The Blueprint · The Durability Curve

Everyone Hit the Target

Sun, 04 Oct 2026 00:00:00 GMT

In 1902, the French administration running Hanoi had a rat problem. The city’s new sewers, the pride of its colonial engineering, had turned out to suit the rats perfectly, and the rats carried plague. So the authorities reached for what looked like a clean solution. They put a bounty on rats. Anyone who brought in a rat’s tail got a small payment, and the city would buy its way to fewer rats.

The tails came in by the thousand. Then officials started noticing something out in the streets: rats, alive and healthy, running about with no tails. Only the tail was paid for. People caught rats, cut off the tails and let the rats go, perhaps so they could breed more. The more enterprising skipped the streets altogether and set up rat farms on the edge of the city. The bounty was paid, the tally of tails climbed, and the rat problem did not improve. The administration’s own conclusion, once it understood what it had built, was that the scheme was breeding more rats than it caught.¹

The mistake is easy to name once you see it, and easy to make. What the city paid for was a stand-in, a tail handed over a counter, and the cheapest way to produce that stand-in had nothing to do with the goal of fewer rats.

Whatever you actually care about tends to be hard to measure head-on, so you reach for a number that moves with it, and that number is an honest signal only for as long as nobody is pushing on it. Attach a reward and you lean on it with your whole weight, and the two come apart the instant moving the number is cheaper than delivering the real thing. The cruel part is that the number keeps climbing. It looks like triumph, right up until someone checks the thing it was supposed to stand for and finds that it never moved, or quietly got worse.

You do not have to go back to colonial Hanoi for this. When Wells Fargo set its staff aggressive targets for how many products each household held, the staff hit those targets by opening millions of accounts that customers had never asked for and did not know existed. The measure, products per household, looked spectacular. The thing it was meant to stand in for, customers happily choosing more of the bank’s services, had been faked account by account, and when it came to light it cost the bank billions and a great deal of its name.²

The same shape appears wherever a clock is the target. Give hospitals a four-hour limit on emergency waits and the four hours quietly become the job. In England, some hospitals kept patients waiting in ambulances queued outside the doors, where the four-hour clock had not yet started, until staff were sure they could deal with each one in time. Waits did fall, though the trusts’ own figures ran well ahead of what patients reported: 96% within four hours officially in 2004/05, against 77% of patients surveyed. Bevan and Hood called the gap an apparent discrepancy, and the survey itself warns that its figures are not directly comparable with the trusts’ data.³

Once you have the shape you find it wherever a number with a shortcut carries a reward or a punishment. Put schools in a league table and some will teach to the test, because the test is what gets counted. If there is a cheap way to move the count, paying for it gets you more of the count and less of what it was counting.

What you can carry away from all this is a single question, put to any number someone holds up as proof that things are going well. What is the cheapest way to move this number, the cheapest way that has nothing to do with the real work? If such a path exists, and anyone is rewarded for the number going up, assume it has already been found. A figure that has become a target should send your attention straight past it, to the thing it was only ever standing in for, which is what you wanted to know in the first place.

None of this means you should throw the numbers away. You cannot run a hospital or a school or a company on feeling alone, and a measure that no one is gaming tells you something real. The trap is narrow and specific. It springs when a single number that can be moved on its own is handed a reward, so that hitting the number becomes the point and reaching the goal becomes optional.

Everyone hit the target. That was the problem.

Michael G. Vann, Of Rats, Rice, and Race: The Great Hanoi Rat Massacre, an Episode in French Colonial History, French Colonial History 4 (2003): 191–203. Vann documents the tail bounty, the cut-and-release of live rats, and the rat farms from the colonial archive; the administration concluded the scheme was increasing the number of rats in the city. ↩
Consumer Financial Protection Bureau, Consumer Financial Protection Bureau Fines Wells Fargo $100 Million for Widespread Illegal Practice of Secretly Opening Unauthorized Accounts (8 September 2016). The 2016 penalties totalled $185 million; a later Justice Department and SEC settlement in 2020 reached $3 billion. ↩
Gwyn Bevan and Christopher Hood, What’s Measured Is What Matters: Targets and Gaming in the English Public Health Care System, Public Administration 84, no. 3 (2006): 517–538. In 2002/03, 139 of 158 acute trusts officially reported ninety per cent of patients spending under four hours in A&E, but only sixty-nine per cent of surveyed patients said they had been; by 2004/05 the figures were ninety-six and seventy-seven per cent. The Healthcare Commission’s Emergency department survey 2004/2005 asked patients how long their whole visit to the emergency department lasted (Q7), and says on p. 7 that the waiting times patients report “cannot be directly compared with waiting time data from NHS trusts”. ↩

