A few years ago I got into a conversation with a venture capitalist who was advising a friend’s company. The pitch was for a piece of software that did one thing exceptionally well, for one specific industry. The total addressable market was real but capped — maybe a few hundred million dollars at saturation, which is a great business by any rational measure but unimpressive by the standards of people who get paid to swing for unicorns.
The VC’s feedback boiled down to a single question: can you make this generic? Could the product be reframed as a horizontal platform that served twenty industries instead of one? Because if it could, the market got 20x bigger overnight, and the math started to look interesting.
I’ve been thinking about that conversation for years. It explains almost everything I notice about the software industry — including why so much of the software businesses actually use is, frankly, terrible.
The tax of being generic
Here’s the trade. When you build software for a specific industry, you can encode that industry’s actual workflows in the product. You don’t have to abstract anything. The data model, the UI, the language, the integrations — they all match how the work actually happens.
When you build a generic platform that can be configured for any industry, you have to abstract everything. The data model becomes a soup of “records” and “fields” and “stages.” The UI becomes a kit of building blocks. The language becomes intentionally bland. And the work of bridging the abstraction back down to a specific industry’s reality gets pushed to the customer — usually as a six-month implementation by a partner network, or as a “customer success manager” whose actual job is to translate the generic platform into something that resembles the work the customer was already doing before they bought the software.
The customer pays for that translation. They pay in money (the implementation fee, the partner hours, the ongoing CSM relationship). They pay in time (the ramp before the team is productive again). And they pay in lost fidelity — the parts of their workflow that are too specific to encode in a generic data model just don’t get encoded, and the team has to remember them in their heads, which is the thing the software was supposed to fix in the first place.
Multiply this by the number of generic platforms a typical small business uses, and you understand why the SMB software market is the way it is: a kingdom of janky configurations, abandoned implementations, and quiet resentment.
The economics that VCs are right about
To be clear, the VC isn’t wrong about the math. If you’re trying to build a $10B company, you almost have to build something generic. The TAM has to be enormous. The product has to slot into a horizontal layer that exists across most businesses. The expansion motion has to be repeatable across industries.
So the venture-backed software industry is heavily biased toward generic platforms. That bias becomes a self-reinforcing prophecy: the talent flows toward the funded companies, the funded companies build generic platforms, the generic platforms become the default, and the niche products that should exist either never get built or get built by underfunded teams that struggle to compete on polish.
This is why so many specific industries — window tint shops, FBOs, marketing agencies that resell Vendasta, Part 142 flight training centers — run on a combination of generic CRMs, spreadsheets, and tribal knowledge. The software they actually need has been deemed too small a market to fund.
But “too small to fund” and “too small to be a great business” are very different things.
The unfair advantage of operators
The opportunity, if you look at it from a different angle, is enormous. A specific industry with a few thousand businesses in it, where every operator hates the generic software they use, will pay real money for a tool built specifically for them. They’ll pay quickly, because the value is obvious in the demo. They’ll churn rarely, because the alternative is a generic CRM with custom fields that they’d already given up on. And they’ll tell other operators in the same industry, because operators in tight verticals talk to each other constantly.
This is a real business. It’s just not a $10B business.
What makes it work is something most generalist software companies can’t copy: an operator who actually knows the industry from the inside. Someone who ran a tint shop, or managed an FBO, or sold Vendasta services through their own marketing agency. Someone who can encode the reality of the work because they’ve done the work.
That kind of founder doesn’t need market research to figure out what to build. They’ve been frustrated by the gap for years. The product exists in their head before they write a line of code. The roadmap is a list of grievances, in order of severity.
This is the unfair advantage. A generalist building software for window tint shops has to spend two years doing user research to learn what an operator already knows by heart. By the time they figure it out, the operator-founder has already shipped, found the first hundred customers, and locked in the product-market fit that most VC-backed software never achieves.
Why most companies don’t do it
So if niche-first software is such a great business, why isn’t everyone building it? A few reasons.
Founders are rarely operators. Most software companies are started by software people, not by operators of the industries the software is sold into. Software people are good at software, not at understanding the specific frustrations of running a tint shop, an FBO, or a window film install. The intuition isn’t transferable. You can hire a domain expert and listen carefully, but it’s not the same as having lived it.
Funding pressure forces generalization. The minute you take outside money — even a friendly seed round — the pressure to grow the TAM kicks in. The board asks how you’ll expand to adjacent industries. The investor expects a five-year plan that ends in a unicorn. The product gets gradually pulled toward the generic, because that’s the only path that justifies the funding model.
Early customer pain is hidden by patience. Operators are the most patient customers in the world. They’ve been duct-taping together generic tools for years. When they meet a niche product that’s even halfway decent, they’re so relieved they don’t complain about the rough edges. This makes it look, from the outside, like the niche product is doing fine — when in fact it could be doing much better, and the founder doesn’t realize the gap because the customers are too grateful to push.
The exit math is uglier. A $50M-revenue niche software company is a great business but a hard exit. There’s no obvious strategic acquirer in most niche industries. The PE roll-up math works, but it values you as a cash flow, not as a growth story. Founders raised on the venture exit playbook find this uninspiring even when the cash flow is real.
The case for doing it anyway
Despite all of that, niche-first software is the right move for a particular kind of company: one that doesn’t need to be a $10B business, that values customer love over TAM, and that has a founder with operator-level intuition for the industry it serves.
The customers you get are extraordinary. They use the product daily. They become evangelists. They send you screenshots of their workflows so you can understand them better. They renew without negotiation. They refer everyone they know. The CAC is mostly word-of-mouth. The product roadmap writes itself.
You also build something with permanent value. A horizontal platform competes with every other horizontal platform forever — the moats are shallow because the abstraction is the same everywhere. A niche-deep product, on the other hand, becomes the de facto operating system of an industry. Once you’re embedded in the daily work of every shop in your category, switching costs are enormous. Not because you locked anyone in, but because you became the way the work gets done.
That kind of moat isn’t bigger than a horizontal platform’s moat. It’s different — and for a small, deliberate company, it’s better.
What we’re building
Roffik is built on this thesis. We pick one specific industry at a time and build the software that industry actually needs — not the generic platform a horizontal company would tile over the top.
SalesThumb is the operating system for window tint, detailing, and PPF shops. HubWho is the business layer for marketing agencies that resell Vendasta — the invoicing, ACH, and KPI dashboards Vendasta itself doesn’t ship. Vizme.ai is the AI visualizer for window film and vinyl, covering auto, residential, and commercial flat glass. AviationAlley is the platform connecting general aviation businesses, including Part 142 training centers.
Each of these is its own product, built by people who have actually done the work. We’re small, deliberate, and not in a hurry to be everywhere. We’d rather serve five thousand customers in a vertical exceptionally than five hundred thousand customers in twenty verticals acceptably.
That’s the bet. If you’re an operator in any of those industries — or you’re thinking about building software for one — we’d love to compare notes.
