Rory Sutherland on Alchemy
What this covers
Rory Sutherland, an advertising executive at Ogilvy, discusses his book Alchemy with EconTalk host Russ Roberts, arguing that human psychology, perception, and meaning create genuine economic value—contrary to the assumption that advertising and branding are mere trickery or inefficiency. The conversation ranges across consumer behavior, public policy, infrastructure, and corporate decision-making, grounded in behavioral economics and Sutherland's decades in marketing. The central claim is that objective features and incentives pass through a perceptual black box before affecting behavior, so context, narrative, and design shape what people actually choose as much as price or product specifications.
Sutherland develops this argument by reframing supposedly irrational behaviors as rational responses to non-ergodic environments where outcomes compound and past decisions matter. Brand loyalty, switching reluctance, and extreme consumer choices become intelligible as variance reduction in multiplicative rather than additive worlds. He illustrates through concrete examples: the London Underground Map's addition of the Overground raised ridership 400% through pure choice architecture; raising a KFC Australia price that was too low increased demand; Uber succeeds partly because watching a car approach reduces the psychological burden of waiting. Advertising itself is reframed as costly signaling—like a flower's petals—that reliably indicates seller confidence in repeat business. The conversation also covers Chesterton's Fence, the dangers of quantification bias, how intermediaries and repeated relationships carry informational value that linear efficiency calculations miss, and how choice architecture can reshape outcomes without quotas or coercion. Several claims remain contentious within economics; Sutherland and Roberts engage with both the promise and the limits of each argument.
Sutherland argues that human perception, context, and meaning—not just objective features or monetary incentives—drive economic behavior, so advertising, branding, and psychological design create genuine value that conventional economics dismisses as inefficiency or trickery.
- Markets are valuable because they are inventive and solve the same problem multiple ways, not merely because they are efficient
- Much seemingly irrational consumer behavior is rational variance reduction in a non-ergodic, multiplicative environment
- Persuasion and social norms have flexibility (allowing good-reason non-compliance) that law and price incentives lack
Choice architecture and menu design shape consumer behavior more powerfully than price or product specs alone.
- When people gain the option to choose more units (two houses instead of one, two or three cars, more frequent shopping), they buy more extreme and complementary options rather than duplicating a single conventional choice—buying a crashpad near work plus a beach place, more extreme cars, or shopping both upmarket and downmarket—which squeezes mid-market sedans and mid-market retailers.
“If I gave you the option of buying two houses, what you certainly wouldn't buy is two suburban houses that were very similar. You'd possibly have a crashpad right next to your place of work and a place on the beach.”
- Contrary to economic dogma that a non-selling product should be cheaper, raising the price can raise demand—as happened at KFC Australia—because in menu environments people use price to navigate and shop for either a bargain or a treat; pricing in the middle satisfies neither, so a product can be priced too low.
“'Well, try doing the opposite. Try putting the price up.' Weirdly, when you put the price up, demand went up.”
- Asking consumers what class of travel they want before they know the price difference is irrational from the consumer's standpoint—since willingness to pay for premium is a proportion of the economy price, not a fixed amount—and rather than demanding more information, consumers default to 'Economy,' which cost one airline about 15 million pounds a year in lost premium revenue.
“Asking consumers what class of travel they want before they know the price difference is, when you think about it, utterly insane.”
- Property search websites using fixed 50,000 price increments make houses priced at unusual points (e.g., £825,000) effectively undiscoverable, because searchers anchor at round numbers and search high-low from £850,000 or low-high from £800,000—turning a smooth price-demand curve into a price-demand ziggurat and creating arbitrage for buyers who search differently.
“you can't sell a house in the United Kingdom for £825,000... neither of those people will discover an £825,000 house.”
- Consumers use brands heavily as a form of variance reduction: buying a Samsung television is not a bet that it is the best possible TV but a fairly reliable certainty that however good it turns out to be, it won't be dreadful—and the same logic explains habits and social copying as ways to reduce variance of outcome.
“we use brands very heavily as a form of variance reduction in making purchases.”
Perception and context, not objective reality, determine how people respond to the same stimulus or situation.
