How I evaluate business ideas (after getting most of them wrong)
I built a digital marketing agency, a dozen products nobody wanted, and two SaaS businesses that actually worked. Here's the framework that separates the two categories.
By Alex Diaz
I’ve built a lot of things nobody wanted.
A digital marketing agency that made money but ate my life. Products I spent months on that launched to silence. Ideas I was convinced were brilliant — until I actually tried to find someone willing to pay for them.
Two things worked: MetricSpot, a small SEO tool that ran for a decade and funded my entire nomad life. And RevenueHunt, now a 7-figure SaaS. Everything else? Time I’m not getting back.
The difference between the two categories wasn’t talent or effort. It was asking the right questions before I started building.
Key takeaways:
- Existing demand + built-in distribution = business that works. Everything without those two things failed.
- The hierarchy: someone paying for a workaround > someone complaining online > someone saying “yeah I’d use that”
- If you can’t get in front of 100 buyers this week without spending money, you have a distribution problem
- Your product is not your moat — distribution, stored user value, and compounding data are
- If someone can get 80% of your product’s value from a ChatGPT prompt, you have a prompt wrapper, not a product
The pain nobody talks about
You have an idea. It feels right. You can see the product in your head. So you start building. Maybe you spend a weekend. Maybe a month. Maybe three months. You launch. And then… nothing. Or worse: a trickle of signups, nobody converts, and you’re staring at analytics wondering what went wrong.
The advice you’ll find online is useless. “Validate your idea.” How? “Talk to customers.” Which customers? “Find product-market fit.” That’s not advice, it’s a destination with no map.
VC-funded founders can survive bad ideas. They raise money, hire people, iterate for two years, and either find something or shut down and write a Medium post about “lessons learned.” Bootstrappers don’t have that luxury. You’re funding this yourself. Every month on the wrong idea is money out of your savings and time you’ll never recover.
The margin for error is zero. And yet most bootstrappers evaluate ideas the same way VCs do — gut feeling, a quick Google search, and blind optimism.
What I got wrong (and what finally worked)
My early failures all shared the same pattern: I fell in love with the solution and skipped the problem.
I’d think “wouldn’t it be cool if X existed?” and then spend months building X. The question I never asked: is anyone actively paying money to solve this problem right now?
Not “would people use this?” — that’s a trap. People say yes to everything in surveys. The question is: are they already spending money, time, or effort on workarounds? Because if they’re not, the pain isn’t real enough.
MetricSpot worked because website owners were already paying for SEO tools. I didn’t create demand. I entered a market where people were actively buying, and I built something simpler and free (with a premium tier). The distribution was baked in — users analyzed their site, got a branded report, shared it. The product spread itself.
RevenueHunt worked because Shopify store owners were already paying for product recommendation apps. The Shopify App Store was the distribution. Freemium was the growth engine. The “Powered by RevenueHunt” badge on every quiz was free advertising on thousands of stores.
The pattern: existing demand + built-in distribution = business that works.
Everything I built without those two things failed.
The questions that actually kill bad ideas
After enough expensive mistakes, I started keeping a list. Not “things to consider” — hard questions with binary answers. If you can’t answer them with evidence, the idea isn’t ready.
Is the pain real?
Not “do people have this problem” — can you find real humans describing this pain unprompted? In forums, app reviews, Reddit threads, support tickets? If the only person who thinks this is a problem is you, it probably isn’t one.
The hierarchy matters: someone paying for a workaround > someone complaining online > someone saying “yeah I’d use that” in a survey. The first is evidence. The last is politeness.
Before building RevenueHunt, we checked the Shopify App Store. The existing quiz apps had many reviews — and many of them were negative. That told us two things at once: people were paying for this and they weren’t happy with what existed. That’s the strongest signal you can find. Not “would you use this?” but “people are already buying this and complaining about it.”
Who’s the buyer and how do they find you?
This is where most bootstrapped ideas die. You build something great and then realize you have no way to reach the people who need it.
The question isn’t “who would benefit from this?” — it’s “can I get in front of 100 potential buyers this week without spending money?” If the answer is no, you have a distribution problem. And distribution problems don’t solve themselves after launch.
Marketplaces (Shopify App Store, Chrome Web Store, WordPress plugin directory), communities where buyers already congregate, SEO for queries people are already searching — these are distribution channels that work for bootstrappers. “We’ll do content marketing” with no existing audience is not a plan.
Can the product spread itself?
This is what separates lifestyle businesses from businesses that scale. Does usage create visibility? Embeddable widgets, shareable reports, “powered by” badges, output that users show to others — these turn customers into distribution.
MetricSpot had it. RevenueHunt has it. My failed products didn’t.
Do the economics work at small scale?
VC math: “If we get 1M users at $10/mo, that’s $120M ARR!” Bootstrapper math: “Can I cover my costs with 50 paying customers?”
Different question, different answer, different businesses worth building. If you need 10,000 customers to break even, you need venture capital. If you can break even at 100, you can bootstrap.
The unit economics need to work when you’re small, not just when you’re big. What does customer acquisition actually cost through your planned channels? What’s the realistic price point? Does gross margin exceed 60%? Can you raise prices 20% without everyone churning?
Why now?
This is the question people skip. “It’s a good idea” isn’t enough. Why didn’t someone build this five years ago? What changed — in technology, regulation, buyer behavior, or market structure — that makes this possible or necessary right now?
When I built RevenueHunt, product recommendation quizzes already existed on Shopify. The market was validated. But the existing apps were terrible — buggy UX, clunky builders, and pricing that started at $300-500/mo. For a Shopify merchant doing $20K/mo in revenue, that’s insane.
