The End of Cheap Growth

Chinese marketers have a phrase that Western marketers never quite adopted: liuliang hongli, the traffic dividend. It spread through Chinese business writing in the 2010s to describe the years when hundreds of millions of people came online and a merchant could reach them for very little. The word carries a useful humility. A dividend is paid because you held something at the right moment. You did not earn it by being clever, and it can be cut.

Western marketers, living through the same years, chose a different word: performance. One word describes a windfall. The other describes a skill. A whole profession built its self-image, its org charts and its dashboards on the second word.

Our argument is simple. Customers were cheap for roughly a dozen years because several unrelated conditions overlapped. Nobody designed that overlap. It is now closing, at different speeds in different markets, and the closing is not a failure of marketing. It is an auction finishing its work.

Four accidents that happened to overlap

The first accident was a flood of new attention. The iPhone arrived in 2007 and Android followed. Within a few years billions of people carried a screen that sold advertising space every time they looked at it. Inventory appeared faster than anyone could learn to buy it.

The second accident was thin auctions. The big budgets stayed in television and print for years, held there by habit, agency contracts and the way large firms plan. So the early digital auctions were lightly contested. A small brand with a good landing page often faced a handful of rivals bidding on instinct.

The third accident was cheap money. The US federal funds target sat near zero from December 2008 to December 2015, and again from March 2020 to March 2022. Cheap money did not lower the price of a click. If anything, it raised it. What it lowered was the price of waiting. A payback of three years looks fine when investors reward growth and capital costs almost nothing. It also let firms subsidise the customer's side of the trade. Uber lost more than $8 billion in 2019 alone, and every discounted ride made its marketing look better than it was.

The fourth accident was free data. For most of this period, tracking a person across apps and sites required nobody's meaningful consent. Targeting was precise because the raw material cost nothing. It was taken from an account that consumers had not realised they held: permission.

Each condition had its own cause. Together they produced a window in which a modest budget could buy a customer at a price well below what that customer was worth. The window felt like the natural state of digital marketing. It was closer to a gold rush in which the land, the tools and the loans were all temporarily under-priced at once.

This is why we say all growth is borrowed. Cheap growth drew on investors' patience, consumers' unexamined permission and the immaturity of a market that had not yet found its price. None of those balances appeared on a dashboard. All of them have since been called in, or are being called in now.

Why the auction always finishes

The heart of the matter is a piece of auction logic that rarely reaches the boardroom.

In the keyword and social auctions that shaped this era, the price you pay is driven by the next-best bid. Economists described the structure in detail in 2007 (Varian; Edelman, Ostrovsky and Schwarz). The exact rules vary by platform and have changed over time, but the principle holds. The price of a customer is set by what your strongest rival thinks that customer is worth.

Take a hypothetical. A customer is worth 100 to you and 90 to your rival. The rival does not know this. It is unsure, so it bids 30. You win at a little over 30 and keep about 70 of surplus. Your team reports a fine return.

Now the rival gets better data, a better bidding tool and a better estimate of lifetime value. It bids 85. You still win. You now pay about 85 and keep 15. Your creative did not get worse. Your team did not lose its touch. Fifty-five units of margin that you had booked as skill have moved to the platform, because your rival stopped being ignorant.

That is the mechanism behind the long rise in acquisition costs. Advertisers often blame platforms for greed, and platforms do control ad load and reserve prices. But honest competition alone would produce most of the effect. As bidders grow better informed, surplus moves from buyers to the seller of the attention. Every improvement in automated bidding, sold to every advertiser at once, speeds that transfer.

There is a nastier second step. Many bidders value customers using attributed numbers, which often overstate what the advertising actually caused. An auction full of overconfident bidders can clear above true value. Economists call this the winner's curse. In plain words, the firm that wins the customer is often the firm that most overestimated them.

How the closing shows up inside a company

From inside, the end of cheap growth rarely looks like a structural shift. It looks like a run of disappointing quarters. Blended acquisition cost creeps up. Payback stretches. Campaigns that worked stop working. The usual response is to change the agency, the creative or the head of growth. These are reasonable moves against the wrong diagnosis.

Two events show the structure more clearly than any dashboard.

The first is data. In April 2021 Apple's App Tracking Transparency made cross-app tracking a question the user could answer. In February 2022 Meta estimated that Apple's changes would cost it about $10 billion in revenue that year. That is what it looks like when an unpriced input acquires a price. Europe had started the same process with GDPR in 2018, and India passed its Digital Personal Data Protection Act in 2023.

The second is a new bidder. From 2023, Temu and Shein spent heavily in Western auctions, and advertisers widely reported higher prices as a result. These firms valued a customer on a different basis from a domestic retailer, with different cost structures and strategic aims. When US rules on low-value parcels changed in 2025, Temu reportedly cut its American advertising sharply, and some advertisers reported softer prices. If those reports hold, this was close to a natural experiment. The price of a customer in Ohio moved because of a bidder based a long way from Ohio, and nothing about the customer had changed.

