Case study — Technology · 12 min read · 5 questions
Why Temu and Wish can't make money
The thesis
What made a $5 item shippable across the Pacific for free was not technology. It was the Universal Postal Union, an 1874 treaty that heavily subsidized mail leaving developing countries for wealthy ones. China kept that classification into the 2010s, so a package from Guangdong to Ohio cost less than the same package from California. The treaty covers items under 4.4 pounds, which is exactly why everything on Wish, AliExpress and Temu is a trinket or a gadget, and why none of them has ever moved into furniture.
Wish built a company on that arbitrage and dressed it as software. It called itself a data-science platform and spent under 5% of revenue on R&D while spending 70–90% on sales and marketing — $1.58 billion in 2018 against $1.73 billion of revenue. It bought customers for $17 to $25 who were worth $10 to $30. Active buyers plateaued around 62 million from 2018 through three consecutive billion-dollar ad years. It lost money every year of its existence.
Then the subsidy went. The US threatened to leave the UPU, rates were reformed, and postage from China jumped roughly 50% overnight. Wish had to become the thing it promised investors it never would: warehouses, logistics, inventory. Gross margin fell from 84% to 29% — below Nike, below Vans — so you would make more money selling branded clothes than running the platform. AliExpress reached the same place and has never profited either; Alibaba carries it at under 6% of revenue. Temu is the same business again. The software platform was never the business. The postage was.
How do you think about this? 5 strategy questions this case raises and does not answer.
Read the comments, or add yours →
The statistics
By the numbers — swipe or use arrows
Wish revenue, operating income, gross profit, operating expenses by line, active buyers, advertising spend, adjusted EBITDA and cash position from ContextLogic Inc. S-1 and 10-K filings for the years shown; comparative sales and marketing ratios and gross margins for Amazon, Airbnb, Uber, Spotify and eBay from company filings; brand and manufacturer gross margins from company annual reports; AliExpress and Alibaba segment shares from Alibaba Group annual reports, converted from RMB; Pinduoduo revenue and US-market valuation from company filings and Nasdaq data; cross-border postal rates are Universal Postal Union terminal dues for items under 4.4 pounds
Key takeaways
The whole category rests on a treaty from 1874. The Universal Postal Union subsidized mail from developing countries to wealthy ones, China kept that classification into the 2010s, and shipping a parcel from China to the US cost less than shipping it coast to coast inside the US.
It only covers parcels under 4.4 pounds, which explains the entire product catalog. Trinkets, gadgets, compact electronics and anything that folds — and no furniture, no appliances, no home decor, on any of these platforms, ever, despite all three having had war chests.
Wish grew revenue $144M, $445M, $1,101M, $1,728M, $1,901M, $2,541M and lost money every year: −$522M, −$141M, −$147M, −$223M, −$144M, −$631M. It never had a profitable year in its existence.
The self-declared data-science company spent almost nothing on data science. R&D was $6M, $10M, $28M, $45M and $74M against revenue of $144M to $1,901M — under 5% of revenue every year. The founders were engineers. They knew where the returns were not.
They were in advertising, and it ate everything. Sales and marketing ran $560M, $429M, $989M, $1,576M and $1,463M — 70% to 96% of revenue. For every dollar Wish earned, up to 90 cents went straight back out to Facebook, Instagram and YouTube.
That is not what platforms do. Sales and marketing was 96%, 90% and 91% of Wish revenue across 2017 to 2019 against 12%, 12% and 14% at Spotify, with Airbnb and Uber both around 30%.
Gross profit was never the problem: $66M, $314M, $896M, $1,450M, $1,458M. Operating expenses were $588M, $455M, $1,043M, $1,673M, $1,602M. The marketplace worked. Everything downstream of it did not.
The unit economics were inverted from the start. Customer acquisition cost ran $17.31, $23.44, $22.58 and $25.00 while lifetime value ran $10, $20, $25 and $30. At a 15% commission, a $30 lifetime value means a customer spent about $50 a year on the platform.
And the spending stopped working. Active buyers went 18M, 31M, 52M, 64M, 62M, 64M — flat from 2018 — while ad spend went $900M, $1,500M, $1,400M and $1,600M. Three consecutive billion-dollar years bought no growth at all.
Adjusted EBITDA is the metric a company defines for itself, and Wish's was no better than the audited one: operating margin of −363%, −32%, −13%, −13%, −8%, −25% against adjusted margins of −349%, −30%, −12%, −12%, −7%, −9%. There was nothing left to spin.
Then the arbitrage closed. The US threatened to leave the UPU, rates were reformed, and cross-border postage from China rose roughly 50% overnight — from a maximum of $5.41 for a sub-4.4-pound parcel to $10.77, against $11–25 to send the same thing inside the United States.
So the software-only platform had to build the thing it said it would never build. Sales commission went $1,053M, $1,508M, $1,473M, $1,827M, $1,177M, $220M while logistics services went from $6M to $743M — table stakes for any marketplace, arriving years late.
Owning distribution destroyed the margin story. Wish gross margin fell 84%, 81%, 71%, 63%, 53%, 29% while Amazon climbed 37% to 44%. The disruptor converged on the incumbent from above and kept going.
At 29%, Wish now runs a worse gross margin than the manufacturers it was supposed to be disintermediating: 44% at Amazon, 45% at Under Armour, 46% at Nike, 47% at adidas, 53% at Vans, 67% at Canada Goose, 70% at eBay, 73% at Airbnb and 82% at LVMH. Selling branded clothes pays better than running the platform.