The Planes That Didn't Come Back

Wed, 09 Sep 2026 00:00:00 GMT

In the Second World War, the American military had a problem with its bombers. Too many were being shot down, and the obvious remedy was to fit armour to them. But armour is heavy, a plane can carry only so much, and cover it everywhere and it will barely leave the ground. So the real question was where. Which parts of the plane most needed the protection.

They had data to answer it. Returning bombers were inspected, and the engineers could see where the hits had landed. In the telling that became famous they clustered in a familiar pattern: heaviest along the fuselage and the wings, the engines coming back comparatively clean. Reinforce the parts taking all the fire, the reasoning went, and you will save the most planes. It is a hard argument to fault. You have the data, the damage is right there in front of you, and you are only following it.

A statistician named Abraham Wald looked at the same figures and reached the opposite conclusion. The armour belonged where the holes were not. Protect the engines, the clean areas, the places nobody thought to patch.¹

The reason, once you hear it, rearranges something in your head and does not put it back. The engineers were studying the planes that came back, because those were the only planes they had. A bomber covered in holes across its wings and fuselage was a bomber that had been hit in those places and had still flown home, which meant those were exactly the spots where a plane could take a beating and survive.

The clean areas were clean for a reason the data could not show. The planes that were hit there did not return to be inspected. They were somewhere in the sea. Those absent holes marked the wounds that killed.

That is the whole trap, and it is worth seeing plainly, because it has nothing to do with aeroplanes. Whenever you draw a lesson from a set of examples, something decided which examples reach you, and that something is rarely chance. It is a filter. Survival, success, fame, memory, simply staying in business: each of those filters does its work by removing the failures before they ever arrive.

So the set in front of you is only the residue left once the filter has run, a long way from a fair slice of everything that was tried, and because the filter’s whole job was to take things away, the things it took away are the ones you cannot see. They are also, very often, the ones you most need.

You feel how strong this is the moment you start looking for it. Pick up any book about how some billionaire built their company and you will find a handful of habits offered up as the cause: the early mornings, the ferocious focus, the refusal to hear the word no. What you will never find, because nobody writes that book, is the far larger pile of people who rose at the same hour and focused just as fiercely and refused just as hard, and went broke regardless.

If the people who failed had every one of the winning habits too, the habits cannot be the thing that set the two apart. You are being handed the survivors and asked to reverse a recipe from them, with the one ingredient that could tell you what actually mattered, the failures, quietly deleted from the page.

It runs through much smaller things as well. When someone tells you they do not make things like they used to, waving at a hundred-year-old chair that is still rock solid, remember that you are looking at the one chair that lasted a century. Most of the flimsy furniture of the past broke and was thrown out generations ago. You are holding the best of the old, the sliver that survived, up against the everyday run of the new.

So here is the move, and it is a single question you can put to almost any claim built on examples. What decided which cases I get to see, and what would the ones it left out have looked like?

Before you copy the habits of the successful, go looking, in your imagination if nowhere else, for the people who did the very same and failed, and ask whether they shared the habit. Before you decide the old ways were better, ask what broke and disappeared long before you arrived to inspect what was left. The absence is shaped, and its shape is the part of the answer that did not survive to be seen.

None of this means every set of examples is lying to you. Sometimes your sample really is a fair one, gathered without a filter quietly picking the winners, and then this particular problem does not arise. The trap springs only when the very process that produced your evidence is the same process that removed the counter-examples. The tell is easy to learn once you have it. Your data is made of survivors, of winners, of the things that lasted and the stories that got told. The moment you notice that, you know to go looking for the planes that did not come back.

