You already know. Nobody arrives at this page needing to be told their idea is too broad — they arrive having noticed, and what they get back is the advice they already gave themselves: narrow it down. Pick a niche. Find your ICP.
The part that is actually decidable is what a narrowing has to contain before it counts as one, and whether a narrowing exists at all for you specifically. Those are different questions, and only the second one can come back as no.
Two answers, and one of them is a refusal
The analysis returns either a wedge or a refusal, and the refusal is the reason the mode is worth anything. The instruction is blunt about it: if every remaining slice still needs enterprise sales, still needs a conversation, or still runs through a platform this founder said no to, then say so — and do not invent a fake wedge.
That last clause exists because inventing one is easy and rewarding. A plausible narrowing can be generated for any idea in a sentence, it reads like insight, and the founder cannot tell it apart from a real one until several months have gone.
The refusal reads your profile, not just your idea
Here is the thing that makes "how to niche down" posts unable to answer this. Of the blockers that trigger a refusal, most are not properties of the idea:
| Blocker | Read from | What it means here |
|---|---|---|
| Enterprise sales | the idea | Every remaining slice buys through procurement. |
| Mandatory calls | the founder profile | No slice can be sold without talking to someone. |
| An intolerable platform | the founder profile | The only route to a buyer runs through a platform this founder said no to. |
The refusal fires only when EVERY narrowing hits one of these. The middle column is the point: the same broad idea can produce a wedge for one founder and a refusal for another.
So "is this too broad" has no answer in the abstract. An idea that narrows beautifully for someone who will get on calls has no wedge at all for someone who will not, and a generic post cannot know which you are. This is also why the profile is treated as binding rather than as a preference: a narrowing you refuse to execute is not a narrowing.
What a narrowing has to contain to count
When a wedge does exist, the output is asked for 7 things. Not all of them are mandatory, and the split is worth seeing, because it is the difference between a rule and a wish.
narrowed_wedgeenforcednew_buyer_segmentenforcednew_one_linerenforcedanti_positioningenforcedNot a list copied out of a prompt. Each field was blanked in turn and fed to the invariant that guards real output; these are the ones it refuses to pass. If the rule changes, this figure changes.
Read the second one again: a named buyer. Not "small businesses", not "teams" — a description specific enough that you could go and find twenty of them this afternoon. The most common fake narrowing is a broad idea with a narrower adjective in front of it, and the named-buyer requirement is what that fails.
And the honest other half, since we are publishing the derivation: 3 of the 7 are asked for and not policed.
diagnosisWhy the original idea was too broad.removed_risksThe risks the narrowing takes off the table.validation_hintOne async next step to test the wedge.Requested by the contract, and a missing one will not stop the output. Stated because a page claiming an enforcement guarantee should say exactly where the guarantee ends.
The refusal branch is even thinner. It requires a reason and nothing else — checked the same way, by building a refusal with every other field empty and watching the invariant let it through. A one-sentence no is a legitimate output here, which is the correct design and also worth knowing before you receive one.
One complete before-and-after
The valid sample used to probe that invariant starts broad: a buyer spanning several industries with no shared workflow. Here is the narrowing it accepts:
| Buyer | Independent bookkeepers with 10-40 clients. |
|---|---|
| One workflow | Month-end close checklist for independent bookkeepers. |
| Explicit exclusion | Not a general accounting suite. |
| First test | Publish the checklist and count signups. |
The same valid sample the invariant tests. The narrowing changes buyer, workflow and exclusion; it does not merely add an industry adjective.
Anti-positioning is the part that costs you something
Of the mandatory fields, the one people quietly skip is the explicit "what this is not". It is mandatory for a reason: a narrowing you have not stated as a cut is not a narrowing, it is an emphasis. Everything you kept in reserve is still in the product, still in the roadmap, still in the conversation with the first customer who asks for it.
Writing the anti-position down is what makes the next twelve decisions cheap. Someone asks for the adjacent feature, and the answer already exists.
What this analysis will not tell you
It does not re-score your idea. The projection is deliberately neutral: the decision column comes back UNKNOWN, the score is a placeholder 50, and the confidence is fixed at MEDIUM. A narrowing recommendation is not a verdict, and letting it produce one would mean scoring an idea nobody has evaluated yet.
If you want the wedge judged, that is a separate pass on the narrowed idea, which is the honest sequence: narrow first, then score the thing you actually intend to build.
The next step it suggests is also constrained. It has to be async — no call, no interview, no demo — and the output is scanned against 24 banned phrases before you see it. Not a style preference: a validation step that requires a conversation is a step most people never take.
Narrowing it yourself
- Name a buyer you could list. If you cannot write down twenty specific instances of your buyer, you have a category, not a segment.
- Cut to one workflow, not one industry. "For dentists" is a market, not a wedge. "The bit where the receptionist reconciles insurance claims" is a wedge, and it might sell to four industries.
- Write the anti-position first. Before the pitch. It is faster, and it tells you immediately whether you are actually giving anything up.
- Check the narrowing against yourself. If the only reachable buyer needs a phone call and you will not make phone calls, the wedge does not exist. That is a finding, not a failure of nerve.
- Then test it async. One page, one real call to action, and a count — the no-call method is the same one this mode's suggested next step is restricted to.
The short version
- "Too broad" is a diagnosis you already made. The decidable questions are what a narrowing must contain, and whether one exists for you.
- The refusal branch is the valuable half, and its trigger is mostly read off the founder — the same idea narrows for one person and refuses for another.
- 4 fields are mandatory in a wedge, derived by blanking each and watching the invariant reject it. A named buyer is one of them; a narrower adjective is not a buyer.
- 3 more are asked for and not enforced. Published because a guarantee is only as good as its stated edge.
- It does not score your idea. The decision comes back UNKNOWN by design: narrow first, then judge the thing you intend to build.