ATTENTION, PEOPLE HOLDING A PROS-AND-CONS LIST WITH FORTY-SEVEN PROS!
PEOPLE WHO HAVE OPENED A HEAP OF TABS TO DECIDE WHICH KETTLE TO BUY! PEOPLE WHO ASKED A GROUP CHAT FOR INPUT AND RECEIVED A GIF, A VOICE NOTE, AND A LINK TO A VERY OLD KETTLE REVIEW!
Are you tired of not turning a tiny decision into a limited-series documentary?
Have you spent another evening comparing options that are all probably fine, hunting for one fact so decisive the universe will fax over a certificate reading CORRECT CHOICE, CONGRATULATIONS?
Then congratulations.
You qualify for the Five Number One Rules of Making a Small Decision Without Small Thinking — a compact decision aid for choices that don’t deserve a committee but do deserve a check.
“Wait,” you’re yelling at the screen, “they can’t all be Number One!”
That’s exactly what Big Numbering wants, while it sells one magnificent trick as a substitute for the other four.
Small decisions fail in specialist ways.
A stopping rule ends the searching — it won’t tell you which option is disqualifying.
A diagnostic reason kills a bad option — it won’t tell you what failure rate to expect.
A base rate corrects a personal story — it won’t give you a reason to reopen the choice later.
And an update trigger can’t retroactively decide how much research was enough.
This isn’t a countdown where Rule One gets a sash and Rule Five goes home with a tote bag. It’s a small architecture: classify the decision, stop searching, eliminate with one good cue, check the rate, and write down what would change your mind.
So here they are.
Five rules.
Five Number Ones.
No substitutions. No “I did a spreadsheet.” No waiting until the choice becomes irreversible merely because a fifth review site remains unconsulted.
RULE #1: CLASSIFY THE DECISION BEFORE OPTIMISING IT
Introducing DOOR-O-SCOPE™, the remarkable device that asks one question before permitting another spreadsheet cell to be born:
Can this be reversed?
Bezos set out the useful split in a 2015 shareholder letter. A two-way door is reversible and deserves a lightweight process. A one-way door is consequential and irreversible, and warrants a slow, deliberate, heavily consulted one.
The cue is reversibility.
Not whether the choice feels important. Not what it costs. Not whether it will look impressive in a meeting, and emphatically not whether it has acquired a comparison chart.
That’s the first compression, and it comes before everything else. Decide what sort of decision this is before deciding which option wins.
A reversible choice can be made, observed and corrected. An irreversible one cannot safely borrow that tempo just because it happens to fit in a calendar slot.
Bezos gave the companion intuition a year later: most decisions should probably be made with around 70% of the information you wish you had. Waiting for 90% is usually too slow.
That missing 20% isn’t an argument for carelessness. It’s an argument against giving every two-way door the evidence standard of a launch sequence.
But use it as a process rule, not a measuring instrument. “Around 70%” doesn’t prove an option is good. It says a reversible decision needs a predeclared point where learning gives way to trying.
The door tells the process how much ceremony it needs. It doesn’t tell the process what’s true.
Reversibility decides how long a decision may borrow from tomorrow.
Don’t optimise the menu before checking whether the restaurant has a fire exit.
RULE #1: SET A STOPPING RULE BEFORE THE SEARCH STARTS
From the manufacturers of ONE MORE TAB™ comes the sequel: ONE MORE TAB: THE TABENING.
Contains every review ever written. Except the one that ends the search.
A stopping rule defines satisfactory before the available options begin negotiating with your nervous system.
Write the requirement that makes an option good enough. Write the information that would still matter. Write the condition that ends the search.
Then — and this is the hard part — actually stop when something meets it.
This is satisficing: choosing once an option clears an aspiration level, rather than treating the entire market as a hostage situation.
It is not lowering your standards for the sake of efficiency. It’s distinguishing information that could change the decision from information that merely gives the same uncertainty a new hat.
The operational question is straightforward. What’s still missing, and is this decision cheap to correct if that missing thing turns out to be a surprise?
