Hacker News Reader

I came here to say this. The author of the article does not understand what an MVP is.

An MVP is the eventual answer to a metric fuckton of customer development, not what you go into customer development with.

If you think you already know what the customer wants, I wish you luck at your next startup after this one fails; you don’t know that, and neither does your customer, or they would have built / bought it already. It’s your job to find the real problems they’re facing, and then build a solution that makes them rip it out of your hands before it’s even done. then, iterate.

With Ansible, Mike already had a successful prototype (func) which had been developed and used by the Fedora infrastructure team for a while. He then left Red Hat and worked on Ansible. So he did in fact already know what the customer wanted! However that's an unusual case for a start-up. https://opensource.com/article/21/2/ansible-origin-story

I think in that case you're not building a minimum viable product because that's already happened. The MVP was what Ansible was when the Fedora team were using it. When Mike left to form a company around the product it was much more than that - because there was already a really good understanding of what people would pay for.

Mike is also a great developer and infrastructure person so in a sense he is his own customer.

Based on my experience func and ansible aren't very comparable and it's hard to call func a prototype for ansible. My recollection is that it was much more of a disruption to puppet, chef, and saltstack that had all become much more complicated and heavyweight by that time.

What I took away from this article is that an expert in that space (DeHaan) knew there was a gap and opportunity in this area, and the point of this article is that because he was able to capitalize on that opportunity because of his existing expertise and familiarity.

It's an interesting view compared to some interpretations of lean startups.

"The author of the article does not understand what an MVP is" - it feels like that's not the case, based on their background? Reading the comments here, I see a lot of people with their own (and conflicting) ideas of what MVP is in the context of a startup. I also didn't get the idea from reading the article you are disagreeing with their point, and that they weren't opposing iterative learning.

Some startups are created by experienced customers, and in these cases you often absolutely do know what the customer wants. Empirically, these can succeed without any significant customer development. Customer development becomes much more important if the startup lacks real domain and market expertise.

Also, while customer development might work well for web apps and such, there are classes of software where customers have no desire to interact with you until the product is almost fully baked. Trying to do customer development in these markets is often an exercise in futility. It is a different kind of process to arrive at an MVP, and in these cases it requires that the founders have strong domain and market expertise.

Customer development is always important. What constitutes an MVP may change, based on the industry or the customer, but the concept of talking to your customers regularly, understanding their problems, and iterating quickly doesn’t change.

As much as a founder would like to say they are the customer… they aren’t. They are a customer, and finding the commonalities amongst potential customers so you can build something they want is important. The founder being a customer just kickstarts that process and means the founder can ask better (aka “the right”) questions.

Which markets are you referring to? Gov requires customer dev, which is why there are often grants. Med requires customer dev, as building simply a better device rarely wins - it needs to be much better than existing solutions, and the only way you get there is by finding out what the problems are with existing solutions. Same goes for banking, etc

Domain and market expertise is always a benefit, but rarely a strict requirement for success. The ability to listen carefully, and iterate quickly, is much more important.

Agreed, super super important to do customer dev.

What I haven't seen mentioned here is that if you do customer dev right, you get a leg up on sales.

One of the slickest founders I know would get conversations with his target customers to ask about their jobs + workflows. He'd learn from everyone but he kept note of the people who shared the same problem he was targeting. He now has a list of dozens of people who he has a relationship with who would make great first customers.

He didn't need to build a landing page or even a prototype in figma because he was just talking to people to learn about what they do.

I've worked on products where customer development was essential and also on products where it was largely useless for product development. You are assuming things about the product development process and customer interactions that are not always true. That doesn't imply that you don't talk to customers or don't understand their problems, just that this can be disconnected, often necessarily, from the product being developed.

As an example, hard tech products in data infrastructure often have this property. Famously, there are no incremental product outputs to show customers (or investors) as a matter of fundamental engineering; the product has to be almost fully baked before it can demonstrate anything of value at all. Customers have a difficult time understanding the implications of fundamentally new capabilities until they have time to play with them; they provide no useful feedback until the product has already reached a stage where this is possible (read: almost fully baked). And of course, there is no "iterate quickly" because the product value follows from architecture; if the architecture isn't correct the first time then iterating on the product means starting over from scratch. Consequently, the only people that can successfully build these products are people with serious expertise in the product and its use case. Even if you try to do customer development, there is nothing for the customer to engage with until you are well past the point where their feedback can materially change the product.

In these cases, no amounts of customer development will make up for a lack of deep domain and market expertise. The lean startup/MVP/customer development etc literature tacitly assumes that every product is an "app" of some type. If you work on software products that are not "apps" then you may find some of the advice to be misguided or useless.

