How to find problems worth solving
Published · By Daily Product Idea · Our methodology
The short answer
A problem is worth solving when four things are true: specific, identifiable people have it repeatedly; it costs them money, time or risk they can describe; they already work around it, ideally by spending money; and you can reach them. The first three can be observed in what people write and do. Whether they will pay you is usually a hypothesis until you test it.
The four tests
Plenty of problems are real and still not worth building a product for. They happen once a year, they cost nothing anyone would pay to avoid, or the people who have them cannot be found. Before building, check a problem against four tests:
- Who: can you name the people who have it, specifically enough to find them?
- How often: does it recur, for many independent people, not just once for one person?
- What it costs: can the people who have it say what it costs them in money, time or risk?
- What they do now: do they already work around it, and do they spend money doing so?
A fifth question, whether you can reach these people at a sensible cost, decides whether the problem is worth solving for you. It is part of how every Daily Product Idea opportunity is scored.
Who has the problem
"Small businesses" is not a customer. "Owner-operators of small service businesses who quote each job individually" is. The narrower description tells you where to look for them, what words they use and what they already own. Deal Floor scores reasonably on customer clarity because forum threads identify a group who recognise themselves (owners arguing about whether to charge for estimates) and it scores lower where the research could not show how large that group is.
How often, and for how many
One vivid post is an anecdote. Look for the same problem described by different people, in different threads, ideally in different communities. Mail Vault is an honest example of the minimum: it rests on a single Hacker News thread asking how to archive 25 years of email, and its evidence level is Limited for exactly that reason. The problem may be common, but one thread cannot show it.
What it costs them
The most useful signals name a consequence. Deal Floor's owners describe jobs that look profitable on the invoice but end at zero or a loss once fees and unpaid sales time are counted. A consequence like that lets you price a product against the loss it prevents. A problem nobody can put a cost on is usually a nuisance, and nuisances are hard to charge for.
What they do now, and whether they pay for it
A workaround proves the problem is felt. Paying for one proves more. Research for Deal Floor found paid pricing guides and calculator bundles sold to contractors at between $37 and $197: people already spend money on this arithmetic. That is encouraging, and it also sets a low ceiling on price, which the assessment records as a risk.
Absence matters too. For Flight Deck, research looked for any sign that developers pay for this kind of tool, such as a job posting, a freelance listing or a paid product with reviews, and found none. The nearest alternatives were free and open source. That gap is recorded on the idea's page under "Looked for, not found", and it is the main reason its market-potential score is low despite strong evidence of the problem.
Check it outside the community it came from
A problem first noticed in an online community is still, at that point, one community agreeing with itself. It becomes more credible when a different kind of source points the same way:
- product reviews and support complaints about existing tools;
- search demand for the problem or its workaround;
- job or freelance postings paying someone to do the work by hand;
- public issue and feature-request trackers; and
- direct evidence: interviews, sign-ups, pre-orders, observed use.
Flight Deck is corroborated this way. Three separate issue-tracker threads, two on the Claude Code repository and one on another multi-agent tool, describe the same difficulty of telling which session needs attention. Vendor marketing pages do not count, however relevant: a vendor saying its customers want something is the seller talking.
Look just as hard for evidence against
The research that makes a problem look worth solving often also shows it is already being solved. Deal Floor's research found quoting tools that already display margin on every quote, and many free margin calculators. Flight Deck's found that several open-source tools already run one isolated workspace per agent, and write-ups reporting that the platform had added built-in support. Neither finding kills the idea, but each narrows it, and each belongs on the page next to the evidence in favour.
Evidence or hypothesis?
How Daily Product Idea records this
Every researched idea carries an evidence level counted from its sources, not judged: Limited, Indicative (two or more independent voices or sources), Supported (corroborated by two or more sites of the kinds listed above) or Direct (direct evidence from customers). Each level above Limited also needs the supporting sources to outnumber those that weaken the idea. See how ideas are researched and assessed for the full rules, and browse the ideas to see the sources behind each one.