The pipe test for a new idea

I get more ideas than I can build, which is a normal condition and not a boast — the constraint on a portfolio is never idea supply. Choosing badly is the expensive failure, and it's expensive in a slow way: you don't find out for a year, and the year is gone regardless.

So I run everything through five questions first. They're all versions of the two laws, applied before there's anything to look at.

1. What flows through it?

Name the unit. A signed contract. A resolved ticket. A released track. A completed tax filing.

If the answer is abstract — "we help teams collaborate better" — there's no pipe yet. There's a hope about an outcome. You cannot measure, instrument, price, or improve a thing you can't count, and the inability to name the unit at the idea stage reliably predicts the inability to name it two years in.

2. What are the fittings?

What does this connect to on either end? Where does the input come from today, and where does the output go?

A product with no defined fittings has to become someone's whole world to be adopted, which is a much harder sell than being a segment they can snap in. The best businesses I've been part of attach to something already flowing. The worst asked everyone to redirect their existing flow into a new system first, and were rewarded with pilots that never expanded.

3. What's the gauge?

What single number tells you it's working — and could you show it to a customer without translation?

If the only available answer is a vanity number (signups, usage, "engagement") the value proposition isn't clear enough yet. Real products have a number that a customer would recognize as a description of their own life getting better: days saved, dollars found, errors avoided.

4. What goes dark if this succeeds?

The one people skip, and the one that has saved me the most.

Every automation replaces some human observation, and that observation was doing work nobody wrote down. If this idea works at scale, what will nobody be watching anymore? What will you find out late that you currently find out early?

If you can't answer, you're not ready to build it. If you can answer and the answer is frightening, that's your first gauge — build it before you build the feature.

5. What's the smallest complete segment?

Not the smallest feature. The smallest thing with a real In, a real Out, and a gauge on the joint.

An MVP that does half a job for everybody is worse than useless: it produces feedback about a thing you'd never ship. A complete segment that handles one narrow case end to end produces feedback you can act on, because someone actually ran work through it and either got value or didn't.

The distinction is: does the unit come out the other end? If yes, it's a segment, however narrow. If no, it's a demo.

What the test is really doing

Every question is the two laws in disguise.

What flows, what are the fittings, what's the smallest complete segment — that's pipes connect. Can this thing exist as a piece of a larger system, or does it demand to be the system?

What's the gauge, what goes dark — that's glass reveals. Will you be able to see whether this is working, and will it make anything else harder to see?

Twenty minutes, five questions. Most ideas die on question one, which is a gift — a bad idea killed at the whiteboard costs nothing, and the same idea killed by the market costs a year and a team's belief that you know what you're doing.