For a CFO, the lesson is uncomfortable. Your acquisition cost is not a property of your business. It is a property of your competitors' beliefs.

The window closes at different speeds

Treating this as one global story would repeat the original mistake. The dividend is paid and cut market by market.

China reached the end early. By the late 2010s its business press was already writing the dividend's obituary, as user growth slowed and merchants moved towards private traffic: customer groups held in WeChat and other owned channels. That move is worth studying. It was an attempt to stop renting customers from an auction.

India received a second dividend. Reliance Jio's launch in 2016 collapsed the cost of mobile data, and hundreds of millions of people came online through cheap Android phones. The ban on TikTok in 2020 then pushed that attention towards Reels and Shorts, and venture funding poured into consumer brands. The funding slowdown from 2022 exposed the weakness. India's lesson is that a cheap click is not a cheap customer. Reach is inexpensive in much of the country, but revenue per user is low too, so the ratio of value to cost can be poor even while the media looks like a bargain. Brands selling to the first hundred million affluent consumers already face mature, contested auctions. The cheaper reach lies further out, where baskets are smaller. Indian marketers should therefore distrust both halves of the common story: the country is not in a Western-style squeeze everywhere, and its low media prices do not mean easy growth.

Southeast Asia lived through its own version in the subsidy wars among super-apps and marketplaces. Around 2022, firms such as Sea and GoTo turned towards cost-cutting and profit once capital grew more expensive. Much of Africa still has growth in connected users, but thin purchasing power and fragmented payments limit how much of it can be turned into revenue.

Latin America offers the most useful exception. Nubank in Brazil was reported to acquire customers at a very low cost, mainly through word of mouth, in a market where people disliked their banks. It did not win the auction. It avoided it. Its growth was still financed from an account, but the account was the goodwill of customers who felt a real difference.

What comes after the dividend

If the price of auctioned attention tends towards the value of the customer, then the auction cannot be a durable source of advantage. It can only be a place where you pay fair price, or more. Advantage must then come from one of two places.

The first is making a customer worth more to you than to anyone else. Better retention, better margin and products that spread by referral all do this. A firm whose customer is truly worth 150 can pay 85 and live well.

The second is reaching customers without passing through the auction at all. That means memory, habit, distribution, owned audiences and physical presence. These take years to build, which is why cheap growth discouraged building them. While the auction was a bargain, nothing could justify the slower investment. For a generation of firms, the dividend paid for not learning how to do anything else.

Consumers feel the change too. More ad load, more paywalls, more loyalty schemes that ask for data in exchange for a discount: all are attempts to replace an input that used to be free. Expect permission to be bought more openly, and expect people to notice its value.

The strongest objection deserves a straight answer. Cheap pockets keep returning. Early TikTok was one. New retail media networks may be another, and AI assistants could become a new channel with thin competition. Regulation has also been less decisive than predicted. Google gave up on removing third-party cookies from Chrome after years of promising to do so. And if AI lifts conversion rates, each customer becomes worth more.

We accept all of that, with two qualifications. Pockets are not the same as a regime. Every new channel opens under-priced and is then competed towards value, and bidders now arrive faster, because automated tools move budgets within weeks rather than planning cycles. And a rise in conversion that everyone enjoys simply raises what everyone can bid, so it flows to the seller. Our claim is about the direction of surplus, not about any single year's prices. We could be wrong if attention supply grows much faster than advertiser demand for a long period, or if platforms choose to hold prices down to win a new market.

Scenarios

These are our subjective estimates of which pattern will best describe acquisition economics in large consumer markets by 2033. They are judgements, not measurements.

The Long Squeeze (confidence: 50%). Auction prices keep tracking customer value. Paid acquisition remains large but stops being a source of margin. Winners are firms with superior retention and strong direct demand. Performance teams shrink into smaller, more financial functions.

Pocket Arbitrage (confidence: 25%). New surfaces open repeatedly, including retail media, connected TV, creator networks and assistants. Skilled, fast firms earn short windows of cheap growth. The dividend survives in small, brief instalments, and speed of entry becomes a real capability.

The Assistant Reset (confidence: 15%). AI assistants take over much of product discovery, and auctions lose their grip on the moment of choice. Prices collapse in old channels and a new gatekeeper sets new terms. The early period looks like 2010 again, then repeats the same cycle.

The Rationed Market (confidence: 10%). Privacy and competition law in major markets sharply restrict targeting and platform power. Precision falls, reach-based buying returns, and acquisition becomes cheaper per impression and vaguer per customer. Large brands benefit most.

The dividend has been paid

Every generation mistakes its tailwind for its talent. Bond traders did it in the long fall in interest rates. Digital marketers did it in the years when attention, money and data were all on sale together.

There is no shame in having collected a dividend. The shame lies in having spent it on the belief that it was salary. The useful question for a leadership team is not how to get cheap growth back. It is what the firm bought while it lasted, and whether any of that still earns a return now the payments have stopped.