How do you validate a startup idea before building it?

To validate a startup idea you need one thing: strangers revealing whether they want the product, before you spend months building it. The methods below are ordered by how hard the evidence is — and the last sections cover the signals that feel like validation but aren't, and what to do when the answer is no. (Disclosure: published by BidSurvivor, which appears in method two.)

Method 1: The landing page smoke test

One page: the promise, the price if you dare, and a call-to-action that captures intent (email signup at minimum). Build it in an afternoon. The page is the experiment — every visitor either acts or doesn't, and the ratio is your first real number. Homepage vs landing page covers the mechanics.

Method 2: The ad-click test

Before visitors can convert on the page, the idea's one-line pitch has to earn a click from a stranger — and that's testable separately, cheaply. The classic version buys a handful of search clicks (~$5+ each). The free version is ours: claim an empty slot on BidSurvivor — first claim is $0, every visitor starts with $100 of house credit, no account needed — put your pitch line and link on the board for twelve hours, and read the published click-through numbers. A pitch that can't earn cold clicks at $0 will not earn them at $5, and learning that before building is the entire economics of validation. (Small board, small samples — run it more than once before concluding anything, per the vanity-metrics rule.)

Method 3: The presale

The hardest currency: money before product. A "founding member" price, a deposit, a paid pilot. Ten emails is interest; three payments is validation. Uncomfortable to ask for, which is why it works — discomfort filters wishful thinking on both sides.

Method 4: The concierge test

Deliver the outcome manually to five customers before automating anything. Slowest method, richest information — you learn the workflow, the objections and the price tolerance while "being" the product.

Two afternoons, in order

Methods 1 and 2 fit in two afternoons if you do them in the right order, and the order matters because the second one tells you whether the first one's number means anything.

Afternoon one — the page. Write the promise as one sentence a stranger could repeat. Put a price on it, even a placeholder; a page with no price validates curiosity, not demand. Add one form. Before you publish, write down a threshold: "if fewer than 3 in 100 visitors leave an email, the promise is wrong." Writing it down beforehand is what stops you moving the goalposts afterwards.

Afternoon two — the clicks. Put the same sentence in front of strangers: a free slot, a few dollars of search clicks, a post where your buyers gather. Read two numbers the next morning: how many saw the line and clicked, and how many who arrived did the thing on the page. If the first number is low, the pitch failed and the page was never tested. If the first is fine and the second is low, the pitch works and the offer does not. Those are different problems with different fixes, and running the two afternoons together is the only way to tell them apart.

The fashionable name for doing many of these small tests fast is "vibe marketing"; what vibe marketing actually means separates the useful part from the hype.

Reading small numbers honestly

Validation numbers are small, and small numbers are wide. Three signups from a hundred visitors feels like 3%, but the range of true rates consistent with that result runs from under 1% to roughly 8% — so the same afternoon is compatible with "nobody wants this" and "this is excellent". A single run cannot distinguish them. Two rules follow. Run the test at least twice, on different days, and only believe a result the runs agree on. And prefer a stronger signal from fewer people to a weaker signal from more: three payments beat thirty emails, and a stranger who clicks away from something else beats a friend who signs up to be kind.

The founders who get this wrong are rarely careless; they are hopeful, and hope reads 3 of 100 as 3%. Reading it as "somewhere between 1 and 8, run it again" is the whole discipline.

When a smoke test is the wrong tool

Some ideas cannot be validated from a landing page, and pretending otherwise wastes the afternoon. Products that need trust before trial — anything medical, financial or legal — do not get clicks from strangers whatever the copy says, because the buyer's first question is "who are you" rather than "what does it do". Two-sided marketplaces have to validate both sides, and a page only tests one. Hardware needs a price that depends on a manufacturing quote you do not yet have.

For those, the method is conversation, done properly: ask people about their past behaviour, not their future intentions, and never pitch. The best short guide to that discipline is Rob Fitzpatrick's The Mom Test, and Y Combinator's library has a short talk on talking to users that covers the same ground for free. Interviews are slower than a page, but for these ideas they are the only instrument that returns a real number.

The signals that lie

Compliments ("I'd totally use this") — social kindness, not intent. Waitlist signups from friends. Upvotes — a launch spike measures curiosity, not demand. And your own conviction on a good day. The test for any signal: did a stranger pay a cost — money, an email address, a click away from something else? No cost, no signal.

What to do with a no

A no from a smoke test is a no to the sentence, not necessarily to the idea, and there is an order to what you change next. First, the pitch: write three different lines for the same product and run each — most ideas are described badly before they are proven bad. Second, the audience: the same line in front of a different group of strangers, because "nobody wants this" often means "the people I showed it to do not". Only third, the idea. Change one thing per run, and keep the runs cheap enough that a no costs an afternoon — the one-day campaign is the right size for exactly this.

Validation doesn't have to prove success — it can't. It only has to make building the wrong thing expensive to ignore. Two afternoons and $0, using methods 1 and 2, buys that much certainty before the first line of code.