Most startup ideas start as a solution.
"An app that does X." "A tool that uses AI for Y." It's an easy trap, especially as an engineer. You see a technology and you think of things it could do.
The trouble is that a solution without a clear problem has nothing to push against. You can't tell if it's good, because you don't know what it's supposed to fix.
So this chapter goes one step earlier: find the problem first.
What this means
A problem is something a specific person struggles with, whether or not your product exists.
An idea is one possible way to fix it.
That difference matters because problems can be checked and ideas mostly can't. You can ask people about a problem they have today. You can watch them deal with it. You can see what they already do about it. An idea is a guess about the future, and people are bad at predicting whether they'd use something that doesn't exist yet.
When you start from the problem, you get something to measure your idea against.
What I learned
Here's the problem HairOver is built around, in plain words:
You find a hairstyle you love. You save it. You show it to your barber. You get the haircut. Then you look in the mirror and think: why does it look so different on me?
The reason it looks different is simple. The photo you saved is of someone else: a different face shape, hairline, hair texture and density. The haircut can be perfect and still look wrong on you.
What I like about this as a problem is that you don't have to explain it. Most people have had some version of it. That's a good sign. If you have to argue with people to convince them your problem exists, it probably doesn't exist for them.
But I also want to be honest about where it's weaker, because this chapter would be useless if I only showed you the good side.
Real example: checking HairOver's problem
I'll run HairOver's problem through the questions I think matter most. This isn't a scoring system. It's how I think about it.
Do I understand the problem?
Mostly, yes. It's a common, visual problem, and I can see the whole path clearly: find a photo, show a barber, get a result that doesn't match. That clarity is what made the core product line easy to write: photo → analysis → recommendations → try it on your own face. I explain the first version screen by screen here.
How painful is it?
Medium, I think. A bad haircut is annoying and it's on your face for weeks, but it grows back. It's not like losing money or losing a customer. That means people will care in the moment before a haircut, but they won't be desperate.
This matters for pricing. Mild pain usually means people are less willing to pay a lot.
How often does it happen?
This is HairOver's honest weak spot. People don't get haircuts every day. Even frequent haircuts happen every few weeks, and a lot of people change their style far less often than that.
A problem that comes up rarely makes a product harder to build a business around. People don't naturally come back, and you have to give them a reason to.
I haven't solved this. It's one of the main reasons I care so much about repeat usage. My distribution plan tracks "repeat generation" and 7-day retention for exactly this reason.
Will someone pay?
I don't know yet, and this is where a lot of consumer ideas break.
What I do know is the cost side. Every generated look costs me money: roughly NPR 15 per image when I worked out the unit economics. So "free forever" isn't an option, and people need to value the result enough to pay for it. The full math is in the pricing breakdown.
What do people already do about it?
This question gets skipped the most, and it's the most useful one.
Today, people deal with this problem by saving photos on Pinterest, looking at celebrities with a similar face shape, reading "best haircuts for a round face" articles, or just trusting their barber. None of those show you with the haircut.
That's where HairOver fits: not "hairstyle ideas" (there are millions of those) but seeing a specific haircut on your own face before you commit.
Competition, and why "no competitors" is a warning sign
Search the Play Store for "AI hairstyle" and you'll get a page of apps. Hairstyle changers, try-on apps, hair-color filters. HairOver is not alone, and I'd be lying if I said otherwise.
It's easy to see that and feel discouraged. I actually think it's reassuring, for two reasons.
Competition shows there's demand. Many companies wouldn't keep building hairstyle apps if nobody downloaded them. When you find an idea with no competitors, the most likely explanation isn't that you're a genius. It's that nobody wants it, or that someone tried and it didn't work.
Competition tells you what's already solved. If ten apps already let you change your hair color with a filter, building an eleventh doesn't add much. The interesting question is: what are they not doing well?
For HairOver, my current answer is about trust and specificity. For example, telling you which cuts are not ideal for your face, not just which ones look good. And basing recommendations on your analysis instead of a generic gallery. I think those are real differences. Whether users see them as important is something I still have to prove.
Finding a narrow starting point
The widest version of HairOver's problem is "people want to look good." That's too broad to act on.
The narrow version is: the moment right before a haircut, when you're deciding what to get.
That narrow moment turned out to be useful in a practical way. It told me where to find people. Before a haircut, people search things like hairstyles for a round face or best haircuts for a receding hairline. So the HairOver website is mostly guides for exactly those searches, each one leading to the same action: try it on your own photo.
A narrow problem gives you a clear customer, a clear moment, and a clear place to find them. A broad problem gives you none of those.
What to do
- Write your problem as a short story, without mentioning your product. Like the haircut story above. If it doesn't make sense without your solution, you haven't found the problem yet.
- Ask five people how they deal with it today. Not "would you use my app?" Ask "what did you do last time this happened?"
- Be honest about frequency and pain. Write down how often it happens and how bad it is. If both are low, you'll need a strong reason why people will still pay.
- Search for competitors on purpose. Install three of them. Read their one-star reviews. That's where the unsolved problems are.
- Find the narrow moment. When exactly does someone feel this problem most? That's your starting point, and it tells you where to find them.
Mistakes to avoid
Starting from a technology. "What can I build with AI?" is a fun question, but it isn't a problem. Start with the person.
Asking people to predict the future. "Would you use this?" gets polite yeses. "What did you do last time?" gets the truth.
Only looking at the good side of your problem. Every problem has weak spots. HairOver's is frequency. Find yours early, because they don't go away when you ignore them.
Panicking about competitors. Or dismissing them. Neither helps. Study them.
Starting too broad. "Helping people look better" isn't a starting point. "The five minutes before a haircut" is.
My current thinking
I'm confident the HairOver problem is real. Almost everyone understands it right away, and people already put effort into solving it with saved photos and articles.
What I'm less sure about is whether it's frequent enough and painful enough to build a big subscription business on. That's an open question, and I'd rather say so than pretend.
If it turns out the problem is real but too occasional, that doesn't mean the idea is dead. It might mean the business looks different, maybe more about one-off purchases or about the moment before a big change, and less about daily use. I'll find out from how people actually use the app, not from thinking about it.
Key takeaway
Start with a problem you can describe without your product. Check how painful it is, how often it happens, what people already do about it, and whether they'd pay. Treat competition as evidence, not a threat. Then shrink the problem down to one clear moment. That's where your first customers are.
Next: Validating the Idea. Once you have a real problem, the job is reducing uncertainty about your solution, not proving you're right.