How to come up with business ideas: a method that works

The reliable way to come up with business ideas is a four-step method: mine problems you have personally paid to solve, map unbundling opportunities inside large products, run structured divergence to 30 or more candidates, then score each on impact versus effort and pick one.

Waiting for a lightning bolt is not a method. Founders who generate good ideas on demand do the same things every time: they look in places where problems are already proven, they force volume before judging anything, and they pick with a scoring system instead of a mood. This post walks through that method end to end, with a worked example.

Mine your own problems first

The cheapest market research you will ever do is an inventory of your own frustrations. You already know the problem is real, you already know what a fix is worth, and you already speak the customer's language because you are the customer.

Run the inventory as a structured pass, not a vague reflection. Go through these five sources and write down every friction point, however small:

• Your bank and card statements. Every recurring charge is a problem someone convinced you to pay for. Which of those tools do you tolerate rather than like?
• Your browser history and open tabs. Repeated searches are unresolved problems. If you have googled "how to export invoices from X" three times, so have thousands of others.
• Your last job. Which task did everyone dread? Which spreadsheet held the company together? Internal duct tape is a product category waiting to happen.
• Your messages. What do friends and colleagues ask you for help with? Repeated requests mean you have expertise people value and no product is serving.
• Your workarounds. Anywhere you chain three tools together to do one job, there is room for a product that does the job in one.

A concrete example: a support lead notices she spends every Monday copying churn-risk signals out of three dashboards into a spreadsheet for her weekly review. That is not a complaint, it is a product spec. "A tool that watches product usage and flags accounts likely to churn, with the reason attached" came out of exactly this kind of inventory, and it is a clearer starting point than "something in AI."

Map unbundling opportunities

Every large product is a bundle, and every bundle has a feature that some customer segment would pay for on its own, done ten times better. Craigslist unbundled into dozens of companies. Excel keeps unbundling into vertical SaaS. The pattern repeats because big products must serve everyone, which means they serve nobody's edge case well.

To hunt unbundling opportunities, pick a product with a huge user base and ask three questions:

1. Which single feature do reviews complain about most? Complaints inside a popular product are demand with the payment already proven.
2. Which user segment is the product clearly not built for? A tool built for enterprises leaves freelancers underserved, and the reverse.
3. What job do people do next to the product, in a spreadsheet or a doc, because the product stops one step short?

Worked example: take a mainstream email platform. Reviews complain that its deliverability reporting is shallow. Newsletter writers, a fast-growing segment, get generic dashboards built for e-commerce senders. The unbundled idea writes itself: deliverability analytics built only for newsletter operators, priced per subscriber. You did not invent demand, you found it already queued up inside someone else's product.

Run structured divergence, not passive waiting

Once you have two or three promising problem areas, resist the urge to run with the first idea. The first idea is almost always the obvious one, which means it is also everyone else's first idea. You want volume: 30 or more distinct directions per problem area before you judge anything.

Your first idea is the market's first idea too. The advantage is in idea number nineteen.

Structured divergence means using frameworks that force you off your default track. SCAMPER makes you substitute, combine, and reverse parts of an existing solution. Reverse brainstorming asks how you would make the problem worse, then inverts the answers into fixes you would never have listed directly. Rapid ideation techniques like timeboxed quotas push you past the comfortable first dozen. If you want the machine to do the heavy lifting, a startup idea generator can take one problem statement and produce dozens of genuinely different directions, each tagged by angle, in under a minute.

Whatever the method, apply one rule: every idea gets written down in full sentences with a named customer. "AI for fitness" is not an idea. "A workout app that builds the week's plan around the equipment in the user's hotel gym, for consultants who travel weekly" is an idea you can evaluate.

Score the shortlist honestly

Picking is where most founders go wrong, because they pick the idea they are infatuated with rather than the one the evidence supports. The fix is a scoring pass with criteria you write down before you look at the list.

Score each shortlisted idea from 1 to 5 on four criteria:

• Pain intensity. Do people already pay, or burn hours, to solve this?
• Reachability. Can you name three places the customer already gathers?
• Effort to first version. Could a rough version exist in 60 days?
• Your unfair advantage. Why you: distribution, domain expertise, an existing audience?

Plot the results as impact versus effort and the decision usually makes itself: one or two ideas sit in the high-impact, low-effort corner, and the seductive moonshot reveals itself as a five-year project with no wedge. An idea prioritization pass like this takes twenty minutes and saves months of building the wrong thing.

A worked example, end to end

Put the method together on one case. A freelance designer wants a business idea. Problem mining surfaces a repeated pain: chasing clients for feedback scattered across email, chat, and calls. Unbundling analysis shows that big project management suites all treat client feedback as an afterthought buried in comments. Structured divergence produces 34 directions, from "feedback inbox that turns emails into tracked tasks" to "a client portal that charges late-feedback fees automatically." Scoring puts the feedback inbox first: intense pain, reachable audience (design communities), a 60-day first version, and the founder's own network as an unfair advantage. Total time from blank page to committed pick: two sessions. Tools like Brainstormer compress the divergence and scoring stages into one sitting, but the method works on paper too.

Common mistakes to avoid

Three failure modes account for most dead ends. First, judging ideas during generation: every premature "that won't work" costs you the three better ideas that would have branched from it. Keep the modes separate. Second, mistaking a market for an idea: "pet owners spend a lot" is an observation, not a business. Third, skipping the honest scoring because one idea feels destined. Feelings are data, but they are one criterion among four, not a veto over the other three.

The method is repeatable on purpose. Run it quarterly and you will never again sit in front of a blank page asking where ideas come from. They come from problems already being paid for, bundles already straining, and a divergence process that refuses to stop at the obvious.