Why Lean Startup Isn't Dead (And When You Need Something Bigger)

Is the Lean Startup methodology dead? Peter and Dave tackle the growing criticism around MVP approaches and explore why this foundational model still has its place in modern product development.

Drawing from George Box's famous insight that "all models are wrong, but some are useful," they discuss how tools evolve but don't necessarily become obsolete. With AI making prototyping faster and cheaper than ever, the conversation explores what's changed about experimentation and what hasn't.

The hosts dig into common misapplications of Lean Startup principles, from "MVPs" that take months to build to organizations that skip the crucial feedback loops. They also explore when incremental learning isn't enough and you need those bigger strategic pivots.

Plus, they make the case for Wardley mapping as an underutilized tool for spotting step-change opportunities that incremental approaches might miss.

Key Takeaways:

• Lean thinking remains relevant: Lean Startup is still a valuable operational tool for continuous learning and incremental improvement, but it works best when properly implemented with complete feedback loops
• You need multiple models: Organizations need both incremental change capabilities (like Lean Startup) and strategic step-change tools (like Wardley mapping) to navigate complexity effectively
• AI commoditizes development, not strategy: While AI tools make prototyping easier and faster, they don't replace the need for good product thinking, user experience design, or asking the right strategic questions

Whether you're questioning your current approach to product development or looking for ways to balance tactical execution with strategic vision, this episode offers a pragmatic perspective on using the right tools for the right challenges.

Chapters

• 0:00 All Models Are Wrong, Some Useful
• 4:09 The Evolving Role of Lean Startup
• 7:26 Commoditization of Development
• 11:35 Missing Pieces in Implementation
• 13:20 Strategic Thinking and Wardley Maps
• 17:07 Key Takeaways and Conclusions

Transcript

Peter: 0:04 Welcome to Definitely Maybe Agile, the podcast where Peter Maddison and David Sharrock discuss the complexities of adopting new ways of working at scale. Hello Dave, how are you today?

Dave: 0:14 Peter, doing very well. Good to catch up and have a bit of a conversation.

Peter: 0:18 We had a plan, right?

Dave: 0:19 So we're going to tackle a small topic today—MVPs, lean startup—and see where that one goes.

Peter: 0:25 Yeah, we started there when we were warming up for this conversation and we landed on the rather famous phrase of "all models are wrong and some are useful" by George Box, who was a British statistician. I always like to pull that one out, especially when I run into over-application of a particular model without necessarily thinking—is this the right place to apply this model? Is there a better way to do this?

Dave: 0:54 Well, it's interesting. I find that we use models a lot. We talk about different models—I was going to say flavor of the month type of model—but we use models a lot in our work because, while the quote is true that they're all wrong, they can be valuable at a particular point. I think one of the things that's probably pretty clear is that, for the most part, we don't have one or two favorite models. We're just hungry for different models in different scenarios.

Peter: 1:23 Yeah, for sure. Part of the reason is we use models because they're simplifications of the complex world that we live in. As a consequence of being simplifications, you're leaving out some information. It's just the nature of it being a model. That's what it is, and yeah, that means that's where the original statement came from.

Dave: 1:47 It was like the simplification, so therefore it can never be perfectly accurate to represent the system. And I find that part of where this came from is both of us are beginning to see comments around the lean startup being passé. It's something that people want to move away from. Certainly both of us have spent, I would say, quite a bit of our career talking and using the lean startup in very specific contexts. Neither of us, I think, are fully bought into the idea that somehow we need to throw away this concept of the lean startup, minimum viable product, various other concepts that come with that, just because certain things have changed. Coding is much easier today than it was 15 years ago.

Peter: 2:28 Yeah, I think perhaps some of the problems that people have encountered when following that methodology may be that they have not necessarily adopted everything that's in there. But one of the common comments that I have seen on it is that incremental improvement from a base—like I create my MVP, I put it out there, I get feedback, I increment from there and I continue to increment from there to improve—where people are saying, well, that's not enough. That's incomplete. It's not enough. I have these step changes where something major happens or I realize this isn't the right direction. I need to pivot. And if I'm continuing just to increment in the same direction, I never talk about that piece as well.

