Guide
How to validate a startup idea before you raise
Last updated: November 8, 2026
Written by the Zylar team. How we research and check our guides
The short answer
Before you pitch, collect evidence for the questions you'll be asked anyway: who has the problem badly enough to pay, who already sells a fix, whether the hard part can be built, and how you'll reach the first customers. Then run the cheapest test that could prove your riskiest assumption wrong, before you spend a raise on building.
- 6
- questions you'll be asked anyway, with the evidence for each
- 10 to 20
- problem interviews for a first round
- 7.9%
- of drug programs that entered phase I trials reached FDA approval (BIO, 2011 to 2020)
The questions you'll be asked anyway
| Question | Evidence that answers it | A cheap way to get it |
|---|---|---|
| Who has this problem, and how badly? | People describing the last time it happened and what they did about it | 10 to 20 problem interviews |
| Will they pay? | Money or a signed commitment with a price on it | A paid pilot, a deposit or a letter of intent |
| Who already sells a fix? | Named rivals, their prices and what customers complain about | An afternoon mapping competitors |
| Can the hard part be built? | A prototype, a bench result or a clear regulatory path | The one experiment that could kill the idea |
| How will you reach customers? | Your first 10 customers by name or channel, and what each cost to reach | A small channel test |
| Why you, and why now? | What changed recently: a rule, a cost, a technology | A short note you can back with sources |
Run the experiment that could kill it first
Every startup idea rests on one assumption that sinks everything if it's false. Find it and test it before anything else. It's usually in a different place depending on the kind of startup:
- Software: will customers switch from what they use now, and pay your price? In our example run of an AI scribe for vet clinics, the agents found about 20 rivals already on sale, most of them cheaper.
- Hardware and health: is there a regulatory path you can afford? A self-fitting hearing aid kit, in the same examples, needs FDA 510(k) clearance before it can be sold.
- Services: does each job leave a profit once you count travel and labor? In the EV charger repair example, the agents' main question was whether a 48-hour repair promise still pays at $35 per charger a month, and the next step was three paid pilots that track the real cost.
- Deep tech and biotech: does the science work outside the lab, at the speed and cost a customer needs? Market research can't answer that; only a bench or field result can.
Deep tech and biotech carry the longest odds on that last question. Of drug programs that entered phase I trials from 2011 to 2020, 7.9% reached FDA approval, according to BIO's clinical development success rates study. The science test comes first, and nothing else can stand in for it.
Where AI helps, and where it can't
AI is good at the desk research before you raise: listing competitors and their prices, finding the rules that apply, and flagging what published research says is still unproven. Ask it to look for the reasons the idea fails, since chatbots tend to agree with you. OpenAI rolled back a ChatGPT update in April 2025 for being overly flattering and agreeable.
AI can't replace a customer paying you, a lab result or an expert's review of your patents and regulatory path. Use it to decide which of those to get first.
FAQ
Do I need a prototype before I validate? No. Most of the questions above can be answered with conversations, a competitor map and a paid pilot offer. For deep tech, the prototype or bench result is often the validation, so plan it as the first experiment.
How many customer interviews are enough? Plan 10 to 20 for a first round and stop when the answers start repeating. If you keep hearing new problems, your customer is too broad.
Is a letter of intent as good as a sale? No, but it's better than a compliment. A letter of intent with a price and a start date shows intent. A deposit or a paid pilot shows they'll pay.
Can Zylar tell me whether investors will fund my idea? No. Zylar checks the idea: demand, competitors, how hard it is to build and the risks, with a verdict and what to prove next. Fundraising, valuation and legal advice stay with you and your advisers.
Zylar is being built to do the desk research for you: AI agents check the market, competitors, build and risks, cite their sources and score the idea out of 100, then tell you what to prove first. See three startup ideas we ran through Zylar. It's in early alpha. Join the waitlist.