Validate demand before you build

LAST REVIEWED 2026-08 · SOURCED FROM 1 SESSION, AUG 2026

TL;DR

Think slow, act fast: most of the work happens before anything gets built. Write a numeric hypothesis — X% of a defined market will pay $Z — and test it with a cheap, low-fidelity prototype until it comes back true 70%+ of the time. Money is the only real signal; interviews and surveys are nearly worthless. Then run a fatal flaw analysis on every predecessor who died on this hill, and pressure-test the deck with investors before you write real code. If your LTV-to-CAC math can’t clear 4:1, push the big red pause button.

Think slow, act fast

The session’s frame, from a serial founder who built a category-defining e-signature company: roughly 99.5% of projects fail, and the majority fail because founders act too fast up front — building before they’ve done the thinking that would tell them whether to build at all. “Acting” is building and delivering. Everything before that is the thinking phase, and it’s where confidence is earned: product, pricing, packaging, and market all have to point the same direction, because if any one vector is off, the endpoint is wrong.

The XYZ hypothesis

Borrowed openly from the book The Right It: state your belief as X% of market Y will do Z — where Z must contain a dollar figure and, for long-sales-cycle businesses, a time bound. “5% of mid-size logistics operators will pay $5K within 60 days.” Then build the cheapest possible artifact that can test it — today, that’s an AI-generated demo built in hours, not an engineered product — and collect fresh, first-hand data from real target customers. Not survey data, not industry reports, not your own expertise.

“Money is the most important signal. They better be willing to write you a check for something, or you probably aren’t on the right path.”

— Serial founder, category-defining e-signature company · session, Aug 2026

The bar: the hypothesis should come back true 70–90% of the time across enough prospects to matter. While testing, don’t collect full price — anchor the real number and discount it: “This is a $1,000 product; because you’re early, pay $100 now, and respond to me as if I’m going to charge you $1,000.” The response to the full number is the data.

What you are explicitly not testing yet: whether you can build it, or build it cheaply. Engineering answers come after demand answers.

The math gate: 4:1

Before committing, run the ratio: expected lifetime value against acquisition cost. The rule of thumb quoted in the room — investors won’t look below 3:1, and his own bar is 4:1, because AI-era variable costs (tokens you can’t control) eat headroom the old software economics never had to give up. A hypothesis whose numbers can’t clear that ratio isn’t a business, no matter how much people like the demo.

Fatal flaw analysis

Once the XYZ tests positive, find everyone who tried something similar and died. Look them up, call them — a call, not a text thread. Founders and executives of failed companies almost always talk, almost always explain what killed them, and almost always refer you to more people if you ask. Your job is then to solve every flaw that killed a predecessor, because unsolved, it kills you on the same hill.

Pressure-test with investors — before you’re raising

The closing step: build a deck (still no product) and walk it past a few investors with no ask attached. You’re not pitching for money; you’re asking how the thing reads as a fundable opportunity. Two things happen: they ask the hard questions you’ll need answers to later, and they surface predecessors you missed — “someone pitched me this two years ago” is a gift; ask for the introduction and add them to the fatal flaw list. Investors read that behavior as signal, and the coffee meeting sometimes turns into a warm intro you didn’t ask for.

A caution from the room, seconded by the founders present: every investor’s take is filtered through their own biases. Run it past several, and treat their input as data about how the story lands, not as a verdict on the idea.

Two words of warning

“Design partner” means unfinished product looking for an early buyer — that’s what an investor hears, and it’s what it is. This validation process gets you to real data faster than a vague design-partner arrangement, and any partner worth having should still put money in (see First customer contracts).

Time counts as price. If every prospect takes months to respond, the Z fails on its time bound. One founder in the room spent nine months on a $100K deal that resolved to nothing — the check-sized interest felt good, but it never converged. Runway is part of the equation.

Numbers from the room

Figure	Value
Project failure rate cited	~99.5%
Hypothesis pass bar	True 70–90% of the time
LTV:CAC to proceed	4:1 (3:1 is the investor floor)
Test discount vs. anchored price	e.g., 90% off, full price stated in writing
Value advantage to target vs. alternatives	~10x rule of thumb; less in big-ticket categories

Sources

One working session (Aug 2026) with a serial founder — multiple companies, including a category-defining e-signature company — walking six founders through his validation methodology. Published work referenced: The Right It (Alberto Savoia).