If yes, a small stopping rule protects your attention. If no, Rule One already sent this decision to slower treatment and you shouldn’t be here.
Statistics adds a warning worth carrying. The number “100 conversions per variant” circulates in A/B testing as a floor; some vendors say 200 to 400. There’s no named coiner, no simulation, no shared study justifying 100 rather than 200 or 50.
A number that appears frequently is not automatically a rule with a derivation behind it.
The same applies to the familiar n ≥ 30 convention. No primary source derives 30 as a normal-approximation threshold regardless of population shape — with heavy skew, a reliable approximation can need hundreds of observations.
A neat threshold becomes a decision aid only after somebody names its conditions.
A stopping rule ends search. It does not manufacture evidence.
The perfect choice is usually an ordinary choice with an infinite return policy in the imagination.
RULE #1: LET ONE DIAGNOSTIC REASON ELIMINATE BAD OPTIONS
NOW AVAILABLE: PRO-CON-VERTER 9000™, which takes one clear disqualifier, buries it under charming features, and announces the option has won on points.
Don’t let it.
A diagnostic gate asks one question first and ends the comparison when the answer clearly discriminates.
It’s non-compensatory by design. A clear no doesn’t get outvoted by a nicer colour, a lower price, or a paragraph opening “on the other hand.”
Other factors still matter — for the cases where the gate doesn’t settle anything.
Take-the-best gives this structure a track record. Czerlinski, Gigerenzer and Goldstein compared it against multiple regression and equal weighting across twenty real-world datasets.
On data already seen, regression scored 74.3%, take-the-best 74.2%, equal weighting 74.1%.
But across ten thousand repeated half-sample splits — predicting data it hadn’t seen — the figures were 72.2%, 71.9% and 71.4%.
Read that carefully, because it isn’t a bumper sticker saying one-reason rules always beat models. It says extra cues and fitted weights can cost more in estimation error than they return, in a particular noisy, finite problem.
The cue needs a job. Not a dramatic entrance.
Recognition makes the same point with fewer confetti cannons. For Swiss cities, simple recognition predicted population with validity around 0.86 — and predicted distance from the country’s centre at about 0.51, which is essentially chance.
Same people. Same city names. Different criterion, and the cue collapsed completely.
The best diagnostic cue is domain-specific. There is no general-purpose one, and anybody selling you one is selling something else.
A good reason is not one more reason. It is the reason that changes the route.
If every virtue can compensate for every defect, the list isn’t deciding. It’s decorating indecision.
RULE #1: USE A BASE RATE BEFORE A PERSONAL STORY
BEHOLD ANECDOTE-A-GATOR™: a story goes in, a universal law comes out, and nobody asks how often the thing happens outside a podcast microphone.
A personal story is evidence that something can happen.
A base rate is evidence about how often it happens across the relevant set of cases.
Before letting the story choose your future, turn the rate into a natural-frequency question: out of a hundred comparable cases, how many does this rate permit?
That question forces the denominator back into the room, and the denominator is what the story quietly removed.
The rule of three is the anti-vibes version of this move. If something occurs zero times in n independent trials, the 95% upper bound on its true rate is roughly 3/n.
It does not say the rate is zero. It says what zero was capable of hiding.
The clinical example is deliberately humbling. After 1,500 patients with zero adverse events, the 95% upper bound is about 0.2%.
Fifteen hundred clean cases. And the honest statement is still “probably under one in five hundred” rather than “safe.”
Small samples are where this stops being funny. At thirty trials, zero events leaves an upper bound near 10%. At ten trials, around 26%. At five, roughly 45% — and the shortcut itself starts breaking down that low.
So when somebody says nothing has ever gone wrong, the useful reply isn’t skepticism about their honesty. It’s a question about n.
Zero observed is a sentence about the sample, not a coronation for the world.
Don’t let a survivor’s story become a population parameter in a trench coat.