Wald’s wartime analysis survives as eight memoranda for Columbia University’s Statistical Research Group; the standard scholarly reconstruction is Marc Mangel and Francisco Samaniego, Abraham Wald’s Work on Aircraft Survivability, Journal of the American Statistical Association 79 (1984): 259–267. The memoranda estimate the vulnerability of each part from the damage on returning aircraft and a distribution of hits over the plane that is assumed in advance or measured by firing dummy bullets; the familiar bullet-hole diagram and the scene of engineers overruled in a briefing room are later retellings, not Wald’s own. ↩

The Extra Lane Fills Itself

Wed, 02 Sep 2026 00:00:00 GMT

A city looks at a motorway that crawls every rush hour and does the obvious thing. It widens the road. More lanes, more room, more cars moving at once, and for a while the traffic loosens and the journey gets quicker. Then, within a few years, the road starts filling again, the familiar crawl creeping back onto the wider road that was meant to end it. Somewhere a lot of money went into a fix that did not fix the thing it was meant to fix, and everyone quietly agrees they should have built it wider still.

The intuition underneath that decision feels like common sense. Traffic is a fixed lump of cars trying to squeeze through a narrow pipe. Widen the pipe and the same lump flows more easily. If it clogs again, the lump must have grown, so widen it again. Roads as plumbing, congestion as a volume problem, more capacity as the answer.

The pipe picture is wrong, and it is wrong in a way that explains the whole thing. The traffic you can see was never the whole demand. The jam itself was holding some of the rest back. Every day the road crawled, some people looked at it and chose not to be on it. They took the train instead. They shifted their trip to before the rush or after it. They bundled three errands into one, worked from home, or simply did not make the journey at all. The congestion was a wall, and behind it sat the trips it was holding back, some of which would become worth making the moment the wall came down.

So you add the lane and the wall comes down. Driving gets quicker, and quicker driving is an invitation. The person who used to take the train may get back in the car. The trip that was not worth the crawl becomes worth it. Trips the old jam had quietly discouraged start returning to the road, eating into the improvement the new lane was meant to deliver. The refilling slows as the road clogs again, until the next trip that might have joined it is once more not quite worth making.

That is the short loop, and it is worth saying plainly. Travel time is a price, paid in minutes rather than money, and like any price it holds demand down. Add capacity and demand can rise to take up the room, and how much depends on how much the old conditions were suppressing. Where that suppressed demand is large, the road fills until much of the improvement is gone and you have moved more cars for less benefit than the map promised. Where it is small, the extra lane stays useful. The loop runs wider over longer periods, as people change where they live and work, firms follow the new access, and trips that the old conditions ruled out entirely begin to make sense.

You do not have to take this on faith. One of the clearest findings comes from a study of US cities by Duranton and Turner: the miles people drive rose roughly in step with the interstate lane miles built, which is a polite way of saying new roads fill themselves.¹ Houston is the vivid illustration, a motorway widened to as many as twenty-six lanes at its broadest point, whose rush-hour journeys grew sharply longer again within a few years of the work finishing.² The reverse case points the same way. Across more than seventy cases where road space was reallocated away from traffic, the traffic problems predicted for the surrounding streets were generally much smaller than expected.³ When Seoul pulled down an elevated expressway and restored the stream it had been built over, road trips fell and subway ridership rose, rather than all the displaced traffic simply reappearing elsewhere.⁴ Some journeys find another route, some shift to another mode, some move to another time, and some are simply no longer made, melting back behind the very wall the road had been holding down.

Here is why this is worth carrying around, because it is not really about roads. The friction was doing a second job. The wait, the queue, the crawl, whatever the painful thing was, was also a filter, quietly turning away demand you never saw because it never arrived. A support team drowning in tickets hires more people and the replies get faster, and some customers who would once have given up on a small problem now bother to report it. A clinic adds appointment slots, and some patients who would have gone elsewhere, put the visit off, or never booked at all start filling them. Not every capacity increase works like this. The tell is that the old crush was already making people give up, postpone, reroute, or go without.

Once you can see it, the fix that everyone reaches for starts to look naive. When something is overwhelmed and the instinct is to add capacity, the first question is whether the crush is also holding demand back, and how much is waiting behind it. Where a lot is waiting, much of the added capacity can fill again, and you have bought a larger operation running at much the same strain. Then capacity alone will not get you the outcome you wanted, and you need a lever on the demand as well, whether that is a price in money, a priority, or a rule about who gets on. Shape the demand too, because added capacity gives suppressed demand somewhere to return.

