How business-model-first thinking can slow consumer innovation

One of the first questions people ask me about Roomix, the real estate search engine we are building, is:

“So... what is the business model?”

Reasonable question. Companies need money. Funds need returns. Founders need rent.

The problem is not the question. The problem is when it shows up.

If it is the first lens on a new idea, it quietly starts steering the product. The conversation jumps from:

“What insanely useful thing could this become?”

to:

“What version of this is easiest to explain on a slide?”

At that point the product stops being an exploration and starts behaving like a spreadsheet with a UI.

I do not think business models are evil. I do not think founders should “just build cool stuff” and hope money magically appears. If you never think about where value might accrue, or how distribution might work, you are just being irresponsible.

What I think kills a lot of consumer innovation, especially in AI, is front loading the business model: forcing a neat, specific monetization story before you have any real signal on what people actually care about. The product shrinks to fit the slide, instead of the slide growing to fit the product.

The business model reflex

The default AI startup template looks like this:

find a workflow → add AI → wrap it in SaaS → sell to businesses → show a path to $10M ARR

A lot of good companies come out of this pattern. They solve real problems and make real money. Some of my favourite products are “just” workflow tools that quietly make teams 10x more effective.

Some of the most important AI infrastructure will probably look like classic B2B software from the outside. That is not the enemy here.

The issue is when this template becomes the only acceptable starting point. At that point, very few products feel genuinely new.

The ecosystem nudges everyone toward the same safe shape:

• predictable pricing
• predictable sales cycles
• predictable margins

Great for fund math. Not great for weird, ambitious ideas.

It is simply easier to raise money if you have an enterprise story and a neat go to market slide. So founders design for that from day one. The product is built to fit procurement, not to stretch what people think is possible.

On the consumer side, the same reflex shows up. Everything wants a subscription and an LTV chart. The slightly crazy “this might be huge or might be nothing” ideas die early, not because they are wrong, but because they are hard to underwrite.

The nuance here is important: it is one thing to have a few hypotheses about where the money might come from later. It is another to lock the company into a specific model on day zero and treat the product as a proof-of-concept for that model.

B2B and workflow products are often the most straightforward way to turn real problems into real businesses. The issue is monoculture: when every new product feels pressured to dress up as an AI workflow SaaS company just to be taken seriously, a lot of stranger, more ambitious consumer ideas never make it out of the sketchpad.

Paul Graham, Google and getting upwind of money

This is where I keep coming back to Paul Graham.

YC’s motto is “Make something people want.” It sounds soft. It is actually aggressive.

In one essay he talks about getting "upwind" of revenue. If you are upwind in sailing, you choose when and how to engage. For startups that means: if you sit upstream of where money flows, you have options.

Craigslist is his classic example. It sits "upwind of enormous revenues." Once you are there, figuring out how to make money is not the hard part. The hard part was becoming the default place people go.

That is the real meaning of "make something people want". It is not a motivational poster. It is a strategy: earn power over your own business model.

Google’s early years show the same pattern.

For a while, Google Search looked more like a research project than a company. It was just a shockingly good way to find things on the web. Fast, relevant, clean. No ads on the results page at first. AdWords only arrived later, and it started small.

Roughly in order:

1. Build one of the best search engines on the planet.
2. Earn massive user trust.
3. Then add a money machine that does not destroy that trust.

Even “Do not be evil” worked as a constraint on monetization. It ruled out the worst tricks that would poison search. That discipline is part of why the ad business worked so well.

It is also true that for every Google, there are thousands of “people loved it but it never made money” stories. So the lesson is not “ignore revenue entirely and hope.” The lesson is: treat business models as hypotheses and constraints, and give yourself enough room to discover a product people really care about before you harden those choices.

Now compare that to the common sequence today:

1. Start with a monetization story that looks tidy in a deck.
2. Wrap a product around it.
3. Hope people care enough to use it.

You get a lot of reasonable businesses. You do not get the next Google.

The structural bias for legibility

I am not anti-VC. Some investors are still fundamentally driven by paradigm-shifting risk, rather than simple pattern-matching.

Founders, Inc is a good example. They actively hunt for crazy ideas and let people build before everything fits neatly in a model. That feels closer to the original startup game: take a serious swing at something that might not look like common sense until it suddenly looks inevitable.

But if you look across most early stage funds right now, the pattern is to converge on the safe bets.

The default checklist is:

• AI
• B2B
• SaaS
• Clear path to ARR

To be clear: this behavior is rational. This combination has produced a lot of real, durable companies.

The bias towards this shape is not necessarily about individual investors lacking imagination. It is structural.

