Why brainstorming produces bad ideas
A brainstorm rewards ideas that sound good in a room. That's a different quality from ideas that survive contact with someone's budget. Sitting down to think of a business idea puts you in the position of guessing what other people need, from memory, under time pressure — which is close to the worst possible conditions for the task.
The alternative is not cleverness. It's noticing. Almost every idea worth building is already visible as work someone is doing badly, slowly, or with a spreadsheet that has become a small tragedy.
Four places ideas are already visible
1. Work you personally do by hand and hate
The strongest starting point, because you can verify the problem without permission. If you've built a spreadsheet to survive a recurring task, that spreadsheet is an unbuilt product with one confirmed user.
2. Jobs you can watch closely
Any job where someone re-types the same information into two systems, or where a task ends with "and then I just do it manually." You don't need to be an expert in the field — you need to watch an hour of someone's actual afternoon.
3. Existing tools with visible cracks
Read the one-star and three-star reviews of software people already pay for. One-star reviews are often about billing. Three-star reviews are gold: someone who paid, stayed, and is specific about what's missing.
4. Communities that keep asking the same question
Forums, subreddits, Discord servers, and trade groups where the same question is asked every few weeks and the answer is always a workaround. A repeated question with no clean answer is a market signal.
Separating irritation from a real problem
Not every complaint is a business. The filter is cheap to apply and saves months.
| Question | Weak answer | Strong answer |
|---|---|---|
| How often does it happen? | Once a quarter | Weekly or daily |
| What does it cost now? | It's annoying | Hours, or money already being spent on a worse fix |
| Who feels the cost? | Someone downstream | The person who could choose to buy |
| What do they do today? | Nothing | A workaround they built and maintain |
The last row matters most. An existing workaround is proof of demand that costs you nothing to collect. Someone maintaining a fragile spreadsheet has already told you the price they'll pay in effort; you're only asking them to pay it in money instead.
Writing the idea in a testable form
"An app for travel agencies" isn't testable. It has no person, no task, and no current cost. Rewrite every idea into this shape before you do anything else with it:
[Specific person] spends [time/money] doing [specific task]today by [current workaround],because [existing tools don't do X].A one-week idea search
- Days 1–2: collect without judgingWrite down every manual process you see or perform. Thirty entries, no filtering. Judging while collecting kills the collection.
- Day 3: apply the four questionsRun every entry through frequency, cost, who pays, and current workaround. Most will die here, which is the point.
- Day 4: rewrite the survivors in the testable formIf you can't name the specific person, the idea isn't ready — go back and watch someone.
- Day 5: pick one and talk to five people who have itNot to pitch. To ask what they do today. That conversation is the next lesson.
Common mistakes
- Starting from a technology ("something with AI") instead of a task someone repeats
- Falling for ideas that are fun to build rather than painful to live without
- Dismissing boring industries — invoicing, scheduling, and compliance are boring because nobody fixed them
- Keeping the idea secret. Almost nobody will steal it, and describing it out loud is the cheapest test there is
Key takeaways
- Ideas are noticed, not invented — look for manual work, not gaps in a market map
- An existing workaround is the strongest free evidence of demand you can find
- Filter on frequency, current cost, who pays, and what they do today
- Rewrite every idea to name a specific person and a specific task before testing it
Try it yourself
Open the last spreadsheet you built for yourself — any spreadsheet that exists because software didn't do something. Write the four filter questions against it. If it passes three of four, you've found a candidate without leaving your desk.