- Raj Chetty's work on Denmark reducing the tax break for pension saving revealed that the majority of people carried on saving as before, with only a small group of super-rich people with financial advisors responding to the incentive—people who did not need government help anyway—illustrating that objective reality passes through a perceptual black box before affecting behavior.
“what happened in Denmark when they reduced the tax break for pension saving... the majority of people just carried on saving as before.”
- Small contextual acts produce outsized loyalty effects: a coffee shop that closes at 5:00 but makes a show of unlocking the door to serve a desperate late arrival gains a customer for life, while one that turns someone away at two minutes to five loses them for a decade—because the same act can read as generosity or insult depending on context.
“If someone runs up desperate for a coffee, make a big show of unlocking the door and let them in and serve them, and you've got a customer for life.”
- Because behavior is driven by perception, and perception has not evolved to approximate objective reality, designing things around stories is as important as designing the things themselves—analogous to how TVs generate the whole visible spectrum, and even non-physical colors like magenta, by stimulating only the eye's three cone types at different ratios.
“TV manufacturers don't bother to produce the whole spectrum of colors because they know that human or higher primate vision is only sensitive to three.”
- Economics modeled on context-independent physics is fundamentally doomed for human behavior, because a single stimulus can produce opposite behaviors: rising London house prices lead some people to take gains and move out while leading others to stay put for fear of never moving back, so models must adapt to context, circumstances, and the behavior of others.
“people can respond to the fact that their house price is going up and up and up in London in one of two ways. They can either take the gains and move out, or they can stay put”
- Humans respond to emotion, emotion derives from meaning, and meaning derives from context as much as from any objective circumstance—so the same act can be either generosity or insult (a gift can be so generous it obligates, while returning a gift is a massive insult), making any single-line mapping between environment and behavior doomed.
“humans respond to emotion. Emotion is derived from meaning, and meaning is derived from context as much as it's derived from any kind of objective measure of circumstance.”
Markets succeed through inventive, decentralized problem-solving and repeated relationships, not through efficiency or price competition.
- Economics likes free markets for the wrong reason—because they are notionally efficient through competition—whereas the truly magical thing about markets is that they are inventive: they solve problems through market-tested innovation and often find more than one solution to the same problem, solving it differently for different people rather than optimizing for the average.
“it likes free markets because they're notionally efficient, whereas I like markets because they're inventive”
- Most problems have already been solved somewhere else but no one has transferred the solution from one domain to another (lateral category analysis / TRIZ); examples include biomimicry, where the Japanese bullet train's beak was modeled on the kingfisher to reduce tunnel noise, and the rail industry's failure to adopt the airline industry's yield management.
“problems have mostly been solved somewhere else, but no one's transferred them from one domain to another.”
- Intermediaries and repeated relationships have informational value that economics disparages as inefficiency: there is a real difference between ten people buying once and one person buying ten times, because the repeat-dealing intermediary's long-term interest is served by square dealing—so when Marseille fish dealers were to be replaced by a centralized auction, even the restaurant buyers resisted because they valued buying from their guy in ongoing mutually beneficial exchanges.
“there's a difference between 10 people buying from someone once and one person buying 10 times over time. And so the intermediary knows that his long-term interest is served by square-dealing with the customer.”
- Advertising is very old—a flower is a weed with a marketing budget—and its function is reliable signaling: a bee can only discover nectar by visiting, so a flower's costly investment in large petals only makes sense if it expects repeat visits, making the upfront cost of advertising a reliable indicator of seller confidence in repeat business.
“a flower is effectively a weed with a marketing budget.”
Psychological innovation—reducing uncertainty and managing expectations—drives consumer satisfaction as much as underlying service quality.
- Uber is very largely a brilliant psychological innovation that changes the nature of waiting: being able to watch a car approach makes a 10-minute wait inordinately less frustrating than waiting in a state of ignorance, and this reduction of uncertainty is as significant to Uber's success as scale or economies—even when the displayed car position may not be perfectly accurate.
“Uber is very largely a brilliant psychological innovation which changes the nature of waiting. So that if you can watch a car approach, the experience of waiting 10 minutes for a car is inordinately less frustrating than if you're in a state of ignorance.”