Dave: 3:12 So there's a couple of things that jump immediately to mind. One of them being, as you're describing it in that way, I know early on in the implementation of the Lean Startup and minimum viable products, it was actually about those pivots, those significant shifts away from the direction. In fact, Eric Ries talks about how they used it to pivot away from a gaming platform to one that's based around the avatars, chat, various things like that. So there are significant pivots kind of embedded into how minimum viable products were used in the early days, I think. At the same time, now, more and more it's used as a "is this feature going to work?" So let's validate feature by feature or small bits of functionality over time, which is much more in terms of that incremental approach.

Peter: 3:58 Yeah. So there's this piece right there where we need to be considering what is the purpose of this next thing that I'm going to build? But also, am I every once in a while going back and checking that this is taking me in the right direction at all? Do I need to make a major change in a different direction? And then there's lots of examples of this in terms of understanding, well, what's coming down the pipe that might cause me to need to make a major change into the way that I do things. Am I entirely in the wrong space, wrong business, that I need to rethink entirely what it is that we do and take this somewhere else?

Dave: 4:37 Well, as you're describing that, it's making me think there's a big difference between strategy and making predictions based on where we are right now and what we have around us, and that sort of strategic thinking of "is there an opportunity for us to angle towards various things like that?" And I don't think we think of Lean Startup as a strategy.

Peter: 4:57 No, no, we think of it as an operational effectiveness tool. It's like, how do we effectively go and find the direction? How do we effectively learn the next thing that we need to do to understand how we move forward? And that idea of creating a system which is continually learning, continually looking at what is the next thing I should try. That makes a lot of sense.

Dave: 5:28 For sure. Now what would you say you have seen change in recent years around lean startup and minimum viable product?

Peter: 5:37 Well, one thing, for sure, especially in the last few years, it's become much, much easier to try something, to create a new direction or idea. There's a myriad of AI tools out there. Take something like Replit or something like that, where you could just say, "Hey, I want to try this idea, build it." It'll come up, it'll set all different agents up for you and it'll create something which you can then even push out on that platform to try it out and see if this works, if this does what I wanted to do. So experimentation becomes much easier. That's at a small scale, and in large organizations I can start to do similar pieces. I have less reliance on waiting for an engineering team to be available to build this thing for me so that I can then go and try and see if this is the right direction to recommend that we take things.

Dave: 6:27 So what I hear you saying—I think of it as commoditization of the actual development of an idea, which, again, if you think back to minimum viable product concepts, we would often use paper prototypes, Figma mockups, various things like that to validate ideas, because they were cheap and easy to pull together and the development of those apps or services, whatever it might be, was more difficult. Well, that's getting easier and easier to do, so I don't necessarily need to use a Figma mock-up. It's nearly the same amount of effort for me to just build an app that uses that Figma design and actually interact with it in a more meaningful way. That's the commoditization of development, which really just pushes the decision further along the line. You're closer to a working product, or at least a potentially working product, let me say it that way, than you are when you're looking at Figma designs.

Peter: 7:22 Now, of course, Figma would be very upset to hear you not talking about their Dev AI capabilities, which will take that design and automatically create the solution based on that design.

Dave: 7:34 I said it was easy! laughs

Peter: 7:36 You can even do it inside Figma, of course, right?

Dave: 7:41 The interesting thing about that is I don't think that invalidates the lean startup, minimum viable product. It makes it easier, it makes it more accessible to many more people. So people who are trying to get startups off the ground can do these things much more quickly and get perhaps more meaningful feedback more quickly. But to your point, it's very much along that incremental path. How do I make something easier and easier to use?