None of this makes capacity useless. The trap is sprung when the crush is suppressing demand that lower friction can release, and a surprising number of the queues you fight with, in traffic and far beyond it, are doing exactly that.

Add room to a queue that was turning people away, and the crowd it was hiding comes back to claim the room.

Gilles Duranton and Matthew A. Turner, “The Fundamental Law of Road Congestion: Evidence from US Cities” American Economic Review 101, no. 6 (2011): 2616–52, which found that the vehicle-kilometres people drive rise roughly in proportion to the interstate lane-kilometres built, and probably somewhat less rapidly for other kinds of road. The extra driving comes from several sources, including more driving by existing residents, commercial traffic, and migration into the area. ↩
In 2008 Houston finished widening the Katy Freeway (Interstate 10), which reaches as many as twenty-six lanes at its broadest point counting the frontage and managed lanes, at a cost of about $2.8 billion; a later analysis found that between 2011 and 2014 the morning commute along it grew roughly 30 per cent longer and the afternoon commute roughly 55 per cent longer. Kinder Institute for Urban Research, Rice University. ↩
Sally Cairns, Stephen Atkins and Phil Goodwin, “Disappearing traffic? The story so far” Municipal Engineer 151, no. 1 (2002): 13–22, reviewing more than seventy cases where road space was reallocated away from general traffic. The results varied so widely that the authors caution against any single rule-of-thumb figure, but the traffic problems predicted for surrounding roads were generally much milder than feared, and most measured cases showed a net reduction in traffic. Reported responses included rerouting, retiming trips, switching modes, combining or forgoing journeys, and changing destination. ↩
Jin-Hyuk Chung, Kee-Yeon Hwang and Yun-Kyung Bae, “The loss of road capacity and self-compliance: Lessons from the Cheonggyecheon stream restoration” Transport Policy 21 (2012): 165–178. After the elevated expressway was demolished and the surface road beneath cut from four lanes to two in each direction, subway ridership rose and the number of road trips fell, rather than the displaced cars simply crowding the surrounding streets. ↩

The Option You Were Never Meant to Buy

Tue, 01 Sep 2026 00:00:00 GMT

Next time you buy popcorn, look properly at the sizes. Say a small for three pounds, a large for seven, and wedged between them a medium for six pounds fifty. Fifty pence less than the large, for a good deal less popcorn. The exact prices vary; the shape is familiar. You will not think about it for long. You reach for the large, feeling like you spotted a bargain.

You have not spotted anything. The large looks like a bargain because the medium is sitting beside it. The medium does not need many buyers to matter: beside the large it loses the comparison, and in losing makes the large look like the obvious, thrifty choice. It can sell almost none of itself and still change what gets bought.

To see why that works, notice how you judge a price. You do not weigh it against some settled sense of what popcorn is worth. Nobody carries a true price in their head. You judge it against the other numbers in front of you. And that is the opening. Whoever sets the board controls what each option gets compared against. You can shift how good a thing looks without altering the thing itself, only what sits beside it.

The most famous demonstration used prices The Economist once listed. Offered a web subscription for fifty-nine dollars and a print-and-web bundle for a hundred and twenty-five, most people take the cheaper web one. Add a third option, print alone at a hundred and twenty-five, the same price as the bundle that throws in the web edition for nothing extra, and the choice flips. That useless third option changes the whole comparison: the great majority now take the bundle, because beside it a hundred and twenty-five dollars looks like a gift.¹ One option nobody chose changed what almost everybody chose.

The dominated decoy is only the cleanest version. A comparison can be shaped in other ways. Set one option far pricier than the rest and it drags your sense of normal upward, so the one beneath it reads as sensible. Offer three tiers and the middle can gain simply by being neither extreme. Different mechanisms, one shared privilege: whoever chooses the alternatives chooses the comparisons you make. The set is the argument, and someone else wrote it. The pull goes deeper still. Often you have no finished preference when you begin, and the set becomes your yardstick, what counts as dear, what counts as generous, before you have judged anything at all.²

None of this is a spell. The decoy effect itself is real but modest, and it comes and goes. It has proved most reliable in the kind of choice the Economist offered, where every option is a tidy number and the comparison is easy to run in your head. Give people messier, more realistic options, and it becomes much less dependable. In a study of millions of supermarket wine purchases, the presence of a dominated decoy was associated with a small shift toward the target, and least of all among the shoppers who bought wine most often.³ Experience seems to blunt it.

