The problem
Most startup ideas are not exceptional. That is not an insult. It is the normal math of innovation.
Ideas are cheap to generate and easy to love. What is rare is an idea with a sharp buyer, painful timing, believable economics, and a distribution path that is not fantasy. Founders get in trouble when they treat every exciting idea as if it has already earned survival.
The brutal truth is useful: if most ideas are weak, your advantage is not genius. It is the speed and honesty with which you filter, test, and kill them.
Why it fails
Weak ideas survive because humans are bad at separating identity from hypotheses.
The founder wants the idea to win. The team wants morale. Advisors often reward creativity more than disciplined elimination. So the company carries too many half-alive concepts, too many “maybe” pivots, and too few hard criteria.
That is how mediocre ideas stay on life support long enough to consume roadmap, narrative, and capital.
A concrete method
Method: generate wide, filter cold, kill fast
Criteria before passion. Predefine: pain size, ability to pay, channel access, plausible differentiation, legal or technical timeline. Simple score per dimension; reject mushy averages.
Time-box each idea. Give every candidate a short window (for example one week of discovery) before go / iterate / kill. No automatic extension without a new signal.
Pair reviews. Have a second person score the same idea independently before resources move; divergence in scores is a signal to redesign the test, not to average opinions into mush.
Team rituals. Weekly “idea graveyard” meeting: one idea buried, one lesson written. Celebrate kills as much as wins to normalize quitting.
Portfolio mindset. Keep one primary idea and one or two backups explicitly labeled “awaiting proof.” No parallel building without dedicated resources.
False-positive journal. When an idea felt good then failed, log the bias (confirmation, friendship, vanity). You sharpen your personal filter.
Anti-patterns. Watch ideas that never die because they are “almost ready”: endless extension usually means nobody dares decide. Require a date or an internal sponsor who owns the risk; otherwise park the idea with explicit wake-up conditions instead of letting a ghost project consume attention.
Example
Example: aborted B2B marketplace
A team imagines a marketplace between industrial subcontractors and buyers. Brainstorming yields dozens of niches. They run the grid: weak channel access (hostile intermediaries), long cycles, compressed margins. Four niches remain; three die in five days of calls (no urgency). The fourth shows unsolicited follow-up emails—a weak but coherent signal.
They still kill the generic marketplace and pivot to a workflow tool for one subcontractor type. “Ninety percent” here was not an insult: it was the list of blind alleys they avoided coding. The founder calls that week the most productive of the year because unwritten code carried no debt.
Without accepting a high bad-idea rate, they would have merged niches into a fuzzy vision nobody could buy.
Pivoting to a single workflow tool let them reuse learnings—onboarding process, domain language—without throwing away all cognitive effort spent on the marketplace. Killing a bad idea can preserve intangible assets when you document what the market taught you about real user constraints.
Share the kill memo with advisors briefly: external eyes pressure-test whether you quit too early or exactly on time, and they learn how you make trade-offs under uncertainty.
What to do now
List the ideas, features, or wedges you are still protecting emotionally. Then ask one cold question for each: what observable event in the next 7 days would prove this deserves more time?
If you cannot name that event, the idea does not need more love. It needs a graveyard or a smaller test.
Related reading
Lumor was built for this filtering job: 13 AI roles attack weak assumptions from different angles so you can kill faster, learn faster, and protect time for the ideas that deserve to survive.