See, I’m not convinced that’s true. Having built hard tech products before, there are plenty of places for iteration.

The catch is that you iterate the plan and architecture, before you write any significant code. In this case, it’s even more important to listen carefully to the customer and pay attention to the underlying problem.

For example, it may not have made sense to build Docker because “everyone uses VMs” and Vagrant was already a thing. But when talking to customers, they would have described ample problems with the existing approach - if you carefully listened, you could surmise that something like basic Linux containers would help, and you could provide a basic prototype of that to start. When they complain it lacks features, that’s okay, if and only if they still want to use it - THAT is the sign of PMF.

Want to build a new data warehouse? That’s hard, I agree - but the important part isn’t that it’s hard but why it’s better which is going to come from customer development. If listening to customers isn’t driving your product roadmap, you are literally guessing at what they need, and unless you’re Steve Jobs and your next product is the iPhone, you’re in for a rough time.

(And even the iPhone had plenty of customer dev; they interviewed thousands of people about their cell phones and surmised the solution wasn’t simply to build a better keyboard for a Palm Pilot, but to build a brand new device entirely, because people loved their phones but they had limited use. The Newton, on the other hand, had no market because people weren’t ready for it and it was built in a lab with no customers to talk to.)

You are inadvertently reinforcing my point.

Customers do not have a mental model of the very complex theory that underpins the design of new kinds of data warehouses; they routinely ask for features and capabilities that are mutually exclusive or violate the laws of physics or computer science theorems if you ask them. There have been many attempts to "customer develop" a new kind of data warehouse, and the result is invariably naive architecture with too many deficiencies to be successful (basically the proverbial faster horse instead of a car).

In practice, the person building the product is usually a deep expert on the use cases for the product they are building in cases like new kinds of data warehouses. Indeed, there is the opposite problem that it is too customer and use case driven. Most such products fail due to insufficient expertise in the technical execution because the product was built by an expert user instead of an expert in the theory of building those types of systems, without a good understanding of the many subtle tradeoffs. If someone need to do customer development to understand why a new kind of data warehouse is better then they lack the deep technical expertise required execute ipso facto, since it trivially follows from low-level architecture.

I think you’re missing mine. You don’t listen to the users’ feature requests; you listen to their problems and then devise solutions that solve a significant cross section of user issues.

Agreed.

The management of requirements in building anything needs to be structured properly to manage client expectations because it's often the first point of failure on development projects. MVP is a concept that is implemented on mission critical projects in order to insure a minimally acceptable point of project success, not to "be" project success... Anything else that can be added (within the implementation time frame) after base goals are met is a bonus.

Scope and schedule creep are 2 of the biggest threats to any major project, and it takes technical oversight just as much as managerial oversight... Technical Architects usually bridge this gap on projects.

An MVP is meant to be something that is generalized enough to fit KEY customer needs in a limited time frame, not ALL of the customer's needs, and customers inherently try to squeeze changes in during execution in large-scale projects. MVP for small-scale and startup projects may not mesh well, unless needs are well defined by an accountable customer.

An MVP is a contractual deliverable, so it shouldn't be delivered or even started before that contract is properly written and signed off.

Key aspects of a successful MVP -

It needs to be scale-able and mindful of future integrations and functionality.

It needs to be able to pivot and adjust to change, especially changes outside of your control (e.g. The Pandemic).

The data therein needs to maintain integrity and transfer-ability.

This is why experienced IT/Business architects are necessary on key projects. They incorporate business drivers, requirements, project management objectives, and technical teams into meaningful language, documentation, and project outcomes.

When you struggle on without an architect, projects suffer from a lack of direction, mixed communication, and often waste lots of money on back tracking solutions.

MVP is also a "version control" and project/client management concept/methodology, not a development language. The author broke down the title based on the words used. Taking time out to really investigate an understand why so many credible people use the term is a better move than declaring it obsolete.

It seems like a lot of YC companies are just drinking the MVP kool-aid. And there are 400 companies per batch now because they can just play the numbers game: throw them all at the wall and see who sticks.

It's a very interesting and nuanced distinction between:

"just build something and launch it as soon as possible"

...vs...

"think really hard about how to solve a particular problem in a novel way, then build the simplest possible solution that lets you iterate, but no simpler"

The most apt analogy for me has been multivariate optimization of a nonconvex function: the MVP is the initialization conditions. If you start near a local maximum then there is no way that hill-climbing (SGD, L-BFGS) will get you to a global maximum. So you need to make sure your starting point (MVP) lets you actually iterate to the global maximum through small local optimizations.