Once you have the shape, you see it well past the shop. A hiring shortlist with one candidate clearly weaker than your favourite, so the favourite reads as the obvious call. A house shown first, tired and overpriced, so the next one arrives as relief. And the version that should trouble anyone who reads a chart: a new system announced at eighty-seven, sat beside an old baseline at fifty-four, so the eighty-seven lands like a triumph. Put the real rival back, an existing system quietly scoring eighty-two, and the triumph shrinks to the five points it always was. Nothing about the new system changed. The set it was measured against did.

A weak baseline is a decoy with a chart around it. A field can flatter a winner two ways, by adding something weak or by leaving out something strong. The baseline at fifty-four matters less than the rival at eighty-two you were never shown. When a result looks unusually good, look hard at what it beat, then ask what should have been in the field and was not.

A menu is sometimes just a menu, and an inferior option can survive for reasons that have nothing to do with persuasion: legacy pricing, a segment you are not in, plain incompetence. The tell to watch for is an option plainly worse than another for the same money, or the same money for plainly less. Finding one does not prove someone planted it, but it is a good reason to inspect the comparison it creates. The options around a thing can carry real information; the mistake is letting someone else’s set become your standard without noticing.

The test takes about ten seconds. Strike out the option that makes your choice look good, and ask whether you would still want what you wanted. Then ask what is missing. Would you still choose it beside the rival you would really buy, the house you could actually afford, or the number it truly has to beat? And does it still make sense against what you need and can spend? If your choice survives that, it survived the comparison. If it weakens, part of what impressed you belonged to the field, not the winner. Most choices still need comparison, so do not stop comparing. Choose the comparison yourself.

The board was built to be read across, each option lending meaning to the next. The comparator is part of the claim: a winner is only as impressive as the field it was allowed to beat. Before you trust one, look hard at that field, and ask whether it still wins in the set you would have chosen.

The option you were never meant to buy is the one doing the selling.

The pricing came from a real Economist subscription page; the choices were measured by Dan Ariely in Predictably Irrational (HarperCollins, 2008), chapter 1. Offered all three options, about 100 students split 16 for web-only, none for print-only, and 84 for the print-and-web bundle. With the print-only option removed, the same offer drew 68 for web-only and 32 for the bundle. ↩
People often do not arrive with settled preferences waiting to be revealed. They build them while choosing, from the options in front of them and the way the choice is framed: Bettman, Luce and Payne, “Constructive Consumer Choice Processes” (Journal of Consumer Research, 1998). ↩
The pull is strongest in stylised, all-numbers choices like the Economist’s and weakens with realistic options: Frederick, Lee and Baskin, “The Limits of Attraction” (Journal of Marketing Research, 2014). In a study of 3.6 million UK supermarket wine purchases, a dominated decoy shifted preference toward the target by roughly one per cent overall, and least of all for the shoppers who bought wine most often: Devine, Goulding, Harvey, Skatova and Otto, “How decoy options ferment choice biases in real-world consumer decision-making” (npj Science of Learning, 2025). ↩

You Were Never the Customer

Tue, 11 Aug 2026 00:00:00 GMT

Some of the best things you use cost you nothing. The map that knows every road. The inbox that holds a decade of your life. The feed you scroll before you are properly awake. You pay nothing for any of them, and somewhere along the way you half-decided it was either generosity or some trick of scale.

Money does not work like that. A service with millions of users and no charge is not running on goodwill. The cost of it is real, and someone is covering that cost, which means money enters the system somewhere. It just enters somewhere you cannot see, from someone who is not you.

That is the whole thing to understand about free. A price of zero for you does not mean nobody is paying. It means the payer is someone else, and once somebody other than the user is paying, the user is no longer the one the arrangement answers to. You have heard the sharp version of this: if you are not paying, you are the product. As far as it goes that is right, and it goes about half the distance. It tells you where you stand in the deal. It says nothing about what the deal does to the thing you are using, which is the part you can actually feel.