RULE #1: WRITE DOWN WHAT WOULD CHANGE YOUR MIND
And now the exciting follow-up product: DECISION-SEALER™.
Once applied, it converts every subsequent fact into “interesting, but not relevant to my magnificent original judgment.”
Before committing, write one observation that would count against the choice. The fact, the outcome, the failure that would make revision necessary.
This isn’t a demand for permanent doubt. It’s a route back from false confidence — and it has to be built before you need it, because afterwards you won’t want it.
The record does one more thing. It separates I changed my mind because the world changed from I changed my mind because a louder opinion entered the room wearing a lanyard.
For a group decision, run a premortem. Before finalising, tell everyone to imagine the project has already failed, then have them write down individually why.
Klein designed it as a fifteen-minute step, and the framing is what makes it work. It replaces a vague invitation to raise concerns with one operational instruction: failure has happened, what caused it?
The mechanism is prospective hindsight. A 1989 study found that framing an outcome as already happened rather than merely possible measurably improves how well people generate specific causal explanations.
It also gives quiet reservations a route around a group that has started treating momentum as evidence — which is the real reason to run it.
A premortem doesn’t predict the future. It names what you’re going to watch afterwards.
And that’s a different job from the other four. Reversibility sets the process. The stopping rule limits search. The cue compares options. The base rate disciplines prediction.
This one gives the decision somewhere to receive bad news.
A decision is provisional when it has a way to hear bad news.
The opposite of stubbornness isn’t indecision. It’s an address for the evidence to send its reply.
BUT WAIT, THERE’S MORE!
“What about choosing a contractor?”
Same rules. Classify how easily it unwinds. Set your minimum terms before calling anybody. Use a genuinely disqualifying question. Ask about the relevant rate rather than one heroic testimonial. Record what would show the choice needs correcting.
“What about a hire, a medical treatment, a legal claim, or an investment?”
Same rules — with a louder warning label. These five organise thought. They don’t replace formal analysis in high-stakes domains, and a tiny sample plus an attractive narrative doesn’t become sufficient by fitting on a sticky note.
“What if it’s a restaurant, a software tool, a weekend plan, or a kettle?”
Same rules, shrunk. Reversible choices need less ceremony, not less thought.
The goal was never to make every small decision thrillingly statistical. It’s to stop a small decision eating the attention a real one is going to need.
The details change.
The architecture doesn’t.
THE FIVE, WITHOUT THE FORTY-SEVEN TABS
Classify reversibility before comparing anything. Two-way doors get a light process; one-way doors get a slow one.
Set the stopping rule before searching, and stop when it’s met.
Use one diagnostic gate before a long comparison. A real no doesn’t get buried under favourable details.
Check the base rate before the personal story — and treat zero observed events as a bound, not proof.
Write the observation that would change your conclusion, before you commit.
DO THIS TONIGHT
Pick one small decision currently squatting in your browser tabs.
Write whether it’s a two-way door or a one-way door. That single word changes everything that follows.
If it’s reversible, write the condition that makes an option good enough, and the one piece of information that would genuinely change your choice. Then close the search when those arrive.
Write the first diagnostic question — and check that it’s capable of eliminating something, rather than producing one more decorative score.
Ask for the relevant frequency. Out of a hundred comparable cases, how often does this happen?
If the answer is “nothing bad happened in the ones I saw,” write 3/n beside it and find out what n actually was. That number is usually smaller than the confidence it produced.
Finally, write one disconfirming observation before you commit. For a group, run the fifteen-minute premortem and keep the causes people write down — those are your revision triggers.
Then choose.
A reversible choice doesn’t get wiser by marinating in tabs. It gets wiser by being made, watched, and corrected when reality earns the right to object.
For the low, low price of refusing to treat indecision as rigour, the complete Five Number One system is yours.
No committee. No ceremonial spreadsheet. No hotline staffed by an actor in a headset beside a plastic ficus.
Just five rules that don’t impersonate one another.
Operators are no longer standing by.
The operator is the person who has to live with the kettle.