A product pivot is basically a random jump in the objective function hoping to get closer to a point where the function's gradient can lead you to a global maximum.

We did that with our product and thankfully the new initialization conditions are way better, and we are at a point where we can iterate to PMF. But there was no way we could have done that with our old product. (Here's the story of that pivot: https://parsnip.substack.com/p/a-new-hope)

The MVP strategy is entirely about validating your product. Unless you decide it has a good market fit, there is nothing in it telling you how to look for alternatives. There's nothing telling you to go by hill-climbing, or even by any gradient based search at all.

What it says is that once you find product-market fit, you don't destroy it by jumping around at random. But at this point, it's about disruption and the Innovator's Dilemma, where you are the incumbent. It's not about starting-up anymore.

This is a great analogy, basically plotting startup success on an evolutionary NK landscape [0]. Finding optima rests on a healthy balance of exploration and exploitation, as in the behavior of ant colonies searching for food. Apple has been a good historical model of this in business, though for a startup it's often necessary to keep the chips all in on one product. Ideally though, a pivot isn't a random jump, but is rather guided by an incomplete map of the landscape. Herein lies the value of the MVP, but it's curse is being difficult to throw away, especially once it has some traction.

[0] https://en.wikipedia.org/wiki/NK_model

The other big reason for an MVP is to get from 'customers told us they want this' to 'customers are actually paying for this' as quickly as possible. All to often potential customers say they want X, but really have no idea what they want. The only surefire way to know if a product is on the right path is to have customers using and paying for the product.

All to often potential customers say they want X, but really have no idea what they want.

Or they actually do really want and need X _right now_. However when you come back with X in 18 month then the customer has already solved X themselves, found another company that can do X or don't really need X any more.

It’s quite rare that a customer knows exactly what they need; often, if they do, they build it.

The wins happen when a customer knows what problems they have, and you can present a solution that solves them better than they could otherwise get elsewhere.

Otherwise, as you say, they will build it or buy it elsewhere.

I think there's some degree of nuance that isn't captured with comments like yours. I agree with your comment in principle, but there are interpretations of this method that are too extreme and that's where most of the bad MVPisms come from (and what I assume this blog post is a response to). As with any process, what most people are doing is not what was intended, and so we have to discuss what people are doing more so than the theory.

Data is only as valuable as the insight, understanding and intuition at the disposal of those gathering and actioning the data. There are organisations that truly embrace this idea of "we know nothing, the data will guide us" and they get lost because they're chasing a bunch of numbers for the sake of numbers on weak experiments that don't marry up to anything of meaning.

You need some amount of difficult-to-define "understanding" and "intuition" to be able to turn any learnings (whether gathered from experience or data) into a product: without it, you'll fail, regardless of process. There are many successful products that started out as someone absolutely committed to a belief in what is needed by the market, there are many successful products that were developed by following the data from day one, and likewise there are many unsuccessful products following both patterns.

My personal disdain for MVP-ing is exactly because it is so difficult to get right, and is very easy to get wrong, and it leaves little room for corrective action: I'm sure we all have experience with zombie startups, years old, still frantically "MVPing" everything under the sun. Personally, I'd much rather work within an organisation that has a clear vision about the problem they're trying to solve with flexibility around how it is to be solved, even if it ultimately leads to failure.

Customers lie, but so does data.

it leaves little room for corrective action

Building an MVP means getting something in front of customers as soon as possible, in order to test your product works for customers and change it if you're wrong. The entire point of an MVP is to give you as much time to change things as possible.

Everything you do prior to putting your product in front of someone might be wrong. The only validation that counts is the customer handing you their money. The longer you spend building before getting something out there the more expensive any mistakes might be. Consequently, making an MVP and only building the absolute minimum viable solution to a problem, give you as much correction time as possible.

I'd much rather work within an organisation that has a clear vision about the problem they're trying to solve with flexibility around how it is to be solved

We all would. The question is about how the org goes about getting clarity about the problem. If they think they fully understand it without talking to customers or putting something in front of customers in order to give the customer something to talk about, run the hell away. Those discussions will happen as soon as the customer sees something. If you've built a full product that you think is ideal, and then the customer doesn't agree, then you either lose the customer or you have to do a ton of work to change the product. Those are expensive mistakes that kill a startup. Having those conversations as early as possible is how you avoid that cost.

> The only validation that counts is the customer handing you their money.

TLDR; of this whole thread.

I wrote something similar in a direct response to your first comment while you were writing this one :)

When I say there's little room for corrective action, I am specifically speaking about the choice to pursue the MVP strategy. A person who starts a company and decides to embrace the MVP strategy is stuck on a hamster wheel: if they've spent a year and not validated anything positively, they cannot "just" change strategy because the organisation has been built atop of the idea of small ideas, not big ideas.