The one who pays is not the one who uses, and the whole arrangement sits on that gap.

Watch what it does to the thing itself. Every product improves over time along some axis, and the axis it improves on is the one that keeps the money arriving. When the money is yours, the product has to keep you willing to hand it over. That is a thin protection and it is a real one, because you can stop. When the money is someone else’s, that protection thins. Leaving is still leaving, but you are one of millions, and no payment stops when you go. What you want still counts, but only as far as it keeps you present for the person actually paying.

The free service grows sharper at holding your attention, because attention is what it sells. It grows more precise at predicting you, because predictability is part of what is sold. And getting better at those things is often the very same motion as getting worse at respecting your time. That is why the free thing you once loved keeps changing in ways you did not ask for and cannot quite explain. You are feeling the design work exactly as intended. It was simply never intended for you.

The clearest version of this is older than the internet. Commercial television looked like a gift: hours of programmes, sent into your living room, free. The customer, though, was the advertiser, and the audience was the thing being sold.¹ The programmes existed to gather that audience and hand it to the adverts, so what got made and when it aired followed from how much of you could be delivered to them. The show was the bait. You were the catch. A whole industry ran on a sale you were not part of and mostly never noticed.

None of this is a reason to distrust everything that costs nothing. Free is not a synonym for sinister. It is worth knowing who the payer is, though, because the common answers point in very different directions. Sometimes the payer is a third party buying your attention, and the arrangement can work quietly against you. Sometimes the payer is a benefactor, a library carried by taxes or a tool released by people who simply want it to exist, and the thing genuinely serves you because the person paying wants it to. And sometimes the payer is a later version of yourself, the free tier that costs nothing today because its job is to turn you into the paying one tomorrow.

The shape is the same each time, in that the money comes from somewhere other than you. Where it points is not the same at all, which is why the name on the invoice is the thing worth knowing.

So here is the habit worth building. When something is free, the zero on the price tag is the loudest fact about it and the least useful, because the cost has not gone anywhere. It has only moved out of sight.

Ask who the paying customer is, what they are buying, and what you have to keep doing for that money to keep arriving.

Answer all three and the product stops being mysterious. The feed that refuses to show you posts in the order they were written is one more setting somebody chose for you. Ranking genuinely helps when you follow more accounts than you can read, and it is also the lever that decides how long you stay, which is the part being sold. The loyalty card that hands you a small discount in exchange for a record of everything you buy makes sense the instant you see what the discount buys: purchases that used to be anonymous, now attached to a name and followed over years. Find the payer and you can see where the product’s interests run alongside yours, and where they quietly part.

And when you cannot find the payer, look harder, because there is one. Sometimes the money is an investor’s, spent now against a payment nobody has invented yet, which is why a free thing can be so good for so long and then turn. The bill was always going to be presented. The only open question was who to.

Free is one of the most honest words in the world about your side of the deal, and one of the most silent about the other. You are told, precisely and truthfully, that you will pay nothing. You are told nothing at all about who will.

Free never means nobody pays. It means someone else is paying, and everything it does has to keep that money arriving.

The idea is older than the phrase. In 1973 the artists Richard Serra and Carlota Fay Schoolman bought airtime to broadcast a short video on television, Television Delivers People. Its scrolling text put it flatly: “You are the product of t.v.” ↩

The Setting You Never Changed

Sun, 09 Aug 2026 00:00:00 GMT

You have settings you have never changed.

Not because you sat with each one and judged it right. Because changing it was one more small task on a day that already had enough of them, and the thing worked well enough as it came. The notifications, the privacy toggles, the plan you signed up on, the way the app opens. Somebody handed you a starting point and you started from there. Almost everyone does.

We treat that starting point as neutral. A sensible middle. The factory had to pick something, so it picked the reasonable option and left the rest to us. The default feels like the absence of a decision, the blank page before anyone chooses.

It is the opposite. The default is a decision, made by someone who is not you, and often the most powerful decision in the whole design. It gets that power from one plain fact about people: many of us never change it.