Large funds need to deploy large amounts of capital on a predictable cadence. They need ownership in companies that can plausibly absorb tens of millions of dollars. Crucially, they need stories that can be summarized in a few lines on a quarterly LP call.

The easiest way to satisfy those constraints is to back things that look like previous winners: enterprise SaaS with a repeatable sales motion, clear unit economics, and a familiar pricing page.

Legibility becomes a filter.

Anything that does not obviously map to "we have seen this movie before" gets scored down, not because it is a bad idea, but because it is harder to slot into the fund’s model and the LP narrative that sits behind it.

You can see the contrast in the few products that actually feel new.

Perplexity is one of the first search products in a long time that genuinely changes how you look for information, not just how the results page is styled. Arc from The Browser Company is one of the first browsers in years that treats the browser itself as a creative surface.

Neither of them started life as a tidy AI workflow SaaS story. Their early backers included a mix of funds, operators, and angels who were willing to lean into the "weird" version before it was obvious.

That is the interesting dynamic for me:

• Capital for ambitious consumer products still exists.
• It is just no longer the default option at seed.

Most early stage VCs are structurally incentivized to optimize for legibility. That makes sense for their business model. But the irony is that the biggest upside in their portfolio will almost certainly come from the thing that did not look like a safe B2B SaaS deal at the beginning.

If I were on the other side of the table, that is the question I would keep asking myself:

What part of our fund is explicitly reserved for the next Perplexity or Arc, rather than just the next workflow tool?

As a founder, my job is different. I have to hold onto the version of the product that users will love before it is fully legible to a model. The best investors I have talked to actually like that tension. They know that if every company in their portfolio looks safe on day one, they have probably already priced away most of their upside.

What we are trying to do with Roomix

This is the tension I feel every day with Roomix.

We are not trying to build a slightly nicer portal for one country. I want Roomix to be the search engine that sets the standard for housing across the world.

We launched first in Argentina, where tens of thousands of people already start their housing search on Roomix every month. It’s still tiny compared to where we want to go, but it’s enough to see how broken the old model feels once you try something different.

If you zoom out, the last few decades of real estate search are almost embarrassing. We went from circling apartments in a newspaper, to clicking the same apartments on .com websites.

Same listing mindset. Same messy photos. Same guessing game. Only the paper turned into a screen.

Most “innovation” since then has been more filters and more ads.

I am not interested in polishing that model. I want to throw it out.

We are trying to build a real estate search engine that behaves like an actual search engine, not a pay to play directory. Less “here are 300 random boxes, good luck”. More “here are the few homes that actually fit your life, and here is why”.

The questions I care about now are not financial:

• Can we help someone find a place they love faster and with less stress?
• Does it feel like Roomix is on the side of the person searching, not whoever pays us most?
• Can it become the default tab people open when they start a housing search, in any city?

If we can do that, I am not worried about the business model. Real estate has plenty of ways to make money.

We already have a short list of paths that could work, some obvious, some less obvious, and we’re quietly testing a few of them in the background. But right now they are constraints, not the goal. The job today is to earn the right to pick a business model, not to pretend we already know the perfect one.

The point is not that money does not matter. The point is, if we force Roomix to look like a neat SaaS business from day one, we will slowly turn it into yet another real estate portal.

And the world does not need another portal that treats users as leads and search results as inventory.

My own test is simple: If it is so easy to slap a subscription on real estate and print money, why has nobody already done it.

Proptech is full of smart, well funded teams. If the obvious SaaS play were enough to reshape the industry, you would feel it already.

You do not disrupt something as old and messy as housing with a pricing page and a couple of SDRs. You do it by making the experience for actual humans unreasonably better, then figuring out how to capture some of that value without breaking what made it great.

We’re starting in Argentina and Latin America, but I don’t think in terms of “regions”, I think in terms of standards. I want someone in any major city to open their laptop and think: of course I start my search on Roomix.

Conclusion

I am not saying we should never talk about business models.

I am saying it should not be the first question.

Instead of opening with: “What is the business model?” start with:

“What makes this ten times better than what people were doing before?”

If a founder cannot answer that, the monetization plan does not matter.

Business models are important. They just belong downstream of products people care about.

For us, the order is simple:

1. Be unreasonably helpful to people looking for a place to live.
2. Earn their trust.
3. Choose a business model that does not betray that trust.

Once you have something people truly care about, then it is time to get very specific about how the money flows, who pays, and why that is sustainable.

If that is naive, I can live with it.

History is full of “naive” products that changed the paradigm, and only later figured out how to make the spreadsheets smile.