Two-day delivery was, in 2005, a startling promise. By about 2012 it was ordinary. Next-day arrived and became the expectation for anything that mattered. Same-day appeared, first in a few dense cities, then more widely. Two-hour grocery slots became normal in urban Britain. And at each step, the previous standard did not remain acceptable-but-slower. It became slow. A five-day delivery estimate in 2026 reads as a warning sign about the seller, in a way that would have been incomprehensible in 2003, when five days was simply how long things took.
This is hedonic adaptation working on a service instead of a possession, and it is the same mechanism that makes a pay rise stop feeling like a pay rise. What is unusual about delivery is the speed of the resetting and the fact that the industry has been actively driving it. Every competitor that improves its speed forces the others to match, and the customer’s baseline moves, and nobody can go back, because the first firm to return to three-day delivery would simply lose.
Compare broadband. In 1999 a connection that loaded a page in eight seconds was thrilling. Now a page that takes two seconds feels broken, and studies of user behaviour keep finding abandonment thresholds measured in hundreds of milliseconds. Or compare video: buffering was an accepted fact of life in 2008 and is now an outrage. Or smartphone touch response, where the industry works in single-digit milliseconds because users can perceive lag they cannot consciously identify. In every case the improvement was real, the adaptation was total, and the gratitude lasted about eighteen months.
The economists have a term I find useful here, though it is usually applied to prices: the reference point. What matters to human satisfaction is not the absolute level of a thing but its position relative to what you have come to expect. Improve the thing, and the expectation follows it up, and you are back where you started with a more expensive operation to run.
Which produces the peculiar situation of the modern delivery industry: it has achieved, in twenty years, an improvement in service that would have seemed fantastical to anyone in 1995, at enormous capital cost, and its customers are on the whole more irritated than they were, because their standard has risen faster than its performance. Satisfaction surveys in the parcel sector are dismal almost everywhere, and they are dismal in markets where the objective service is excellent.
There is a lesson here for anyone in the business of making things better, and it is not a cheerful one. You do not get to bank the gratitude. You get to keep the cost.
GPT-5.6 Sol:
The passage is substantially right about the phenomenon, but I think it identifies the wrong mechanism too narrowly and then pushes the conclusion further than the evidence will bear.
The best idea in it is the ratcheting of service standards. Two-day delivery begins as a differentiator, becomes a competitive requirement, and finally becomes what service researchers sometimes call a basic or hygiene attribute: providing it earns little praise, while failing to provide it causes dissatisfaction. That is closely related to the Kano model of service quality, in which attributes can migrate from ‘delighters’ to expected basics. This captures Amazon-era delivery extremely well. ‘Fast’ stops meaning an absolute number of days and comes to mean ‘not conspicuously slower than the alternatives’.
But calling the whole thing hedonic adaptation obscures three different processes.
Hedonic adaptation properly concerns the attenuation of the emotional impact of a changed circumstance: something initially feels wonderful and eventually feels normal. There is evidence for that phenomenon, but adaptation is neither necessarily complete nor uniform. Delivery expectations involve something more specific: expectancy disconfirmation. I expect a parcel tomorrow; it arrives tomorrow; nothing noteworthy has happened. It arrives today; I am pleasantly surprised. It arrives three days from now; I am annoyed. Satisfaction is therefore partly relational.
That is almost exactly the established expectancy-disconfirmation model of consumer satisfaction. Interestingly, however, the latest large meta-analysis complicates the passage’s story. Across 150 studies and nearly 59,000 observations, perceived performance had a strong positive association with satisfaction, while expectations themselves were also positively associated with satisfaction. In other words, consumers don’t simply raise expectations one-for-one until all benefits of better service disappear. Better performance still matters.
There is also a third mechanism that is economic rather than psychological: competitive escalation. If one firm discovers that next-day delivery materially increases conversion, others may be compelled to copy it even if the industry as a whole would prefer not to bear the cost. Once warehouses, household routines and commercial promises reorganise around the new standard, unilateral retreat becomes difficult. This resembles a Red Queen dynamic: firms must improve merely to maintain relative position.