Watch how that fact does its work. Every change, however small, costs a little effort, and even a little effort is enough to keep many people where they started. A default also reads as advice. Someone who knew more than you set it here, so here is probably fine. And there is the pull of the path already laid: the option in front of you is the one you can take without stopping to think, and not stopping to think is most of what any of us do in a day. Put those together and the person who sets the default is not gently nudging the few who care. They are quietly deciding the outcome for the many who will never look.

You can see the size of that power most clearly where the stakes are life and death.

Take Germany and Austria. Neighbours, with similar wealth and much shared culture. Germany asks you to opt in, rather than presuming your consent. Austria works the other way, presuming consent unless you object, so if you do nothing you remain a potential donor. In the comparison that made defaults famous, Germany’s effective consent rate sat around one in eight. Austria’s ran above nine in ten.¹ Two neighbouring countries are never a controlled experiment, and apparent consent is not the same as an organ donated. But where you can run the clean experiment, the same lever pulls the same way.

The same lever changed how a country saved for retirement. Britain used to make workers opt in to a workplace pension, and only about half of those eligible were saving into one. From 2012 the system began to flip: automatic enrolment was phased in, so you were enrolled unless you chose to leave, and within a decade participation was near nine in ten.² The system did not wait to persuade each worker. Other things moved alongside the default, minimum contributions among them, so this was no clean laboratory test. But the resting state had changed from not saving to saving, and millions simply stayed where it put them. The resting state, the thing that happens while you are busy living, does an extraordinary amount of the work.

Follow that one step further and the incentives start to matter. If the default can decide the outcome for so many people, then the right to set it is worth real money and real power to whoever holds it. So the question to ask of any default is not “is this a fine starting point.” It is: who set this, and what is it set to do?

The pre-ticked “yes, keep me posted” is set to grow a mailing list. Google paid Apple billions to be Safari’s default search engine, betting the one you start with is the one you keep.³ The subscription that renews unless you cancel is set on the company’s calendar, not yours, and it benefits from the same forgetfulness that keeps you from changing any other setting. The tip screen that opens with twenty per cent already lit is set to make twenty per cent feel like the starting point. None of these is an accident, and none is neutral. Each is a resting state that someone tuned toward their own goal, wearing the calm face of a reasonable place to start.

Here is the habit worth building, and it is the whole point of looking under this particular surface. When you meet a setting you did not choose, stop reading it as the way things simply are. Read it as a sentence somebody wrote and left for you: this, unless you say otherwise. The moment you hear it as a sentence, two questions fall straight out of it. Who benefits if I leave this exactly as it is? And what would I have picked if the box had been blank and the choice were genuinely mine? The gap between those two answers is the thing that was quietly taken from you while you were getting on with your day.

Often the gap is small, because the person who set the default guessed well or genuinely wished you well, and you leave the setting alone, now knowing that you chose to. Sometimes it is large, and you change it, and it takes ten seconds you did not know were worth taking. Either way you have stopped mistaking someone else’s decision for the natural order of things.

Once you can see it, you cannot stop seeing it. The box already ticked on the form. The toggle set to share before you looked. The plan pre-selected as the one that quietly renews. The world is full of resting states that were set before you arrived, and every one of them was set by a hand with a reason. This is only one of the structures the world is built from, and it stays invisible until someone draws you the diagram.

The setting you never changed is still a choice. Someone else just made it for you.

Eric J. Johnson and Daniel G. Goldstein, “Do Defaults Save Lives?” Science 302 (2003): 1338–39. The same work paired a European comparison of effective consent rates with a controlled experiment, in which agreement to donate was far higher under an opt-out default than an opt-in one. ↩
Department for Work and Pensions, “Workplace pension participation and savings trends of employees: 2009 to 2025” (30 July 2026): around nine in ten (90%) of eligible employees were saving into a workplace pension in 2025, up from about half when automatic enrolment began in 2012. ↩
In 2022 Google paid Apple an estimated $20 billion for default search placement in Safari, according to filings in the US antitrust case that found Google an unlawful monopolist in 2024. US Department of Justice court filing. ↩

The Blueprint · The Durability Curve

Everyone Hit the Target

Footnotes

The Planes That Didn't Come Back

Footnotes

The Extra Lane Fills Itself

Footnotes

The Option You Were Never Meant to Buy

Footnotes

You Were Never the Customer

Footnotes

The Setting You Never Changed

Footnotes