mean others think so too and think so enough to pay for it. Yeah. I think you’ve answered my next question like so beautifully because my next question was actually about, was there a time when you say you stopped thinking about how do we build this and then, you know, just started asking like should we even build this?

9:22 So I think you’ve, you’ve answered that perfectly. So most of us, especially those of us who are not in the business side, of course I’m a little bit in the business side, I won’t say I’m not in the business side, but for when I was starting up, I think it was always about how to build, what are the bells and the and the whistles.

9:41 I have not actually sat down to ask myself like should I actually build this, and I think that is one major question that anyone who’s building anything, most of the time if it’s not for your portfolio, as long as you want to build it for other people to make use of, you need to ask yourself like should I

Translating client ideas into code and managing feature requests

Diese Stelle ansehen · 9:54

9:56 should I just build this? Yeah, that’s that’s a wonderful, that’s wonderful knowledge. Thanks. Thanks, Kevin. All right, so another question is mostly about like clients interaction. When a client comes to you with an idea, you know, most clients don’t come to you with the idea fully formed. They come with, they come with, like, should I say, like a framework or something, or a feature request.

10:22 How do you get from what they are asking for to the actual problem that they have? What is the process? That’s an awesome question because it basically summarizes the difficulty in building individual software. It’s rarely about, okay, how do I build this from a technical standpoint? Of course if you build Netflix scale or whatever then that is a huge challenge, but in the beginning that’s rarely the issue.

10:54 So it’s usually about, hey, how do I translate what the customer has in his or her head into code. And the beautiful thing with AI now is they are doing that already by vibe coding stuff, and oftentimes or most of the times it’s some throw away prototype, but it helps you get a sense of what they want because they prompted and iterated and got their ideas during that process and added that or removed stuff and so on.

11:28 And that’s usually something that otherwise would have happened and still happens, obviously, during the discovery process where you get to a specification or where you get to an MVP, where they say, okay, that’s good enough to get out there in the first version and then to iterate upon. So that’s, there are many ways, but I think AI changed the process substantially.

11:56 It shortened it, because people are now able to express themselves in a visual manner, which they otherwise try to get down to some sentences or paragraphs or during course meetings. And most people struggle with that. Almost everyone, I would say, because it’s difficult to describe an image, right?

12:20 And then draw it without seeing the image. That’s kind of the difficulty. Right. Thank you. That’s a nice response. And honestly, I’m really, we’ve not even been like how many minutes now? And I’ve gotten a couple of things like in my own, on my own path, I’ve gotten a lot of things and I’m sure our viewers also would have gotten one or two wisdom from you.

12:48 So I appreciate you once again. I know we’ve not gotten anywhere, but I would want to appreciate you a whole ton. All right. So have you ever had to tell a client, like, I understand what you’re trying to build. I understand your ideas, but I don’t think you should build this now. You know, let’s, let me, let me give an example, for example maybe like a store who currently, should I say store, maybe like a client who sells clothes, who is, who is more or less like, should I say like an online vendor or something, who a client, what she just, what he or she just need is maybe, what is, what is it called, maybe like, should I say

13:38 Google, a Google address, a phone number or something, you know, or maybe just like basic websites just to tell the person, like just to tell people or clients like, hey, this is what I do, contact me on this phone number to actually, you know, get what I want. But this person wants something like an Amazon website, like a marketplace.

14:04 Have you ever had instances where you’ve had to tell someone that, like, I don’t think you should build this yet. I think you need to build your market out first. I do that literally multiple times a week. The reasons are different per project and person because every circumstance is different. Sometimes it’s what you mentioned.

14:25 Okay, are you sure you want to build this because competing with Amazon, for example, directly might be incredibly hard and nearly impossible. Why not pick a niche and grow from there? Or as you said, maybe let’s start with an audience first or let’s try to get some kind of a wait list or something like that.

14:50 Absolutely makes sense. Or sometimes maybe the idea is great, but it’s only something usable to yourself. I had that yesterday. Hey, I wanna build this. I actually just need it for myself. Okay, cool, awesome. Are you sure you want to spend money, like a lot of money on building this? Because for small projects that you only use yourself, any money is a lot of money.

15:20 Yeah. So there are many different reasons why you shouldn’t build stuff. And there are very little reasons why you should build stuff. But the main challenge is to not get discouraged by that and automatically say, okay, almost nothing makes sense. Because if you try to pull everything apart, everything is a no.

15:47 Right? So it’s about, okay, what risk I’m willing to take for what reward, what’s the outcome, and how can I reduce that risk? That’s all that any business is, or building software. Reduce the risk enough to increase the certainty to a point, and then try to act in a way to get there as cheaply and fast as possible.

