You have an idea. Maybe three. And the advice waiting for you online splits into two equally unhelpful camps: "just launch and find out" on one side, "run a hundred customer interviews before you write a word" on the other. One risks your savings on a hunch; the other turns a small dream into an unpaid research job.
There's a third way, and it's calmer than both: you can't prove an idea is good before you start — but you can make the ways it might fail show themselves cheaply, and a cheap early "no" is a win. That's all evaluation is. Not a verdict, not a business plan, not permission from the universe. A short series of small tests, run in order of cheapness, that turns "I think" into "I checked."
This matters more now, not less. Tools — AI ones especially — have made building something faster than it has ever been, which quietly means the scarce skill has moved one step earlier: deciding what's worth building at all. This guide walks that deciding, end to end.
What evaluating an idea is actually for
The question "is this a good idea?" has no direct answer, which is why it keeps you up at night. Swap it for two questions that do:
- What would have to be true for this to work? People with this problem exist, they already feel it, they'd accept this solution, they'd pay something, I can reach them.
- Which of those things can I check this week, for almost nothing?
Evaluation is buying information at the lowest price available. Every test you run either raises your confidence honestly or kills the idea before it costs you a season. Both outcomes are the method working — and it's worth saying plainly that most ideas don't survive evaluation. That isn't a talent problem. Brainstorming produces mostly duds by design; evaluation is the sieve, not the enemy.
The evidence ladder: said, did, paid
Not all evidence weighs the same. Picture a ladder with three rungs, each one costing the other person a little more — and therefore telling you a little more.
What people say. The bottom rung, and the wobbliest. Friends encourage, strangers are polite, and everyone answers hypothetical questions with a hypothetical self. Talk to people anyway — but ask about their current behaviour, not your idea. "How do you handle dinner after a night shift?" teaches you something. "Would you buy a meal kit for shift workers?" teaches you how nice they are.
What people do. The middle rung. Someone joins a waiting list, replies to a post asking who wants this, books a fifteen-minute call, shows up. Doing costs a sliver of effort, so it filters out pure politeness.
What people pay. The top rung. A deposit, a pre-order, a first invoice — even a small one. Payment is the only signal that can't be faked by kindness.
You don't need the top rung to make a decision. You need to know which rung your current confidence is standing on — because most abandoned-later ideas were built entirely on the bottom one.
A week of desk research before you talk to anyone
An honest evening or two at the kitchen table answers a surprising amount:
- Who's already doing this? Competition is evidence that people pay for this — a genuinely comforting thought. An empty market is a question ("why has nobody made this work?"), not a green light.
- What do their customers complain about? Reviews and forum threads are a free map of what the incumbents get wrong — and unmet complaints are where new entrants live.
- What do things cost? Note real prices people currently pay to solve this problem, badly or well. You're not setting your price yet; you're learning whether money already moves here.
- Where would your first ten customers come from — specifically? A named group, a real place, a community you're already in. "The internet" is not an answer.
One piece of discipline turns this research from comfort-shopping into evaluation: write down what "promising" would look like before you start looking. Deciding your criteria in advance is what keeps you from grading your own homework generously — the same move that keeps any decision honest, as the guide to deciding whether a tool is worth buying shows on a smaller scale.
Keep the liking separate from the evidence
Here's the trap the ladder can't save you from on its own: enthusiasm feels exactly like evidence from the inside. The idea you love produces the same warm certainty as the idea three strangers have asked you to build, and only one of those certainties is about the market. Before you score anything, it's worth running the tests in how to tell a good idea from an idea you just like — they exist precisely to find that gap, and this guide won't repeat them. The one-line version: liking is a measurement of you; goodness is a measurement of someone else's problem. Both matter. Only one predicts customers.
Comparing several ideas without a spreadsheet spiral
If you're holding more than one idea, resist the forty-column scoring matrix. Three questions per idea, answered in a sentence each, are enough:
- Evidence: what rung of the ladder is this idea's support actually standing on today?
- Fit: does it match your real skills, hours, and energy — not your January self's? (For side projects, the fit-first method in how to choose a side-hustle goes deeper.)
- Cost of the next test: what would the next piece of evidence cost you, in evenings and money?
Then pick the idea whose next test is cheapest relative to what it would teach you — and give the decision one evening, not one month. The unchosen ideas go in a parking-lot note, where they will patiently keep. Testing one idea properly beats keeping three warm forever, because attention, unlike ideas, doesn't multiply.
A worked example, start to finish
Take a concrete case. Dana, a nurse who cooks well, is weighing "batch dinner kits for shift workers" — colleagues come home at odd hours and eat toast.
- Desk research, two evenings. Existing meal services deliver in daytime windows; forum threads from shift workers complain about exactly this. Real money already moves toward meal kits. Criteria written down first: "promising means five colleagues describe the problem unprompted, and at least three later pay for a trial week."
- Said, one week. Five colleagues, asked only about their current after-shift eating. Four describe the toast problem without prompting. No pitch yet.
- Did, one week. A short message in the staff group: "Thinking of doing weekly batch dinner kits sized for post-shift reheating — who'd want details?" Nine replies. That's the middle rung.
- Paid, two weeks. A one-page offer with a real price for a two-week trial, sent to the nine. Three pay. Dana cooks from her own kitchen — no branding, no website, no equipment bought.
Total outlay: groceries and a handful of evenings. If the three re-order, Dana has top-rung evidence and a next step. If everyone loved the idea but nobody paid, the ladder just saved her a year. Either way, she decided once, on evidence, and can stop re-deciding nightly.
When the answer is no
A "no" from a test like this stings for about a day and saves you a season — that's the trade, and it's a good one. Results in any venture depend on skill, timing, market, and honest luck; evaluation improves your odds, it never guarantees them. When an idea falls, go refill the well rather than forcing the next candidate: the brainstorming techniques guide is the systematic route, and why can't I think of ideas is the diagnostic for when the well itself feels dry.
After a yes
A surviving idea earns the smallest real version of itself: one offer, one channel, the manual version of anything that could be automated later. Somewhere in those first weeks it will also earn a name — how to name a business covers that calmly. One honest caveat as you commit: availability of names, domains, and handles changes constantly, and trademark law is jurisdiction-specific, so check the relevant registry — and for anything you'll invest in seriously, a trademark professional.
FAQ
How long should evaluating a business idea take? Days to a few weeks, run in your spare evenings — not months. Set a deadline for the decision when you start; evaluation without an end date quietly becomes a hobby of its own.
Do I need surveys or interviews? A handful of honest conversations about people's current behaviour beats a large survey about a hypothetical one. You're looking for what people already do and pay for, not what they predict about themselves.
What if someone steals my idea while I'm testing it? A common worry, rarely a real one. Ideas are cheap and execution is the work; meanwhile, secrecy blocks the only thing that can make your idea stronger — contact with real people. Test openly.
Is a crowded market a bad sign? Usually the opposite: it proves people pay to solve this problem. The question becomes which narrow group the incumbents serve badly — that's your opening, and their complaints will point to it.
What if I can't afford to test anything? The bottom two rungs — conversations and small asks — cost time, not money. Only the "paid" rung needs a real offer, and even that can be a manual, kitchen-table version of the service.
Evaluation isn't the opposite of dreaming — it's how a dream earns your next three months. And if you're short of candidates to test in the first place, get idea starting points from the Mellow Ideas generators — friendly first drafts to react to, judgment still yours.
