Find the one assumption that could sink it.

Behind every startup is a quiet list of things the founder believes but has not proven. That people have the problem. That they will pay enough to make it a business. That you can build the hard part and run it reliably. That you can reach customers for less than they are worth. Most of that list is fine to carry untested for a while. But somewhere on it sits one belief that, if it turns out to be wrong, takes the whole thing down. Finding that belief, and testing it early, is the most valuable thing you can do in the first year of a company.

The discipline that surfaces it is called assumption mapping, and it is simple enough to do on a whiteboard in an afternoon. You list what you are assuming, you score each assumption on two axes, and the riskiest one falls out of the maths. It is not clever. It is just a way of forcing yourself to look at the belief you have been quietly avoiding, which is almost always the one that matters most.

An assumption is a belief you haven't earned yet.

Start by being honest about what an assumption actually is. It is any statement your plan depends on that you cannot yet back with evidence. "Users will churn at under three percent a month" is an assumption. "Enterprises will let a startup touch their billing data" is an assumption. "We can acquire customers through content for under fifty pounds each" is an assumption. The test is simple: if someone asked you to prove it right now, could you, or would you reach for a hopeful story? If it is a story, it is an assumption, and it belongs on the list.

Write them as specific, falsifiable statements, not vague hopes. "People will love it" cannot be tested. "At least a third of the ops managers we speak to will describe this problem as one of their top three" can. The sharper the statement, the easier it is to design something that could prove you wrong, which is the entire point.

Surface them with three questions.

A blank page is hard to fill, so use a frame. Nearly every assumption a startup makes is really an answer to one of three questions, and running down them in turn will shake loose the ones you have been ignoring.

• Desirability. Will they actually want it? That the problem is real, frequent and painful. That your solution is one they would choose. That they will change their current behaviour to use it.
• Viability. Does the money work? That they will pay enough. That you can reach them through a channel you can afford. That they stay long enough to be worth more than they cost to win.
• Feasibility. Can you build it and run it? That the hard technical part is possible. That you can deliver it reliably, at quality, with the team and the money you have.

Go through all three out loud and write down every belief that each one surfaces. You will usually end up with somewhere between eight and fifteen. Do not filter yet. The one you are tempted to skip because it feels obvious is often the one worth writing down first.

Score each on impact and evidence.

Now rate every assumption on two axes, each from low to high. The first is impact: if this turned out to be wrong, how much damage would it do? A wrong assumption that costs you a fortnight is low impact. One that means you have no business at all is high impact. The second is evidence: how much proof do you already have that it is true? A hunch is low evidence. A signed contract, or a pattern you have seen ten times, is high.

Your risk on any single assumption is impact multiplied by uncertainty, and uncertainty is simply the flip side of evidence. An assumption that would sink the company and that you have almost no evidence for is a maximum-risk bet. One that barely matters, or that you can already back with proof, is not worth testing. That multiplication is the whole idea: it stops you fussing over things that are either harmless or already known, and points you at the one that is both dangerous and unproven.

Map them, and read the top-right corner.

Plot the assumptions on a grid: evidence along the bottom, impact up the side. Four regions appear, and each tells you what to do. High impact and low evidence, the top-right, is where your leaps of faith live, and those get tested first. High impact but well-evidenced can be watched, not tested. Low impact and unproven can wait until it matters. Low impact and already known can be ignored entirely. Most founders spend their energy in the wrong corner, polishing things they already know, while the leap of faith in the top-right sits quietly untested.

Diagram · the assumption map

Assumption number one sits top-right: high impact, almost no evidence. That is the leap of faith, and it is where the next experiment goes.

Test the top one, then look again.

Once you have your riskiest assumption, you do not build around it and hope. You run the smallest, cheapest experiment that could prove it wrong, and you match the experiment to the kind of assumption it is. A desirability leap wants conversations or a fake-door test. A viability leap wants a pre-sale or a pricing test. A feasibility leap wants a technical spike or a hand-run Wizard-of-Oz. There is a whole guide to those six in the companion issue on cheap experiments, so I will not repeat them here.

The part founders forget is the last step. When the experiment comes back, the map has changed. The assumption you just tested now has evidence against it, so it drops down the list, and a different one becomes your riskiest. Assumption mapping is not a one-off exercise you do at the start and file away. It is a loop: map, test the top, re-map, repeat. The company that keeps doing that is always spending its scarce time on the single most dangerous thing it does not yet know.

The traps to avoid.

Three failure modes catch almost everyone. The first is testing the comfortable assumption instead of the risky one, because the risky one is scary and you suspect the answer. That is exactly the one to run at. The second is treating your own conviction as evidence. Being sure is not the same as having proof, and the strength of your belief has no bearing on where an assumption sits on the map. The third is mapping once and never again. Do it at the start and you have a snapshot. Do it every few weeks and you have a compass.

None of this replaces judgement, and not every belief needs a test. But if you can name the one assumption your whole plan rests on, and you have run something real against it, you are already ahead of most founders, who discover their riskiest assumption the expensive way: by building the entire company on top of it first.