16:14 And yeah, there are tons of ways, right? And the better you are in that, the more successful you become, and it’s a learning process, but I’m not [unclear] either, so I have tons to learn too. So it’s, yeah. Thank you, thanks so much. All right, so after building 75 products or more, what what what are those warning signs that you would see that would make you think that, hey, there is a probability that this project might struggle later on, maybe like, should I say client struggle

Handling scope creep and changing client requirements

Diese Stelle ansehen · 16:50

16:52 or something, you know. Struggle in what regard? Now when I said project struggle, let me maybe like, you know, we have some clients who are not satisfied. Mhm. We would, you know, we would have sat down, discussed the product requirements. Hey, this is what the MVP needs. Then maybe a couple of weeks later, we are halfway through the project.

17:17 Then he or she calls back and tells you, hey, I was discussing with my mother-in-law, and we feel like this feature might actually might be a good fit. So those things change. They change the face of the product. Mhm. They might tell you that, hey, you know, we could, it could just be like a client who is not, should I say, really satisfied.

17:43 You might use for the project, let’s, let me just take it maybe like blue color. And then tomorrow he says, hey, I want this to be yellow. And then later on, he changes his mind, like this yellow color does not look, does not look fine to me. I want it to be orange. You know, things, little little things that might be frustrating.

18:04 It’s a fantastic question. There are two facets to it, I would say. On one hand, always consider yourself an advisor, I would say, because in the client provider relationship, you’re trying to satisfy the customer, right?

18:30 Of course, that involves to make them successful. But if they want a terrible color or if they want a feature that in your eyes is absolutely nonsense and you told them, hey, okay, I would try to validate that further or are you sure you want to build it? In my opinion, it doesn’t make sense. And they still want it and they pay for it.

18:53 it, it’s their beer, right? They pay for it and they want to take the risk because maybe I’m wrong. I can just advise them with my own experience. That’s that part. As long as they’re fully informed and know the risks, that’s fine. Because otherwise I would create that negative tension of telling them what not to do and they want, hey, but I want that and I like that.

19:20 And then there’s this kind of competition and you don’t want that, right? They’re the, I wouldn’t say king, because that’s also not true, but they’re your partner and it should be on eye level. Just tell them how you think, what you think, and if they want it anyway, give it to them. The second one is, because you said weeks later, and that’s the more challenging part, is the scope itself.

19:51 You agreed on a certain project scope, you said, okay, I’m charging you X for that. Yeah. And then weeks later, they changed their opinion and want to do something else. The main mistake, or there are basically two ways to solve that is, you either work with retainers or on an hourly basis, which has a lot of disadvantages, as you can imagine.

20:16 But at least you’re flexible and the customer can change their mind how often they want because they have to pay for it. That’s fine. If you have a fixed price, it’s super important to be super clear in the offer. Don’t assume anything. Everything that is listed here is included, and everything that is not listed here is explicitly excluded.

20:41 And if you make that really clear, you will see that customers tend to iterate and discuss the offer a lot, which is annoying, but it’s less annoying than changing the scope layer and they’re pissed and you’re pissed. And so that’s the best solution for that. All right, so now let’s get into the technical parts.

21:05 Can you take us through a system or a product that you’ve built where performance and scalability were important in the beginning? What problem were you solving and what made it technically challenging? I mean, scalability is inherently challenging always because you have to account for concurrent operations.

Designing scalable systems and real-world fintech challenges

Diese Stelle ansehen · 21:24

21:32 Although people overestimate how much concurrent usage they actually will have. We had clients that said, hey, we need to make sure this thing can handle a thousand users concurrently at the same time. But then in the end, they had maybe a thousand users per month. It’s just a very different thing. So, yeah, and that obviously makes a huge difference in how you architect and build something.

22:03 From a technical standpoint itself, this project specifically is a great example because it even was a fintech product, which means it kind of was e-commerce checkout and people had to be able to pay through different means of payment, which included crypto payments, which included fiat payments and so on.

22:28 And we also had to include different payment providers. So you had to deal with a lot of external factors. These things had to be very available. Like the third parties too, because they have their own limits as well. And you obviously also have to auto scale your thing too, because in this case it was a drop based marketing, which means demand searches and then it falls off.

22:57 Just having a really high, like a lot of containers running, a lot of resources in your cloud running, waste a lot of money. You have to scale down and you have to scale up on time. That’s a main challenge to make sure, hey, no matter how the usage grows, we can handle the load, the concurrent load, especially [unclear] a certain point.

23:22 And you do that obviously with, of course, with benchmarking, like actually testing that, making sure, okay, how does the cloud behave? How does your application behave? Does it get slow? And all these things. And is like what happens if something money-wise, like on the transaction side, is happening.