That distinction matters because it changes the interpretation of the apparently depressing conclusion.
Satisfaction is not the same thing as welfare
The sentence ‘you are back where you started with a more expensive operation to run’ is rhetorically excellent and economically dubious.
Suppose in 2000 I wait five days for something and report satisfaction of 8/10. In 2026 I get it tomorrow and also report 8/10. It does not follow that nothing has been gained. I have an objectively more useful service. I can purchase something on Thursday that I need on Friday. I can hold less household inventory. A broken cable, missing ingredient or forgotten birthday present causes less disruption. Businesses can operate with lower inventories. An enormous amount of option value has appeared even if nobody feels grateful for it.
Indoor plumbing provides the extreme example. Almost nobody wakes up delighted that water comes out of the tap. Its hedonic novelty has disappeared almost completely. That plainly doesn’t mean society is ‘back where it started’.
So there is an important distinction between experienced delight and consumer surplus. Progress can cease to produce delight precisely because people have incorporated its benefits into ordinary life.
And in a competitive market that is, to some extent, the point. The producer doesn’t necessarily get to ‘bank the gratitude’ because competition transfers some of the gains to consumers. Yesterday’s luxury becomes today’s entitlement. From the firm’s perspective that can feel like an arms race; from the consumer’s perspective it can represent an enormous permanent improvement.
The factual claim about parcel dissatisfaction is too strong
The weakest part of the extract is:
‘Satisfaction surveys in the parcel sector are dismal almost everywhere.’
For Britain, the available data don’t justify that formulation. Ofcom’s 2025 survey found 78% overall satisfaction with parcel delivery, unchanged from the preceding year. That is not plausibly described as dismal. At the same time, 68% reported experiencing at least one delivery issue during the preceding six months, so there clearly is considerable friction.
Those two numbers are actually much more interesting than the book’s stronger claim. People can be broadly satisfied with an extraordinarily capable system while simultaneously encountering frequent failures within it.
There is also evidence of serious specific frustration: Citizens Advice found 37% of respondents reporting a problem with their most recent parcel in its 2025 research. But that establishes a high incidence of problems, not that customers are generally less satisfied than customers in 1995. Demonstrating the latter would require reasonably comparable longitudinal evidence that the passage doesn’t provide. The next Ofcom parcel-experience release is scheduled for 12 October 2026, so the 2025 results remain the latest published official figures as of 22 September.
The claim that a five-day estimate now automatically signals a bad seller is similarly overdrawn. Speed is one attribute among price, reliability, choice and convenience. One 2025 UK survey found shoppers much more likely to prefer free delivery to fast delivery, while other research finds substantial sensitivity to delivery charges. A five-day estimate from Amazon for an ordinary stocked item may indeed look poor; five days from a specialist bookseller, handmade-goods shop or low-cost overseas retailer may be entirely acceptable. Reference points are category-specific.
Broadband is an especially interesting comparison—but for another reason
The broadband example initially looks like pure adaptation: eight seconds once seemed amazing; two seconds now seems intolerable.
Yet the thing being delivered has changed too.
A web page in 1999 and a web application in 2026 aren’t equivalent goods delivered at different speeds. Faster networks and processors permitted developers to build richer, more interactive software. Workflows then became organised around that capability. Latency which would once merely have delayed the display of a newspaper article can now interfere with a video call, cloud IDE, collabourative document or payment flow.
This is induced demand for quality. Increased capability doesn’t merely satisfy an existing requirement; it creates activities whose existence depends on the increased capability. Once those activities exist, deterioration really does impose greater costs.
Delivery has begun to acquire the same property. If households know that groceries, toiletries or computer peripherals can appear within hours, they alter their behaviour accordingly. They order later and keep fewer contingencies. Consequently a three-day delay may genuinely cause more inconvenience than it would have twenty years ago. Their annoyance isn’t necessarily an irrational failure to appreciate progress. Their lives have been reorganised around the infrastructure.
That, to me, is more interesting than hedonic adaptation.