- Giving any kind of estimate of departure time—even a slightly pessimistic one—is inordinately better for psychological wellbeing than the single word 'Delayed,' because the word 'Delayed' makes people fear the worst (that it is effectively canceled and they are being broken to gently), revealing a deep human need to maintain a sense of certainty and control.
“if you can give any kind of estimate of time of departure, even if it's slightly pessimistic, it is inordinately better than the single word 'Delayed,' in terms of our psychological wellbeing.”
- Rigid travel pricing creates needless anxiety and a sense of injustice: airlines leave business-class seats empty rather than offer them cheaply at the last minute, and when you miss a train the immediate collapse of your ticket's value to zero is perceived by humans as unjust, whereas yield management borrowed from airlines could let early or flexible travelers fill otherwise-perishable empty seats.
“if you do miss your train, the fact that the value of your ticket immediately goes to zero, I think humans perceive as an injustice.”
- Putting existing but previously unmapped rail lines (Silver Link Metro) onto the London Underground Map as the Overground raised usage by 400% in the first month, demonstrating that roughly three billion pounds' worth of effective infrastructure value was created merely with ink (or its digital equivalent) by changing choice architecture.
“When those lines appeared on the London Underground Map, usage went up by 400% in the first month alone.”
- As models become widely used, gains come not from following what the model tells you but from exploiting what it leaves out: because everyone uses the London Tube Map to choose where to live, undervalued property sits next to well-served South London rail stations absent from the map—a decision-making arbitrage available to those using a different model of choice.
“the more people start using them, the more you'll find gains, not in looking what the model tells you, but in looking what it leaves out.”
Economic models based on perfect information and context-independent rationality are dangerously unrealistic and misleading.
- The assumptions of standard economic models—that people have perfect information and decide in an atmosphere of perfect trust, with desires independent of context or meaning—create a fantasy world in which marketing and advertising need not exist, because goods would simply sell themselves; the danger is that this theoretical frame has become the default decision-making frame in business school curricula.
“if everybody knew simply what it was they wanted and how much they're prepared to pay for it, and that their desire for the object was independent of context or meaning. Then patently in such a world, you wouldn't need any marketing activity at all.”
- Sunk cost bias and loss aversion may not really be biases: evolution would likely have corrected them if they were consistently bad, and in a non-ergodic environment where utility is multiplicative rather than additive, variance reduction and the avoidance of significant misfortune is perfectly rational because fortune is intertwined and path-dependent.
“in a non-ergodic environment where utility is to some extent multiplicative, not additive, variance reduction, of course, is perfectly rational.”
- Many consumer paradoxes are explained by avoiding the worst case: men hesitate to order a cocktail for fear it arrives in a hollowed-out pineapple and order a beer instead, and people resist switching banks or broadband because a 1% chance of a nightmarish experience dwarfs a 99% chance of a 5% gain—rational when bad outcomes are multiplicative and clustered rather than spaced.
“A 1% chance of a nightmarish experience dwarfs a 99% chance of a 5% gain.”
- Economists' efficiency fetish drives them to commoditize goods, but commodities inhibit innovation, destroy brands, and destroy trust—because a brand's value depends on having reputational skin in the game; certified, scaled commodities (like Soviet rivets measured only by quantity or weight) remove the incentive to improve quality and the fear of cheating, whereas a named brand is rewarded for producing better goods.
“commodities actually, totally inhibit innovation. They destroy brands and they also destroy trust. Because if you think about, part of the value of a brand is that you have reputational skin in the game that's attached to the thing you make.”
Focusing exclusively on measurable outcomes like speed creates missed opportunities for unmeasured but valuable improvements.
- Travel-routing algorithms assume time minimization is the only objective, so they route travelers London-to-Bordeaux via a station change in Paris rather than an easier Lille connection; a human-curated site like seat61.com adds value precisely by encoding human preferences the algorithm ignores, raising the worry that forcing everyone through identical question paths makes markets stupider by spreading the same distortions.
“an algorithm gives you advice where it assumes that time minimization is the greatest objective.”
