You did the work properly. The criteria went down before you looked at anything, the weights were set in the quiet, every option got scored against every criterion, and the arithmetic came back with a winner.
And you do not believe it.
That is not a failure of nerve, and it is usually not a failure of the method either. It is a correct reading of a result most people misread. This is what to do about it.
First: how close is “close”?
Express the gap between your top two as a share of the winner’s score, not as raw points. Points do not travel — everybody’s scale is different — and a percentage is comparable across decisions and across people.
In the worked decision that runs through How to Make a Big Decision, the numbers come out like this:
Every figure below comes from one invented household’s worked decision, printed so the arithmetic can be followed — the method travels, the figures do not.
| Option | Weighted score | Behind the leader |
|---|---|---|
| Stay and extend | 3.95 | — |
| Trade up across town | 3.84 | 2.8% |
| Move out of the region | 3.01 | 23.8% |
Two things are true at once here, and only one of them usually gets noticed. Third place is settled — it is more than eight times further back than second is, and that option can stop coming up in the evenings. First place is not settled at all.
A margin under about five percent should be treated as a tie until proven otherwise; above about twenty percent you have a genuine winner and can stop reading. Those bands are a working convention the method uses so the next step can be computed from something — they are not a measured finding.
Second: is the margin smaller than your own error?
Look at the scores that produced the gap. They were assigned by judgment — how well the space fits how you live, how good the schooling is, how much you would miss being near people. Ask yourself honestly how far you would move any single one of them if somebody made you defend it out loud.
In practice most people will concede half a point without much argument. A 2.8 percent margin is far smaller than that. The lead is inside the error bars on its own inputs, which means the ranking is not evidence about the options — it is arithmetic performed to more precision than the inputs deserve.
Third: compute the flip distance
This is the part almost nobody does, and it is the part that turns “I don’t trust this” into something you can act on.
For each criterion, ask: how far would this criterion’s weight have to move before my top two swap places? Move it up or down, rescale the other weights in proportion so the total stays a hundred, and find the point where the two totals meet. You can do it by hand — nudge, recompute, repeat — and you only need the ordering, not the decimals. Ten minutes with a calculator is enough.
For the worked decision the answers look like this:
| Criterion | Weight | Must move by |
|---|---|---|
| Space that fits how we live | 13 | +3.1 points |
| Schools and education access | 12 | +4.6 points |
| Daily travel time | 14 | −5.0 points |
| How easily we could undo it | 8 | −5.4 points |
| Five-year housing cost | 20 | −9.9 points |
| Four other criteria | 33 | Cannot flip it at any weight |
Three points out of a hundred. Nobody could tell you with a straight face that the correct weight for that criterion is 13 rather than 16. The whole decision turns on a distinction you cannot make.
Fourth: notice the criteria that cannot matter
Look at that last row again. Four of the nine criteria cannot change the answer at any weight whatever.
For three of them the reason is simple: your top two score the same on them. If both leading options score a five on being near your support network, then however much you care about that, it adds the same amount to both and cancels. Weight it at ninety points and nothing moves.
The fourth is a subtler case worth knowing about: it is scored differently, but it carries too little weight for the difference to reach across the gap — even taking its points away entirely would not flip the result.
Either way, the practical consequence is the same and it is worth having on a Tuesday evening: any argument about how much those four matter is an argument that cannot change the outcome.
What not to do next
The tempting move, once you know the decision hangs on three weight points, is to go back and adjust something. Revise a weight. Add the criterion that has just occurred to you. Merge away the one your favorite keeps losing on.
Every one of those is a thing a careful, honest person does, and every one of them is legitimate before you know which option it helps and corrupting afterward. The arithmetic cannot tell the difference. Only the sequence can — which is why the method asks you to date the page, and to write down any revision with the reason and which option it helps, on the same page.
If four of your five revisions have helped the same option, you have not discovered that option is better. You have discovered that you want it. That is genuinely useful information — it is just information about you, and it belongs in the paragraph you write at the end rather than in the weights.
Four honest ways to break a tie
In rough order of usefulness. None of them involves touching the matrix.
- Which would you regret less? Not which would you enjoy more — regret is the better predictor at this stage. Picture each having gone mildly wrong in two years. One of those is usually noticeably worse to sit with, and that asymmetry is real information the matrix does not contain.
- Which keeps more options open? When two things are equivalent on everything you can measure, the one that is easier to reverse is worth more, because the ability to change your mind has value. Price the unwind for each — see what a one-way door decision is — and this frequently breaks the tie on its own.
- Which could you explain to a stranger? Write two sentences justifying each to somebody who does not care about you. The one that is harder to write is usually the one you are choosing for a reason you have not admitted, which does not make it wrong — it makes it worth saying out loud before you commit.
- Go and find something out. A tie means the information you have is not sufficient to separate the options. Name the specific fact that would change your answer, price the check that would get it, and compare that price against the chance it changes your mind times what changing your mind is worth. If the check is cheaper, you are not deciding yet — you are buying information, and that is a decision too.
Then write it down
A near-tie is precisely the decision you will later remember as obvious. Before you commit, write one page: the options, the weights, which you chose, the margin, the criterion the whole thing turned on, and what would make you say in two years that you had chosen badly.
That last line is the one people leave out, and it is the one that makes a review possible. The Decision Journal & Outcome Tracker keeps those pages across every decision you make; a sheet of paper in the file with the contract works just as well for one.
Where to start
The free Frame the Decision mini-guide covers the three moves that come before any of this — widening a yes-or-no, forcing the option nobody listed, and pricing the unwind. Eight pages, no email required.
The flip-distance calculation, the four rigging moves with their effects computed, and the diligence budget that decides how much checking a close decision has earned are all in How to Make a Big Decision — 109 pages across four PDFs, with every figure computed on one worked decision and every input printed so you can run the whole thing on your own.
General decision-making guidance — not financial, legal, medical, or tax advice. The worked decision and every figure in it are an invented household’s own planning estimates, written down so the arithmetic can be followed; the method travels, the figures do not.