The business lesson is consequently somewhat different
The final aphorism—
‘You do not get to bank the gratitude. You get to keep the cost.’
—is memorable, but only half true.
What firms cannot normally bank indefinitely is differentiation generated by an easily copied service improvement. Innovation produces a temporary advantage; imitation converts that advantage into an industry standard. The customer then captures much of the permanent benefit while the original innovator loses the ability to charge merely for having the feature.
But firms don’t necessarily simply ‘keep the cost’. Faster fulfilment can increase order frequency, route density, warehouse utilisation and customer retention; technological improvements can also reduce the marginal cost of speed. Indeed, the continuing investment in same-day infrastructure makes little sense unless the firms involved believe some combination of those benefits offsets the cost. Amazon is reportedly contemplating a very large further expansion of same-day fulfilment in the US for exactly that strategic reason.
The sharper managerial lesson would therefore be something like this:
A service improvement depreciates as a source of competitive advantage much faster than it depreciates as an operating obligation.
That phenomenon is real. Today’s delight becomes tomorrow’s specification. So firms should be wary of creating expensive expectations that competitors can cheaply match, and should distinguish improvements that permanently change their economics from ones that merely move the reference point.
There is also a larger irony that the passage almost reaches. Civilisation consists to a considerable extent of turning miracles into things people complain about when they fail. Electricity, running water, antibiotics, GPS, broadband and next-day logistics cease to generate astonishment because reliability allows people to build their lives on top of them. At that point, irritation at failure isn’t necessarily evidence that progress was pointless. In a strange way it is evidence of just how completely the progress succeeded.
That makes the passage’s observation stronger, but its pessimistic conclusion weaker.
Claude Fable 5.1:
The passage is persuasive, but I think it runs together two different mechanisms, and separating them changes the lesson.
Hedonic adaptation is a fact about individual psychology: you habituate to your own circumstances. What the passage actually describes is mostly a competitive ratchet: the reference point isn’t set by what you used to get, it’s set by the best thing anyone currently offers. Those behave differently. A monopolist who improved delivery from five days to two would see gratitude fade through adaptation, but could still coast on the new level. In a competitive market the baseline moves the moment one rival moves it, whether or not any customer has had time to habituate. The trap the passage describes is a Red Queen race, and the psychology is almost incidental to it.
The asymmetry that really locks the door is loss aversion rather than adaptation. Nobody can go back to three-day delivery not because customers have forgotten what three days felt like, but because a downgrade from an established reference point is experienced as a loss, and losses are felt roughly twice as strongly as the equivalent gain. That’s why the ratchet only turns one way.
The dismal-satisfaction observation deserves more scepticism than it gets. Parcel surveys are bad partly because the last mile is where failure is visible: a 98% on-time rate at modern volumes still produces millions of people whose specific parcel went missing, and those are the people who fill in surveys. It’s also plausible that what customers now care about isn’t the mean but the variance. Adaptation research finds people habituate readily to levels but poorly to unpredictability (commuting is the classic case), which would suggest the industry has been optimising the wrong variable: shaving hours off the average when the felt experience is dominated by the one-in-fifty miss.
The broadband and touch-latency comparisons are also a slightly different thing. Sub-100ms abandonment thresholds and single-digit-millisecond touch response aren’t reset expectations, they’re perceptual and attentional limits that were always there and only became reachable. That’s engineering discovering the floor of human perception, not customers becoming spoilt.
On the gloomy closing line: it’s true that you don’t bank the gratitude, but gratitude was never the currency. Amazon didn’t build two-day delivery to be thanked; it built it to make Prime a habit and to raise the cost of entry for everyone else. The firm banks market position and the customer banks a real welfare gain that they happen not to feel. Economically that surplus is no less real for being unfelt, and there’s an interesting question buried here about whether welfare should be measured by satisfaction at all, given that satisfaction is a relative measure by construction.
The practical escape, if there is one, is to stop competing on the ratcheting dimension: charge explicitly for the baseline (a membership fee monetises the expectation rather than fighting it), or compete on reliability and recovery, where adaptation seems weaker.