- Making a journey 20% faster might cost hundreds of millions while making it 20% more enjoyable may cost almost nothing, because focusing transport budgets exclusively on journey-time reduction reflects a quantification bias—time is easy to measure and model—while speed actually matters in a nonlinear way and there is a large hedonic opportunity cost in ignoring enjoyment.
“Making a train journey 20% faster might cost hundreds of millions, but making it 20% more enjoyable may cost almost nothing.”
- There is danger in obsessing over what happens to be measurable rather than what matters—the McNamara effect, named for the Vietnam-era focus on kill counts—so you can feel you made an intelligent decision because each step reflected your stated preferences, while the real preferences (like architectural quality, for which there is no Parker score) were never asked about.
“you obsess about what happens to be measurable--in that case, the kill count--and you design a wholly inappropriate, in fact, counterproductive target, simply because it's the thing that you can measure not because it's the thing that matters.”
- Applying economic reasoning to fiction reveals its implausibility, as when pointing out that opening a piano shop in a remote area of France (in Betty Blue) is foolish because the catchment area could support selling at most one grand piano per year.
“their decision to open a piano shop in a remote area of France was patently ridiculous... the likely market within that catchment area for buying a grand piano probably enables them to sell at most one every year.”
Traditional practices and social norms often encode value that narrow economic reasoning cannot see and should not discard carelessly.
- Asking ten people to each hire one person yields a very different, more conservative and conformist mix than asking one person to hire ten people, because single hires trigger high variance-reduction and blame-avoidance while batch hiring leads people to seek complementarity rather than conformity—so changing the choice architecture of appointments can create diversity without quotas.
“when we make a single hiring, our instinct for variance reduction is very, very high, or blame avoidance... If you have 10 people, you're going to go much wider and you're going to look for complementarity rather than conformity.”
- The book is in a sense conservative: it argues you should pay regard to long-established traditions and behaviors that may have value even when difficult to make sense of through a narrow lens—do not chop out society's appendixes willy-nilly, because there probably is a function, as later discovered with the actual appendix.
“the whole book is really about Chestertonian fences in that it's about behaviors which actually may have a value even though it's very difficult to make sense of them when looked at through a very narrow lens.”
- Persuasion should be tried first, then incentives, then compulsion—because persuasion and social norms have a flexibility that law and price lack: people with a good reason not to comply can ignore persuasion without being vilified, whereas legislation or price mechanisms (e.g., charging more to run dryers at night) unfairly penalize night-shift workers and others with legitimate reasons; yet governments, run by lawyers advised occasionally by economists, reach for compulsion or bribery first.
“If you have a good reason not to be persuaded, you can ignore persuasion.”
Minority preferences and veto power reshape outcomes; asymmetric preferences constrain solutions toward accommodation rather than average optimization.
- Products succeed partly because few people veto them: pizza thrives not only because people love it but because almost no one rejects it, and asymmetric preferences (Muslims only eat halal while non-Muslims are indifferent) lead to outcomes like all New Zealand lamb being halal—an instance of Nassim Taleb's minority rule.
“one of the reasons pizza is so successful as a food is not only that people love it; it's that very few people veto it.”
- Demanding that comedy never say anything to which someone somewhere might object is like requiring a restaurant to serve no food to which anyone could be allergic—an extraordinary curtailment because it ignores context; standards of acceptability should vary by setting, since a comedy club is a different context from a job interview.
“What I think political correctness is sometimes demanding of comedy is that you can't serve any food to which someone somewhere might be allergic.”
Advertising's real function is reliable signaling of seller confidence and quality commitment, not manipulation.
- Advertising performs an alchemical trick by directing attention to a comparative strength rather than a comparative weakness, and what we pay attention to becomes more important in our minds—so 'Avis: We're Number Two, so we try harder' reframes a weakness (smaller scale) into a strength (customer effort), since the bare fact 'Avis is Number Two' is otherwise an ad for Hertz.
“advertising can perform this kind of alchemical trick of turning--literally--turning a weakness into a strength by telling a different story about it.”
- To determine what is dishwasher-safe, put everything in the dishwasher and whatever survives is dishwasher-safe by definition.
“just put everything in to start with and whatever survives is okay.”