AliExpress survives only because nobody at Alibaba needs it to work. It has been 5–6% of group revenue since 2018 against 67–74% for Tmall, it ran revenue of $6.8B, $8.5B and $9.6B against adjusted EBITDA of −$0.7B, −$1.2B and −$0.8B, and it has never turned a profit in its life.
Temu exists because Pinduoduo needs a growth story for American investors who cannot use its Chinese app. PDD went from $0 revenue in 2015 to $4.3B by 2019 and its US-listed valuation ran $24B, $44B, $100B, $211B, then down to $73B and back to $200B — level with Alibaba at $195B and four times JD.com at $46B.
Common questions
How is Temu so cheap?
Three things stack. The sellers are Chinese manufacturers, wholesalers and distributors operating at a much lower cost base, shipping generic unbranded goods that cost cents to produce. The products are non-essential impulse items with no brand to protect, so quality is whatever clears the price. And the postage is subsidized: the Universal Postal Union has set below-cost cross-border rates for parcels under 4.4 pounds leaving countries classified as developing, which China was until the rates were reformed. That is why the catalog is all small, light, foldable goods and why none of these platforms sells furniture or appliances. On top of that, Temu is deliberately loss-making — Pinduoduo funds it out of a profitable Chinese business to buy a foothold in the US market.
Is Temu profitable?
Not as a standalone business, and it is not designed to be yet. Pinduoduo has said publicly that Temu's financial impact is small and that investors should not expect continued high growth from the core platform, which is precisely why Temu exists — PDD's group-buying model does not translate to the West, its Chinese user base has plateaued at around 880 million, and its US-listed valuation depends on a growth story. Temu's offer of 90-day refunds, 30-day price adjustments, free shipping and two-week delivery costs more than the margin on a $10 order can support. It is subsidized entry, funded by a mature core business, exactly as AliExpress is by Alibaba.
What happened to Wish?
It ran out of the arbitrage it was built on. Wish grew revenue from $144 million in 2015 to $2.5 billion in 2020 and never posted a profitable year, because up to 90 cents of every revenue dollar went to advertising to acquire customers worth $10 to $30 over their lifetime. Active buyers plateaued around 62 million from 2018 despite three straight billion-dollar ad years. When the US forced a reform of Universal Postal Union rates, cross-border postage from China rose about 50% overnight and Wish had to start building warehouses and logistics — which took gross margin from 84% down to 29%. It has gone through three CEOs in two years and now survives on a 30-to-1 reverse split to stay listed on Nasdaq.
Why does AliExpress still exist if Wish failed?
Because Alibaba does not need it to make money. Alibaba's business is B2B wholesale and Tmall, which is 67–74% of revenue; AliExpress has been under 6% since 2018 and has never turned a profit. It targets emerging markets — Russia, Spain, Brazil — where e-commerce and product standards are less developed, rather than competing for low-income households in Amazon's home market. Its roadmap has been identical to Wish's over the same decade: an unregulated marketplace on flat commission that gradually curates, localises, and starts charging merchants for logistics to cut delivery times. It reached the same destination. The difference is that its losses are rounding errors inside a profitable group.
Why isn't a software marketplace automatically a good business?
Because you are compensated for the problems you solve, and a marketplace that refuses logistics, quality control and returns has only solved the easy one. Wish had software-level gross margins of 84% and still lost money, because the hard part — convincing someone with little disposable income to buy a non-essential item, sight unseen, from an unknown seller, with a four-week delivery — had to be bought with advertising every single time. Taking 10–20% commission on a $5 order is a fundamentally different business from taking it on a $50 order, and no amount of volume fixed the ratio. When Wish finally started doing the hard parts, its margins converged on Amazon's and then fell below the brands' entirely.
Was Wish's data-science moat real?
No, and the spending says so. Wish pitched proprietary data science, personalization and network effects as its competitive advantage, then spent under 5% of revenue and no more than 5% of operating expenses on R&D every year — $6 million in 2015 rising to $74 million in 2019 — while spending up to 96% of revenue on advertising. Companies fund what they believe returns. The founders were engineers who knew that recommendation algorithms deliver marginal improvements, not a moat, and that what Wish was actually doing was what Amazon and eBay had already done years earlier. The discovery feed was the dollar-store treasure hunt in software, which is a merchandising choice rather than a technology.
Discussion
Free shipping for a $5 item across the Pacific was not technology — it was an 1874 postal treaty that subsidized mail from developing countries. How much of what gets called disruption is actually a rule nobody had read?
No answers yet — be the firstThe treaty covers items under 4.4 pounds, which is exactly why everything on Wish, AliExpress and Temu is a trinket and none has moved into furniture. What does it tell you when a company's entire catalog is shaped by a weight limit?
No answers yet — be the firstWish called itself a data-science platform while running an arbitrage. Why do companies reach for the technology framing, and who is it aimed at?
No answers yet — be the firstWhen the subsidy that makes a business viable is a policy, what is the correct discount to apply to its valuation?
No answers yet — be the firstYou are advising a marketplace built on a regulatory advantage that is about to close. What do you do with the time you have left?
No answers yet — be the first
Related case studies
TechnologyWhy AI is just big data againBig data was sold for a decade and produced nothing. The same players are running the identical playbook with AI.12 min read
TechnologyWhy Insta360 beat GoPro at its own gameGoPro invented the category. A startup took 82% of the market and now owns a future that nobody thought was possible.8 min read
TechnologyWhy the scooter startups all went bankruptA Bird scooter grossed $2,000 a year, cost $1,300 to buy and $1,000 to maintain. It lost money on every vehicle, always.11 min read