I've never liked the usual advice about finding startup ideas.
"Study market size." "Brainstorm a hundred ideas." "Find an underserved niche." It all sounds clean on paper. In practice, it's almost useless — at least for how my brain works.
I don't sit down and ask myself, "What startup should I invent?" I start somewhere much more concrete, and usually much more emotional. I start with products people already use, already need, and increasingly dislike.
Not broken products. Not dead products. Products where something in the value exchange has quietly shifted. The pricing stopped feeling fair. The free tier got worse. The UI gained weight. Trust eroded. What used to feel elegant started feeling extractive.
People keep using the product. But they stopped respecting it.
That's where I start paying attention.
I don't search for "ideas." I search for products that stopped earning their position.
The biggest shift in how I think about startups was realizing that "startup idea" is usually the wrong unit of analysis.
Weak ideas almost always start in abstraction. Market maps. Generic problem statements. "AI for X." "Uber for Y." That's not how I want to think. I'd rather study a specific product that used to make sense and doesn't anymore.
I watched this happen with Heroku. For years it was the way to deploy a web app. Push your code, get a URL. Beautiful. Then Salesforce acquired it, the free tier disappeared, the dashboard stagnated, and the pricing started feeling absurd for what you got. People didn't stop needing easy deployment — they stopped trusting Heroku to care about them. Railway and Render didn't need a revolutionary concept. They just needed to be what Heroku used to be.
That's not a dying category. It's a live category with stored-up frustration.
A dead category isn't interesting. A category where people still need the job done but resent the current tradeoffs — that's where openings get created.
I care about what I'd call product drift more than trend reports. Trends tell you where attention is flowing. Product drift tells you where openings are being created. And most of the time, those openings aren't obvious from the outside. The market looks "crowded." But internally, it's unstable.
Resentment is my favorite signal
If I had to pick one thing to listen for, it'd be resentment. Not curiosity. Not launch-day hype. Resentment.
There's a massive difference between "this is solid, but I wish it had one more feature" and "this used to be good, now it feels like they're squeezing me." The second market is far more interesting.
I watched it with Mailchimp after the Intuit acquisition. For years it was the friendly email tool — simple, generous, everyone's first choice for newsletters. Then pricing restructured. Features moved behind paywalls. The UI got heavier. Suddenly the indie-hacker forums were full of people saying "I only use Mailchimp because migrating my list is a pain."
That sentence — "I only use it because..." — is gold. It means the product is surviving on inertia, not on love. And inertia is a weaker moat than most people think.
Users don't leave beloved tools instantly. They leave in stages. First they complain. Then they build workarounds. Then they start recommending alternatives in Reddit threads and forum comments. Then the replacement market appears.
The phrases I pay attention to:
- "We still use it, but..."
- "This used to be free."
- "I only keep it because everyone else is on it."
- "The problem isn't the price — it's how they charge."
- "There has to be a simpler way to do this."
Those aren't just complaints. They're evidence that the company's old product logic is breaking down. And once enough people feel that, the market doesn't need a new idea. It needs a product that feels sane again.
Fragmented workflows are free real estate
This is probably the strongest recurring pattern in how I think. And as a frontend developer, I notice it constantly.
Users don't think in software categories. They think in outcomes. Nobody wakes up wanting "appointment management software with integrated payment support and light CRM." They want to book something, confirm it, pay for it, maybe message the person, and move on.
They don't care that the market split that into six tabs.
I watched this play out with design tools. A few years ago the workflow was: design in Sketch, prototype in InVision, hand off specs in Zeplin, discuss in Slack, manage assets somewhere else. Five tools for one job. Figma didn't invent design. It collapsed a fragmented workflow into one surface — and the entire industry followed.
Same pattern, everywhere. One tool for communication. Another for scheduling. Another for payments. Another for follow-up. From a pitch deck, that looks like a "mature ecosystem." From the user's chair, it looks ridiculous.
The question I keep asking: what are people already doing through too many disconnected tabs?
If the answer is clear, and the fragmentation feels historical rather than necessary, I get interested. The best opportunities here aren't about inventing new behavior. They're about removing handoffs, repeated input, and unnatural boundaries between adjacent actions that the user already performs.
Competition doesn't scare me. Lazy incumbents do.
Some of the most interesting markets look terrible if you just count the competitors. Category is full. Leader is big. Alternatives exist. Surface-level conclusion: too late.
I don't buy that.
Notion entered a space with Confluence, Google Docs, Evernote, and dozens of wiki tools. By market-map logic, it was insane. But Confluence was bloated. Evernote had lost its way. Google Docs was powerful but not opinionated. Notion didn't find an empty niche — it found a crowded space where nobody loved their tool.
The real question isn't how many players are in the space. It's whether the leaders still have product logic strong enough to deserve their position.
Sometimes they do. Sometimes the moat is real.
But often it's weaker than it looks. Maybe the product has distribution but no love. Maybe switching costs are more psychological than technical. Maybe everyone uses it, but nobody would build it that way today.