23:45 How do you handle that? How is it then on the support side? Like when something goes wrong, where can users go to and how is that resolved, both business and technical? So there are tons of things and the higher the scale you go, the exponentially difficult this gets. So that’s why I urge founders to be humble here oftentimes.

24:15 Say, hey, are you sure you will have this usage right from the start? Do you really want to optimize for that? Because it will cost and take a lot more time. But yeah, there are tons of difficulties with that. The tools change, the process changes, everything changes. Now, when you’re designing, when you design that particular system, did you decide to scale from day one or when there was demand, that was when you built in scaling?

24:51 Are you talking about myself or customers or? No, you used the fintech product as an example. Mhm. So I was asking, like, number one, was it a new product or it already had users? That’s one. Then two, when you built it, did you factor in the scaling at the get go or it was later when there was demand for it?

25:20 Like in that specific product, there was a marketing agency involved and they obviously honestly prided themselves for saying, okay, we are able to generate that demand. And they did generate some demand. Like in summary, we have raised or the client has raised over a million bucks, something like that.

25:43 But yeah, it wasn’t like it still was, I think, like just 40% of what they aimed for. So they didn’t achieve their goal. But yeah, in that case, it basically was okay, let’s get this right and scalable from the get go, which makes sense. It’s just the scale of how scalable does it need to be from the start was the wrong decision.

26:13 But that’s more reasoned because, yeah, stakeholders. Thanks. So have you have you ever had any system or any project

Dealing with wrong technical assumptions and heavy server load

Diese Stelle ansehen · 26:25

26:25 where your initial assumptions about the performance and scale actually was wrong? [unclear] that was so honest. And I would believe that’s also the right way of doing it. And that’s why we have fallen back to agile development.

26:51 Now it’s maybe more spectrum development with AI. I don’t know. Or loop-based engineering. But you always make assumptions. And the assumptions you make. And I really love Jeff Bezos' thought here as well. That’s a technical topic, but you can still apply it in some way. Is it a two-door decision or a one-door decision?

27:17 A one-door decision is you go through, you can’t go back without meaningful costs. A two-door decision is a two-way decision is you can go through and you can go back without incurring a lot of cost and resources. So and I think that’s like if you’re in university, you will immediately learn, OK, everything around a database is usually a one way decision and it tends to be true to a certain extent.

27:50 And other things are two way decisions oftentimes. So it’s really, I think all about that and most assumptions are just wrong and being okay with that and being clear with customers or people you engage with, hey, it’s totally normal that we have to change things to get to something that actually works and people want.

28:16 And yeah. If that answers your question. Yeah, it does. It does. It does. Actually, it does. Because as both an engineer and as an entrepreneur, there are times when your assumptions might be wrong from the get go. You might scale for a thousand users and then you have 2000 users. You can scale for a thousand users and just get like 500 users.

28:39 So it’s correct. Look at the big ones. GitHub. How many, like how often have they been down this year? That’s a huge worldwide enterprise product. And they still struggle with that. They didn’t expect that amount of usage. It will always stay that way, being okay with that.

29:06 Honestly, I think we’ve had AWS outages about twice or thrice this year. AWS also had a couple of outages. So even the big ones fail sometimes. Absolutely, yeah. Yeah. All right, so when a system starts struggling under load, when a system starts struggling under load, now this is more of your engineering side, not the entrepreneurial side now.

29:37 What do you look at first? Do you look at the database first? Do you look at the infrastructure first? Do you look at the architecture, the application code, or somewhere else? Depends on the maturity of the application. The more mature, the more, let’s say, alerts, analytics and so on you hopefully have, which means you know, okay, what is wrong and what might not be wrong.

30:03 You also hopefully do have a CI/CD pipeline that catches a lot of things, like software tests and so on. If something breaks, usually you start with, okay, is every component of the infrastructure up or down? Like visibly down. And then you, like as most people, you basically just start digging into it.

30:29 Like which component possibly is it? Then you look at the logs. And then you hopefully are able to fix it quickly. But yeah, obviously, the more mature the product, the way different things look like. But that’s in its essence, as always, from the MVP to enterprise, how it will always look like when you really break it down.

30:55 Someone looks where the issue might be, you nail it down, you fix it as quickly as possible while staying true to your process. And then hopefully everything goes up again. Thank you. Okay, guys, thank you so much for staying to the end of this session. Thanks, Kevin, for your time. Thanks, Kevin, for staying to the end of this session.

31:22 Yeah, so we’ve come to the end of the session. I really appreciate you all. I hope you guys have gotten one or two nuggets of wisdom from our guest Kevin. Please don’t forget to like, share, and if you’ve not yet subscribed, please do so now. Subscribe and I still remain your host [unclear] and I’ll catch you in the next episode.

31:41 See ya, bye.