Peter: 8:10 Yeah, and actually, in that ease of use is an interesting piece too, because one of the things I have noticed with these prototypes is that they don't necessarily produce good user experience, amongst all of the other things that they're doing. I mean the very simple pieces—oh, it's a website form or something that you fill in—yeah, okay, that's fairly straightforward. But when you start getting into more complex systems, more complex interfaces, they run into some of the same challenges. Thinking about who's going to be using this, how are they going to use it, which parts of this are most critical—these are still elements that you need to do. And you can go ask another AI or see how that might work, but it gets circular at some point, doesn't it?

Dave: 8:56 It does.

Dave: 8:58 I just wanted to touch on—I think there's another use case, perhaps, around lean startup, which is misuse of minimum viable products, lean startup and that pivoting thing. We've seen that. We have talked about it before in terms of people talk about pivoting all the time. They talk about minimum viable products which are months and months in preparation, which goes against this concept of minimum viable product. But I think one of the things that I don't think you can talk about the impact of the lean startup without recognizing that, like many great ideas, they get picked up and then just not followed through to its entirety. It's very easy to do the stuff that we like, but not do the stuff that takes more discipline, that requires a little bit more thought and more effort. And therefore we're doing something that we're going to call the lean startup, but we're not really understanding or following through on the pieces that make it work. And then some of that I think is about being available to have the feedback.

Peter: 9:58 So your feedback loops actually creating change in the product to actually improve the product. And if that feedback loop is incomplete or you don't have a solid way of looking at that and then building out in the right directions—and this I think goes back to the other piece—don't just focus on the feedback, because if I ask my customers they want faster horses. I also need to be looking at where is the car? At what point do I need to build a car? Right, it's the conceptual idea.

Dave: 10:29 I'm just even thinking, are we even asking the right people? Are we getting the opinion or the behavior of a cohort that isn't our target customer, isn't going to have, doesn't have the problem we're particularly aiming to solve?

Peter: 10:44 That's the Nike problem we talked about a few episodes back, right? If you only listen to the noisiest customers, and the noisiest customers are going to be the ones who have the biggest complaints, but they could very well—you could end up going down to a niche and suddenly finding that, well, you don't actually have an audience anymore and you shut down all the pieces that you really needed.

Dave: 11:17 So yeah. Now, if we—as I think we're kind of not buying into the lean startup no longer being relevant, it's a model that has its place. You also talked about it being a really good tool, model for incremental change, for incremental learning, small changes as we kind of direct towards a desired outcome, which I think we're on the same page there. Like big step changes. Where do you look for that? Anything that jumps to mind?

Peter: 11:36 And we touched on this around the strategy piece of it. From a leadership perspective, are we re-looking at our strategy? Are we making sure that on a regular basis, we're going back and we're observing what's happening outside as well as inside the organization to understand what might need to change? Are we still going in the right direction? If we continue down this path, is it going to take us towards the outcomes we're looking for? Do we need to make changes?

Peter: 12:03 And those step changes can come in lots of different forms and ways. Of course, the problem often tends to be once we have set off in a particular direction, are we flexible enough to be able to change to a new direction, and do we have mechanisms in place to allow that to happen at the right points in time? Because sometimes it can be—well, we set off in a particular direction to learn something, but we need some time to learn whether this was the right thing to do or not.

Dave: 12:43 That sounds like you're dovetailing into, say, business model canvas coupled with lean startup, and now, all of a sudden, you're back at the beginning of that incremental learning. Something has to be a trigger. There has to be a trigger.

Peter: 12:46 That's a bit tricky, yeah.