I pay special attention when a market leader starts relying on friction, opacity, or inertia instead of product quality. That usually means they stopped being product-first — or the category matured enough that the old design decisions no longer make sense.
I'm not looking for empty white space. Empty white space is often empty for a reason. I'm looking for crowded spaces where the energy has shifted from loyalty to toleration.
Distribution needs to be baked in
A lot of startup ideas die not because the product is bad, but because the distribution model is fantasy. I try to be honest about this early.
How does this thing spread without me manually acquiring every user?
I don't need viral consumer loops. But I strongly prefer products where usage creates exposure. A shared booking link. A collaborative page. A payment request. A public profile. Anything where one user naturally surfaces the product to someone else as part of the normal workflow.
Calendly understood this — every booking link is a tiny ad. Linear understood this — shared project boards expose the tool to every collaborator.
A product that lives exclusively inside one person's account, invisible to everyone else, has a much harder growth path. It can work. But it needs to be significantly stronger to compensate.
Pricing should feel fair. That's it.
A lot of products rot because their monetization drifted away from the value they create.
At first the company wins by solving a problem. Later it grows by controlling access to that solution. Eventually it adds toll booths everywhere — seat limits, feature gates, usage caps that feel arbitrary.
I'm drawn to businesses where revenue scales with user value. More usage, more transactions, better outcomes — more revenue. Not "you added a team member, pay $15 more per month for the same product."
A lot of categories are vulnerable precisely because incumbents monetize lock-in instead of utility. Users are willing to pay. They just don't want to feel punished for growing. That gap between "willing to pay" and "feeling manipulated" is where cleaner alternatives gain traction.
Design quality is the wedge, not the polish
This is where my frontend bias shows up.
I think founders underestimate how many categories are waiting for product clarity more than for more capability. There are markets where the problem isn't missing features — it's that the current tools feel heavier than the job itself.
I've spent enough time building interfaces to know the difference between software that does the thing and software that feels like the thing. Linear feels like project management. Jira feels like enterprise software that happens to track tickets. Both "work." One makes you want to open it. The other makes you want to close the tab.
A clean interface doesn't sound like a moat — until you realize most of the market still feels like billing logic wearing a UI.
I don't mean "make it prettier." I mean: make it faster, calmer, more obvious at the exact moment the user is trying to get something done. In categories people touch every day, that emotional texture matters more than another settings panel.
My actual research process
It's not glamorous.
I don't learn from polished market maps or startup Twitter threads. Those are too abstract and too clean. I learn from the internet when it looks messy.
Negative reviews. Migration discussions. Pricing backlash. Reddit rants. App-store one-stars. Changelog reactions where users ask "why did you remove this?" Forum posts explaining the workaround someone built because the official product flow doesn't work.
I also dig through "X vs Y" comparison threads — not for the conclusion (usually SEO-optimized nonsense), but for the specific complaints people articulate along the way. When someone writes a 400-word comment explaining exactly why they left a product, that's more useful to me than any industry report.
The goal isn't to collect facts about a market. It's to find where users are excusing bad product behavior out of habit — and where a cleaner product could make the old tradeoff feel unnecessary.
Traps I've learned to recognize
Knowing which opportunities are real is only half the skill. The other half is knowing which attractive-looking categories are structurally broken.
The network-effect wall. If the idea basically requires people to leave the platform where all their contacts already live, I get cautious. People complain about entrenched communication products constantly, but switching social infrastructure is far harder than dislike alone suggests. In those markets, I'd rather build on top of existing behavior than try to replace it.
Software costumes on operational problems. Some categories look like software opportunities but are actually trust, fraud, compliance, or logistics problems wearing a nice UI. A prettier marketplace doesn't make supplier vetting easier. An elegant dashboard doesn't make insurance underwriting simpler. If the hard part isn't the interface, a better interface isn't the answer.
Adversarial marketplaces. If one side of the market feels extracted from the moment they join — "pay to get leads, no guarantees" — that's a warning sign. Those platforms build deep supply-side resentment, which is a terrible foundation regardless of how good the demand side looks.
The feature-not-a-company trap. Some products are real but too narrow, too infrequent, or too easily absorbed by a larger platform. The demo always looks great. It's the retention curve that tells you whether it's a business or just a nice utility.
What a strong opportunity actually looks like
After enough looking, you start recognizing a shape.
The underlying behavior already exists — people are doing the job, just through fragmented tools. The market leader is big but coasting. Users complain in emotionally specific language. The product could spread through usage. The business model could feel cleaner. And the interaction could be dramatically better — not more featured, but more obvious.
When most of these line up, I trust the conviction that follows. Not "this market is large" conviction. The more specific kind: "this product category started violating user expectations, and a new product could fix that."
I don't think the best startup ideas come from staring at a blank page trying to be original. They usually come from paying attention to products that already proved demand — and noticing when the relationship between product, user, and business model starts to decay.
It's the same thing I keep finding in my own work. The tools that age best aren't the most complex — they're the ones where the design decisions still make sense. The moment they stop making sense, someone builds something better. That's true for TypeScript patterns, and it's true for startups.
Originally published on maryanmats.com.
