Write the assumption as a sentence that could be false, write its opposite underneath and ask what you would expect to see if the opposite were true. Most assumptions survive that exercise in a minute. The few that do not are the ones worth a day of investigation, which is why the method starts with a list and ends with one check.
The difficulty is rarely a shortage of doubt. Anyone who has carried an idea for a few weeks can produce objections on demand, and producing them feels like rigour while changing nothing. The method below separates the assumptions that need an outside look from the ones your own memory can settle, and it is drawn from reasoning about how plans fail rather than from a study, so treat its numbers as a worked illustration.
Write every assumption down before you argue with any of them
Take an invented plan: a freelance designer intends to spend two months building a library of website templates for small charities, on the belief that charities would rather buy a template than hire a designer. Before challenging anything, the designer lists the sentences that have to be true for the plan to pay off. The useful trick is to go through the plan noun by noun and ask who says so about each one, because the assumptions that sink a plan are usually the ones nobody has stated out loud.
Seven sentences came out of that pass. They appear in the table further down. Some were obvious, such as the claim that two months is enough. Others had been carried without ever being said, such as the claim that small charities build their own sites at all, which is the assumption underneath the word template. The list is the whole point of this first step. Arguing with an assumption before the list exists means arguing with whichever one came to mind, and the one that comes to mind is usually the one you already half doubt.
Rank them by the damage if false and the days to check
Each assumption gets two numbers before any of them is challenged. The first is the damage to the plan if the assumption turns out to be false, on a scale of one to five where five means the plan has no point. The second is a guess at the working days needed to find out, counting the time to reach people as well as the time to talk to them. Both are judgements, and in the example they are the designer’s own, so the chart below shows opinions laid out for inspection rather than measurements.
| No. | Assumption | Damage if false | Days to check | Where the opposite landed |
|---|---|---|---|---|
| 1 | Small charities build or rebuild their own websites rather than paying someone | 5 | 2 | Needs a look |
| 2 | The person who chooses the website tool cares most about price | 3 | 2 | Needs a look |
| 3 | A charity would trust a template from a designer it has never heard of | 4 | 5 | Supported by something already known |
| 4 | A template saves a charity more hours than it takes to learn | 4 | 10 | Needs a look |
| 5 | The templates can be sold without a conversation | 2 | 15 | Supported by something already known |
| 6 | Two months is enough to build the library | 2 | 1 | Answered from memory |
| 7 | Charity websites are alike enough for one library to cover most of them | 5 | 3 | Needs a look |
The seven invented assumptions behind the template plan, with the designer’s own damage rating, days to check and where each opposite landed.
Two points sit high and to the left: assumption 1, that charities build their own sites, and assumption 7, that their sites are alike enough to share a library. Each would end the plan if false. Each could be checked within three working days. Assumption 4, the one about hours saved, would also hurt but needs charities to use a template first, which puts it ten days out and behind the other two in the queue.
Write the opposite and see where it lands
Now each assumption gets its opposite written underneath it, in full, followed by one line on what the world would look like if the opposite held. The opposite of assumption 1 reads: small charities mostly pay someone to build their site, so a template would compete with the person they already trust. A written opposite lands in one of three places, and the chart below counts where the seven in the example ended up.
| Where the opposite landed | Count of 7 | Assumptions |
|---|---|---|
| Needs a look | 4 | 1, 2, 4 and 7 |
| Supported by something already known | 2 | 3 and 5 |
| Answered from memory | 1 | 6 |
Where each of the seven written opposites landed in the invented example, with the assumptions in each group.
An opposite that is answered from memory is one you can dismiss in a sentence with evidence you already hold. The designer has built a smaller library in six weeks before, so the two-month estimate stands for now. The reverse case is an opposite supported by something already known, and it is the most useful thing the exercise can turn up. Every client the designer has ever had arrived by referral, which argues against assumption 3, the claim that a charity would trust a stranger’s template, and against assumption 5, the claim that no sales conversation is needed. Those two had been sitting in the plan next to the evidence that contradicted them, and nobody had put the two side by side.
The third landing place is the one that matters most for what to do next. Four of the seven opposites could not be settled from memory at all, because the designer’s experience comes from charities that chose to hire a designer. That experience cannot say what the charities who never hired anyone do. No amount of thinking will supply the missing sample. The objection worth investigating is one that sits here and is cheap to check.
Why arguing with yourself tends to fail
The psychologist Peter Wason published an experiment in 1960 in which people had to discover a hidden rule about sets of three numbers by proposing their own sets and being told whether each fitted. Most participants proposed sets that fitted the rule they already had in mind and announced that rule with confidence, and only a minority tried a set built to break their own guess. The task was small and artificial, and its sample was a few dozen students, so it cannot say how often people with a business plan behave the same way. It does show the shape of the problem. Asked to test a belief, people gather examples of it.
Writing the opposite is a counter to that habit, because a written sentence can be examined for what it predicts and a vague feeling of doubt cannot. The written form also exposes a comfortable kind of self-challenge, where the assumption chosen for attack is the one you were already uneasy about and the load-bearing one goes unexamined. In the example the designer’s first instinct was to worry about assumption 6, the timeline, which carries a damage rating of two. Challenging it would have felt like diligence. It would also have left assumption 1 untouched, with the whole plan resting on it.
Pick one objection and give it a check
With the opposites sorted, choose the single objection that scores highest on damage and lowest on days, and give it a check with a stopping result written before you start. For the example that is assumption 1. Contact twelve small charities that have never been clients, ask who built their current website and who would build the next one, and count the answers. If nine or more of the twelve say they paid someone and would pay again, the template library is the wrong product for this buyer and the plan stops. Fewer than nine leaves the assumption standing and moves the next check to assumption 7.
Twelve charities cannot estimate how the whole sector behaves, and the ones who reply to a stranger’s message may differ from those who do not. The check screens for a gross error in the plan, which is what a first challenge should do. The earlier note on turning a vague idea into something you can test covers how to write a stopping result so that the answer cannot be bent after it arrives, and the same discipline applies here.
One objection at a time is a deliberate limit. A list of seven assumptions invites a person to challenge all of them in one sitting, which produces seven half-arguments and no contact with anyone outside the room. If you find yourself on the third sitting of rewriting the same list, the note on stopping the loop on a single idea describes how to make each sitting produce something written and when to stop.
What to do this week and what stays unknown
This week, take one plan you are carrying and write the sentences that must be true for it to work, aiming for between five and ten. Give each a damage rating and a day count, write the opposite under each one and sort the opposites into the three groups above. Then send the first message for the one check that scores highest, and write down beforehand the count that would make you stop.
What the method cannot tell you is whether your damage ratings were right, because they were made by the person with the most to lose from a low score. The invented example gives the designer seven assumptions and a tidy ranking, and a real plan may produce fifteen with three tied at the top. Where that happens, check the cheapest of the tied ones first, and expect the result to change the ratings on the rest.