Dave: 12:58 So I mean we're in a world now where the natural response in a lot of cases is to go and ask large language models and start really trying to fine-tune where that might go. I'd say we have had some mixed results with that one, mainly because one of the interesting things from a strategic perspective is to look at the extremes, the outliers, and what you often end up with—at least if you're not forming the prompts in a smart way—you end up in the middle ground rather than looking at the extremes. So you're missing these key opportunities. And one I still come back to, a tool that I think hasn't had enough recognition, and that's Wardley mapping for shifting the thinking. I find the construction of Wardley maps to be convoluted. It's a lot of effort to go and really think it through for your particular area. However, the thinking, the understanding that comes from that about where opportunities are coming from, about things which are really out of view, outside the radar of an organization, but could really come in and impact their area—that domain is so interesting. And so I still kind of, when I think of that step change perspective, I kind of come back to probably my favorite tool there. You know, that doesn't work everywhere, like we have just said, but that favorite model is the Wardley mapping structure.

Peter: 14:07 Yeah, I agree it is very valuable. I have used it on a number of occasions, even in a group setting, although it is somewhat of an individual sort of exercise to put it together. But in a group setting can be valuable just to help, especially a senior leadership team, understand the commoditization of elements of their delivery system. Like here's my value chain. Here are the elements of my value chain. They are slowly commoditized over time. So now, if we start to understand that, what's the impact on the system? Where do we need to focus our energy?

Dave: 14:40 I mean all the noise about how various types of work are becoming open to attack or replacement through agentic AI and various other generative AI solutions. When you start thinking about the pace of that change, I feel that Wardley mapping really helps you kind of tease out some of those areas and look at what happens if you really do commoditize in these areas where we have got a lot of investment as an organization.

Peter: 15:11 Yes, yeah, it's the commoditization of knowledge work and delivery work is essentially what that targets and it has a massive effect. Another space where I have seen impact on what you were describing—means you are going to miss out on what are the consequences of it, because it's also very positive generally.

Dave: 15:51 It's like, no, that's a great idea.

Peter: 15:52 You totally should just upgrade every single one of your systems overnight.

Dave: 15:56 I thought that was just me, that it thought was a really, really smart individual coming up with these ideas. Sorry to burst your bubble. laughs How do we wrap things up?

Peter: 16:06 Well, I can come up with a couple of points. I think one of the main ones is I think it's very true that all models are wrong, but some are useful, and we have talked about some useful models. Lean Startup is still a useful model, but it is important that it doesn't exist in isolation and that it's used well in the organization where it's most applicable. And the tools that it gives you, like the business model canvas and a lot of the other pieces, are very valuable, very useful, can help you think through problems, so use them. Wardley Maps is another key model that is also very, very useful and can help you identify where a strategy might be and point you in the right direction, especially if you marry it with all of the other pieces that fit into that. Spend some time learning about it, and Simon Wardley has put a whole ton of great posts on Medium that you can use for that.

Peter: 17:02 What would you add to that?

Dave: 17:04 So, as you're describing it, here's what strikes me really clearly—you need multiple models, and what we have talked about really is the concept of I need continuous incremental change, innovation, testing, hypothesis testing, and so on. But you also need to be looking at those sort of bigger step changes. So in that particular context, I think at different times, organizations are going to focus more on one than the other, but the reality is they need both capabilities within their organization.

Dave: 17:32 I think that really, you know, that concept of balance and we have often talked about more of one thing, less of another, and so on. How do you get that balance? The other thing that struck me as we're going through—and I think we have talked a lot about AI and it popping up—and what strikes me in this conversation is how pervasive it is in terms of where to go for ideas and how to work with it. And it's really peppered through our whole conversation without it being a "we're just going to talk about AI" thing, which I think just shows how rapidly we're shifting into that sort of paradigm where it's at the table and driving a lot of our conversations and provides a lot of commoditization in certain areas.

Peter: 18:24 Yeah, it's another tool in the toolbox and another critical part of practically any system that you're going to encounter these days. Awesome, well, thank you, as always, Dave. Always good to have these conversations and I look forward to the next one.

Dave: 18:36 Talk to you soon.

Peter: 18:38 You've been listening to Definitely Maybe Agile, the podcast where your hosts, Peter Maddison and David Sharrock, focus on the art and science of digital, agile and DevOps at scale.