The timing wasn’t “this category is new.” The timing was “the incumbents got lazy and overpriced.” Nobody had tried freemium. We came in free, ate the market, and the incumbents couldn’t respond without cannibalizing their own revenue. That’s a timing window.
AI making code nearly free is a different kind of timing shift. New privacy regulations forcing compliance is another. “People have always needed this” is not a timing argument — it’s a reason to be suspicious.
Does your message resonate with a specific buyer?
Most founders obsess over product-market fit. But there’s a step before that: message-market fit. Can you describe what you do in one sentence, and does the right buyer immediately think “that’s for me”?
RevenueHunt is a product recommendation quiz. Generic. Could be for anyone. But 40% of our customers are skincare and beauty brands. When we stopped saying “product recommendation quiz” and started saying “help your customers find their perfect skincare routine” — conversions jumped. The product didn’t change. The words did.
Message-market fit means your target audience sees themselves in your messaging instantly. If you need a demo to explain what you do, the messaging isn’t there yet. If your homepage could belong to any of your competitors, the messaging isn’t there yet.
The test: pull exact phrases from app store reviews, Reddit complaints, and forum posts where your buyers describe the problem. Use their words, not yours. When a skincare brand owner reads “help customers find their perfect routine” — that’s their language, their problem, their world. “AI-powered product recommendation engine” means nothing to them.
Are you the right founder for this?
A great idea with the wrong founder still fails. Do you have domain knowledge — not “I read about it” but “I’ve lived this problem”? Do you have distribution access — an audience, a network, a community you’re already part of? Can you ship a working version alone in under 3 months?
If you’re an outsider with no distribution access building something that requires deep domain expertise and a sales team, it doesn’t matter how good the idea is. That’s a different founder’s business.
Your product is not your moat
This is the part most founders get wrong in 2026.
Code is a commodity. A solo developer with Claude can ship in a weekend what used to take a funded team three months. Distribution is the only moat left. That’s great for bootstrappers — it means you can build anything. It also means anyone else can build your thing.
If your moat is “we built it” — you don’t have a moat. If your moat is “better features” — you have a temporary head start, not a moat. Someone will copy your features in weeks. AI made that inevitable.
Real moats take time to build, and that’s exactly why they work:
Distribution is the moat. If you own the channel — a marketplace ranking, an embedded widget on thousands of sites, an audience that trusts you — copying your product doesn’t help competitors. They still can’t reach your customers. RevenueHunt’s “Powered by” badge on thousands of Shopify stores is a distribution moat. The quiz logic isn’t hard to copy. The installed base is.
Stored user value is the moat. When users invest time building workflows, uploading data, customizing settings — leaving becomes painful. Not because you locked them in with contracts, but because they’d have to rebuild everything. That’s earned lock-in.
Data that compounds is the moat. Every user interaction makes the product smarter, more personalized, harder to replicate from scratch. A new competitor starts at zero. You start with months or years of accumulated signal.
What’s NOT a moat: your code, your features, your UI, your AI model, “first mover advantage,” or “our team.” All of these can be replicated. Some in days.
Before you build, ask: what will make this business harder to kill in two years, not easier to copy? If the answer is “nothing” — the idea might still be worth building, but price it like a commodity and plan accordingly.
The AI substitute test
One more question that matters more every month: could someone get 80% of your product’s value from a ChatGPT prompt?
If yes, you don’t have a product. You have a prompt wrapper. And prompt wrappers are worth $0. The businesses that survive are the ones where the value isn’t in the output — it’s in the data, the workflows, and the network that only exist because users showed up and stayed.
Why I automated this
I used to run all of this analysis manually. Reading market reports, Googling competitor pricing pages, building spreadsheets, checking app store reviews to understand what customers complain about.
It worked. But it took days per idea. And the frameworks were locked in my head.
So I took every question, every scoring rubric, every sanity check I’d built over a decade and turned them into AI-powered skills. 10 research agents run in parallel — analyzing competitors across 16 dimensions, sizing markets bottom-up, stress-testing unit economics, scoring founder-business fit. Each one pulls real data from the web. The whole thing takes about 30 minutes — not days — and goes deeper than most paid consultants.
The scoring is opinionated. Distribution and problem validation are weighted heaviest — because those are what actually kill bootstrapped businesses. Ideas that fail the gates get a clear “don’t build this” signal. No soft landings, no “it depends.”
It’s my decision-making process, automated. It’s opinionated because I am.
FAQ
How do I validate a business idea without spending money?
Find real humans describing the pain unprompted — in forums, app reviews, Reddit threads, support tickets. The hierarchy: someone paying for a workaround > someone complaining online > someone saying “yeah I’d use that” in a survey. The first is evidence. The last is politeness.
What’s the best moat for a bootstrapped business?
Distribution. If you own the channel — a marketplace ranking, an embedded widget on thousands of sites, an audience that trusts you — copying your product doesn’t help competitors. Code and features can be replicated in days. An installed base can’t.
How do I know if my idea needs venture capital?
If you need 10,000 customers to break even, you need VC. If you can break even at 100 paying customers, you can bootstrap. The unit economics need to work when you’re small, not just when you’re big.
Could an AI replace my product?
Ask: could someone get 80% of your product’s value from a ChatGPT prompt? If yes, you have a prompt wrapper, not a product. The businesses that survive are the ones where the value is in the data, the workflows, and the network — not the output.
The Bootstrapper Toolkit is open source. Clone it, run /analyze-idea in Claude Code, and get a scored analysis of any business idea. It’s based on 30+ business books I actually read, filtered through a decade of building things — most of which failed.