Let's be honest with each other for a second.
You have a startup idea. Maybe you've been dreaming about it for months. Maybe you've already bought the domain, sketched the wireframes, told your friends at a dinner party. And everyone said "that sounds cool!"
But here's the thing nobody will tell you: most startup ideas are terrible. Not just "needs more polish" terrible — but fundamentally, structurally, irredeemably bad. And the worst part? The people closest to you are the least likely to tell you the truth.
That's why we built DeathScore — an AI that does what friends won't: tell you the unvarnished truth about your startup idea. We recently analyzed 50 startup ideas submitted by real founders and scored them on a 0-100 scale. The results range from "embarrassing" to "actually promising." Let's get into it.
The Methodology: How DeathScore Roasts Ideas
Before we get to the roast, here's how it works. DeathScore evaluates every startup idea across five dimensions, each scored 0-20 for a total of 0-100:
After the score comes the roast — a brutally honest, AI-generated takedown — and a verdict on whether to proceed, pivot, or go back to the drawing board.
Here are 5 real examples from our database. Names and specifics have been changed to protect the embarrassed.
☠️ Example #1: "AI for Everything" — Score: 11/100
The idea: "An AI-powered platform that uses machine learning and natural language processing to help businesses optimize their workflows through intelligent automation powered by AI."
We see this one constantly. It's the startup equivalent of saying "I want to make a website." Except it's 2026, and you just described 40,000 existing products.
The lesson: Specificity is your startup's only lifeline. "AI for X" isn't a business. "AI that automates commercial lease reconciliation for property managers with >50 units" — now you're talking. The more specific you are, the less competition you have.
😬 Example #2: "Uber for Dog Walking" — Score: 23/100
The idea: "A mobile marketplace connecting dog owners with vetted dog walkers. On-demand, with GPS tracking, in-app payments, and real-time photo updates."
Ah, the classic "Uber for X" pitch. It worked in 2014. It's 2026 now. You're competing with Rover, Wag, and about 400 local services that already do this — plus the fact that most dog owners already have a walker they trust.
The lesson: A marketplace needs a liquidity moat. If both sides (walkers and owners) can easily switch to a competitor — and they will — you're running on a treadmill. Ask yourself: 10 years from now, why will users still be on your platform and not a clone?
✅ Example #3: API for Commercial Real Estate Lease Abstraction — Score: 82/100
The idea: "A developer API that extracts key data points from commercial real estate lease PDFs — rent escalations, termination clauses, renewal options — and returns structured JSON. For proptech startups and CRE software vendors."
This one is boring on the surface. But it's exactly the kind of idea that scores high — and builds a real business.
The lesson: The best startups often sound boring at a dinner party. B2B API plays targeting a specific industry pain point have higher survival rates than consumer apps. If you have to explain your market instead of your technology, you're on the right track.
⚰️ Example #4: Failed Startup Autopsy — Juicero with DeathScore X-Ray
The idea (post-mortem): Juicero — a $700 IoT juicer that squeezed proprietary packs of pre-cut fruits and vegetables. Raised $120M from top VCs. Shut down in 2017 after a Bloomberg article revealed the packs could be squeezed by hand.
Let's see what DeathScore's AI would have told them in 2015:
The lesson: Juicero's failure is a masterclass in what DeathScore catches.Three fatal red flags: (1) premium pricing for a commodity problem, (2) unit economics that rely on customers not doing basic math, and (3) a "moat" (proprietary packs) that disappears the moment someone asks basic questions. DeathScoring an idea before building would have saved $120M of investor money.
Want to run your own founder autopsy? 🫣 Put your idea through DeathScore — see what the AI would have told Juicero's founders.
🦄 Example #5: The Dark Horse — Score: 76/100
The idea: "A subscription service that sends parents of newborns a weekly curated box of age-appropriate developmental toys, with an accompanying app that tracks milestones and suggests specific activities based on pediatric research."
On paper, this sounds like another subscription box. But DeathScore gave it one of the highest scores in our dataset. Why?
The lesson: Dark horse ideas score high because they nail the fundamentals: a real problem for a defined audience, subscription economics that work at small scale, and a defensible distribution channel. The best ideas don't need to sound revolutionary — they need to be reliably useful for a specific group of people who have money and urgency.
🔍 What Separates High Scorers from Low Scorers
After analyzing 50 ideas, clear patterns emerged. Here's what separates the unicorns from the zombies:
☠️ Low Scorers (0-40) Shared These Traits:
- Vague problem statement — "helping businesses grow" is not a problem
- Zero specificity — no particular user, no particular workflow, no particular pain point
- "AI for X" without the X — the solution is the star, not the problem
- Consumer marketplace in 2026 — unless you have a billion-dollar distribution insight, this is suicide
- Unit economics that need a miracle — freemium with 0.5% conversion and $100 CAC
- No defensibility answer — "we'll execute better" is not a moat
✅ High Scorers (60-100) Shared These Traits:
- Boring but painful — a specific, recurring, non-sexy problem in an existing industry
- Definable buyer persona — "property managers with >50 units" vs. "businesses"
- API-first or B2B focus — higher LTV, lower churn, clearer distribution
- Data moat — the product gets better with every use
- Unit math that works at 100 customers — not just at 100,000
- Distribution channel pre-identified — they knew exactly where first users would come from
The single biggest predictor of a high score? The founder's ability to describe their user's current workaround. If you can tell me exactly how people solve this problem today (spreadsheet, sticky notes, yelling at a contractor), you understand the problem. If you can't, you don't.
62% of low-scoring ideas couldn't describe the user's current workaround. 90% of high-scoring ideas described it in painful detail.
📊 The Full Breakdown: Where Ideas Die
Of the 50 ideas we analyzed, here's how they broke down:
- 0-20 (Dead on arrival): 14 ideas (28%) — Vague concepts, no market, no specificity
- 21-40 (Zombie territory): 24 ideas (48%) — Some merit but fatal flaws in monetization or differentiation
- 41-60 (Needs work): 9 ideas (18%) — Real potential but missing key execution details
- 61-80 (Promising): 2 ideas (4%) — Solid fundamentals, worth building
- 81-100 (Go build): 1 idea (2%) — The API-for-CRE-lease-abstraction idea
The distribution is brutally honest: 76% of ideas scored below 50/100. Most startups are dead before they start — and most founders never get the feedback to realize it.
🎤 Your Turn: Get Roasted by AI
Reading about other people's bad startup ideas is fun. But what about yours?
We built DeathScore because we got tired of watching founders burn years on ideas that could have been killed in 30 seconds with honest feedback. The AI doesn't sugarcoat. It doesn't care about your feelings. It just evaluates your idea across the five dimensions and tells you what's wrong — so you can fix it before it's too late.
☠️ What's Your DeathScore?
Describe your startup idea in 30 seconds. Get a brutal, honest score and a personalized roast. Free. No signup. Nothing to install.
Get Your DeathScore →P.S. — The best founders don't get discouraged by a low score. They iterate. A score of 23 today becomes 67 next week after you narrow your focus. DeathScore isn't the verdict — it's the first honest conversation you've had about your idea.