My experience is that most businesses that pursue the MVP strategy never leave the MVP stage, regardless of the viability of the thing that originally inspired them, because they believe to MVP properly is to surrender all creativity, vision and intuition to numbers, which is in turn to surrender what's important when building anything.

No matter what strategy you choose, there has to be a meaningful amount of intuition and understanding to be able to make decisions: a purely data driven approach (with arbitrary goals like "someone paid us") is predicated on product development being entirely quantitative. Customers lie not just in what they say but also in what they do: getting a customer to pay is very easy compared to, say, building a sustainable business. People would have paid for faster horses!

People who espouse the success of the MVP strategy in their own businesses are often people with a level of creativity/intuition/understanding that most people do not have. MVP is a powerful tool in the right hands, absolutely, and I don't doubt that you're one such person, but MVP as a strategy for "normal" people has so many foot guns that I would never recommend it over a grand vision strategy. How many people are capable of picking the right customers to talk to? How many people are capable of understanding customer behaviour? How many people are capable of balancing outlier vs. representative behaviour?

I believe that normal people will have a higher chance of success by pursuing a vision than they would by following customer behaviour from the start.

You're expecting normal people to successfully invent the car? I have much more faith in them forming a relationship with a paying customer, and offering a small process improvement through a MVP.

An MVP will also help with appropriately market timing of your product. If the client is asking for a faster horse, you might want to look at innovating the saddle or horseshoe first. You might be 1000 years too early with your car engine idea.

>My personal disdain for MVP-ing is [...], and it leaves little room for corrective action

The "corrective action" is actually embedded into the MVP / Lean Startup terminology popularized by Eric Reis: https://www.slideshare.net/startuplessonslearned/minimum-via...

It's just the that acronym wasn't longer such as MVPI (Minimum Viable Product Iteration) or MVPLL ((Minimum Viable Product Learning Loop).

>Customers lie, but so does data.

The "data" in the MVP iteration framework doesn't lie because the relevant data is "paying customers". Either the startup has growing revenue from its product or it doesn't.

> it leaves little room for corrective action

The entire idea of MVPs is to leave as much room for corrective action as possible.

It doesn't happen often, but when it happens it really annoying that society is so reluctant to embrace an idea that people won't even acknowledge that the idea exists. If you name it, they will change the word meaning to the opposite of the thing you named. If you show it by example, they will completely ignore your idea on the example, and only ever accept irrelevant details.

We are seeing an example here, with MVPs. People really dislike anything lean and anything scientific. Merge the two ideas and you have a sure mind-repellent.

In your story you thought you knew what to build, but didn't.

The "MVP" theory is that the MVP is part of validating that you really know what to build, as you say.

OP author disagrees. (I'm not sure OP author says anything about when you "start selling" exactly).

Clearly a lot more can be said about how you know what to build if not an "MVP" approach. OP suggests a few things: "cultivated instinct" (which as a phrase alone isn't too useful, how do you get it?); Spending a lot of time talking to potential customers, and not necessarily taking what they say at face value but "averaging" it, and mixing with that "cultivated instinct"; and maybe most specifically/operationalizable -- spend some time working in the industry/market before starting the company at all, so you've seen "how 5 other companies have built something and know all the things that they’ve done wrong, and it will be a lot easier."

"Domain expertise" might be the main thing he's talking about, arguing agianst the idea that you can start a company without somebody involved having significant domain expertise/experience in the industry, but instead develop that market knowledge as you go, through an "MVP" approach.

I'm curious how those ideas relate to your company, how much industry/domain experience you had before starting the company.

>> "Domain expertise" might be the main thing he's talking about

This is clearly what he is talking about. When he is talking about his experience with the "Senior Design project course in Computer Science at my alma mater" he is really talking about how people can "MVP their way into getting domain expertise" at some level. This is a fundamental problem with the "startup industry".

Groups of people from a computing (I want to be broader than CS) that have reasonably strong technical skills and a lot of enthusiasm for applying those skills want to build successful startups. They generally lack domain expertise outside of technology-adjacent products for the most part. It's not completely absent because there is some success, but there isn't a balance for the most part.

Where does the domain expertise come from then? It could come from a slow development process involving the founders immersing themselves in the area. That's really only something that's going to be feasible before the founders create the company. Once the company has been founded, it becomes a scramble to interpret some ideas and build an MVP that can be sold to some of the advisors that the founders could get time with. It's only now that the foundational work is happening. If the right people see the MVP, or the founders make the right connections, it's possible to catapult forward from here, but it's a predictable failure point.