Bad-Good Ideas
Deliberately entertain ideas that sound stupid on first telling, because embarrassment is a poor proxy for market demand.
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 80%
Shaan Puri's core distinction on the Drunk Ideas format is between ideas that are simply bad and ideas that are 'bad-good': ones where the honest first reaction is 'that's the stupidest thing I've ever heard' but which contain, in his words, a nugget of gold. His canonical proof is the woman selling farts in a jar, who made roughly $500,000 — the pitch is humiliating to say and the business works anyway. The framework's real function is procedural: the 'drunk' framing (neither host actually drinks) is deliberate air cover that licenses saying silly things out loud, because the ideas that die are the ones killed silently between having them and voicing them. Bad-good is explicitly not a quality verdict — it is a category of idea worth surfacing so it can be evaluated on evidence rather than on how it sounds. The pattern repeats across the episode: the private photo-vault app that sounds juvenile turns out to be a top-100 App Store product with tens of millions of downloads, and a virtual-Santa studio that sounds like a joke sells out at $35-75 per fifteen-minute visit.
Origin
Shaan opened the second Drunk Ideas episode admitting the ideas he had were bad — but not all bad, they were 'bad good' — and built the whole recurring format around giving that category permission to be spoken.
Core principles
- 01Embarrassment is not a demand signal — the two get confused constantly.
- 02The ideas that die are killed silently, before anyone hears them.
- 03A stupid-sounding pitch and a working business are fully compatible.
- 04Manufacture social permission (a format, a framing) to say the silly thing.
- 05Bad-good is a category to surface, not a verdict to act on.
How to run it
- 1
Create the air cover
Set up a framing — a recurring segment, a 'drunk ideas' label, an explicit no-judgement round — that makes saying a stupid-sounding idea socially free. Shaan notes neither host actually drinks; the framing is the mechanism, not the alcohol.
- 2
Dump the list without filtering
Get every idea out in one flood rather than pre-screening. Shaan describes the ideas all arriving at once and simply dumping them all out.
- 3
Sort bad from bad-good
For each idea, ask whether it is merely bad or whether there is a nugget of gold under the silly surface. Only the second category earns further attention.
- 4
Name the underlying truth
Articulate the real human impulse the idea is serving — privacy, voyeurism, status, novelty. If you cannot name it, the idea is just bad.
- 5
Hunt for existing proof
Search for a product already quietly serving that impulse and pull real numbers: App Store rank, review count, download range, pricing. Evidence, not taste, settles it.
In the wild
Shaan's shorthand for the whole category: a woman selling farts in a jar is the stupidest-sounding pitch imaginable, and telling someone about it produces immediate ridicule.
→ She made roughly $500,000 — the ridicule and the revenue coexisted.
Shaan pitched 'Hideaway', a disguised private camera roll behind a fake bookshelf or calculator app. Sam searched and found Private Photo Vault already existed — reviewed roughly 807,000 times, ranked 86 in the App Store, 10M+ Google Play downloads alone, run by seemingly one person, with a ~$50/year premium tier and even a decoy-password feature Shaan hadn't thought of.
→ A silly-sounding pitch was validated in minutes against a product plausibly grossing tens of millions of dollars lifetime.
Sam surfaced santasclub.com — virtual 15-minute Santa visits, started 2020, operating only two months a year, roughly 15,000 visits in year one at $35-75 per visit.
→ Roughly $500,000 to $1M in two months, and sold out with a pre-sale — a joke-sounding idea Sam graded 'not even a good bad idea, simply a good good idea'.
Common mistakes
Treating embarrassment as market research
The feeling of 'I can't say this out loud' is about your social risk, not about whether anyone will pay. Conflating them systematically removes the highest-variance ideas from your list.
Pre-filtering before the list exists
Screening while generating collapses the pool to safe, respectable ideas. Generation and selection have to be separate passes.
Stopping at 'bad-good' and never checking evidence
The label is permission to look, not a conclusion. Without pulling real download, review, and pricing data, you have swapped one gut call for another.
Is it for you?
Best for
Founders and operators whose idea pipeline has gone sterile because they only write down ideas that sound respectable.
Not ideal for
Anyone who needs to pick and commit to one idea today — this is a generation filter, not a selection filter.
From the transcript
“these are ideas that i know are not great but there's some nugget of gold in them there's some truth to them and so they…”
“it's like the lady who sells farts in a jar like when you're telling someone about it you're like that's the stupidest thing i've ever…”
“i think being drunk is like air cover to saying some silly stupid things neither of us i don't drink sam doesn't drink”
From the episode
Drunk Ideas: Only Fans For Therapy, A Secret Bookcase for your Phone, and More
Drunk Ideas