Primary sources for this piece include Alibaba’s Form F-1 (SEC, September 2014), eBay’s FY2002 and FY2003 10-Ks, and Yahoo’s 2005 and 2011 SEC filings — figures from these are treated as fact. Below that tier: Porter Erisman, an Alibaba vice president from 2000 to 2008, whose book “Alibaba’s World” and documentary “Crocodile in the Yangtze” are the principal insider accounts; Duncan Clark’s “Alibaba: The House That Jack Ma Built” (2016); interviews with Savio Kwan, Alibaba’s president and COO through the dot-com crash; and contemporaneous journalism from the war years. Where a claim rests on a single insider, I name the insider. Where it exists only in Jack Ma’s own retellings, I say so — this story carries more founder self-mythology than any other I’ve researched, and part of the job is separating the filmed record from the stagecraft.
Prologue: Eight O’Clock, Day Four
On the evening of May 10, 2003, at eight o’clock, a shopping website called Taobao went live from a residential apartment at Lakeside Gardens, Hangzhou.
Four days earlier, at 4 p.m. on May 6, Alibaba’s management had told its Hangzhou headquarters staff — more than four hundred people — that the building was closing immediately. An employee named Song Jie had flown to Guangzhou on April 11 for the Canton Fair, at the height of the SARS outbreak; the company had sent her. She returned April 18, developed a fever, and on May 7 was confirmed as Zhejiang province’s fourth SARS case. The headquarters was quarantined from May 6 to May 19; the building emptied in roughly two hours.
The next day, working from home, the company set a record: more than 12,500 business leads posted to its B2B marketplace in a single day.
In the founding apartment, meanwhile, a team of six or seven engineers kept working. They had been there for weeks, since before anyone had heard of Song Jie’s fever, pulled out of headquarters by Jack Ma and sworn to secrecy to build a consumer marketplace aimed at eBay. The taobao.com domain had been registered April 21. The launch page carried a dedication — in TechNode’s translation, “Think of those who start a business in trying times.”
The compressed legend — SARS locked Alibaba down, and in the lockdown they invented Taobao — has the causality backwards. The skunkworks predates the epidemic’s arrival in Hangzhou; the quarantine is not when Taobao was conceived but when it shipped. It’s the pattern of this entire story: the verified record beats the myth, and the myth is usually Ma’s own.
Three and a half years later, eBay — then holding 70 to 85 percent of Chinese online auctions, depending on the measure — announced it was shutting its China site.
Part I: The World They Were Born Into
E-commerce is three stacked promises: you can pay, you can trust, and the goods will arrive. In China in 1999 all three layers were missing — credit cards rare, consumer protection weak, parcel logistics thin outside the big cities. Every credible account of the period treats trust, specifically who sends first, money or goods, as the binding constraint on Chinese online commerce. The company that solved that constraint won — and not on price alone.
The other half of the environment was the export economy. China’s small manufacturers produced at globally competitive cost but were locked out of the state-dominated channels connecting factories to foreign buyers. On December 11, 2001, China joined the WTO — and “help a Chinese factory find a foreign buyer” went from useful service to rising tide.
When Taobao launched in 2003, China had somewhere between 68 and 90 million internet users, depending on the survey — against roughly 300 million mobile-phone users, a gap Forbes later flagged: most of the coming consumer market wasn’t online yet, let alone looking at banner ads. By the end of 2013, per CNNIC figures in Alibaba’s F-1, China had 618 million internet users and 302 million online shoppers. Whoever owned Chinese e-commerce would own the largest retail migration in history.
The incumbent was EachNet, founded in 1999 by Shao Yibo — an eBay-style auction site that eBay itself would buy in stages. Around the completed acquisition in mid-2003, Forbes accounts put EachNet at roughly 85 percent of China’s online-auction market with about 4 million users; the case literature says 70-plus percent. Either way: Taobao launched against a dominant, well-capitalized, American-owned leader.
Part II: An English Teacher and Seventeen Co-Founders
Ma Yun — Jack Ma — was born September 10, 1964, in Hangzhou. He learned English giving free tours to foreign tourists around the Hangzhou International Hotel, in his own tellings cycling some twenty-seven kilometers to do it. He failed the gaokao twice — 1982 and 1983 — passed in 1984, took an English degree at Hangzhou Normal University in 1988, and taught English and international trade at Hangzhou Dianzi University.
Here the record and the stagecraft need separating. The famous KFC rejection — twenty-four applicants, twenty-three hired, only Ma turned away — comes from Ma’s own speeches; no independent record exists. Harvard rejecting him ten times is likewise his claim, with no evidence he ever applied. The thirty-odd job rejections: self-reported. Duncan Clark flags the pattern: where Ma’s tellings and the record diverge, Ma’s version is always more cinematic, always the underestimated underdog. Not false, necessarily — unverifiable, and told by a professional mythmaker. I present them as his tellings.
The documented arc doesn’t need them. In 1995 Ma saw the internet on an interpreting trip to Seattle — in his retelling he searched for “beer,” then for anything Chinese, and found almost nothing. He came home and founded China Pages (domain registered May 10, 1995), one of China’s first commercial websites, and lost control of it in a joint venture with Hangzhou Telecom — his first lesson in what shared control costs. In 1998 he went to Beijing to run an IT unit under the foreign-trade ministry; in that period he guided a visiting Yahoo co-founder, Jerry Yang, along the Great Wall. File that away; it becomes a billion dollars in 2005.
In February 1999, back in Hangzhou, Ma gathered seventeen colleagues in his apartment at Lakeside Gardens. The eighteen co-founders — a third of them women, including Lucy Peng, and including Joe Tsai, a lawyer and private-equity investor who, Clark reports, walked away from roughly $700,000 a year to earn about $600 — pooled RMB 500,000, around $60,000. The legal entity, a Cayman Islands holding company, followed on June 28, 1999, per the F-1.
The founding meeting was filmed; the footage survives in Erisman’s documentary, so for once we need no one’s retelling. On tape, February 1999, Ma says: “Our competitors are not in China, but in America’s Silicon Valley.” He predicts pain: “The dream of the internet won’t burst. We will have to pay a painful price in the next three-five years.” And he sets a target: “The goal is Alibaba will IPO in 2002.”
The verified tape — an English teacher with a failed website, promising eighteen people in an apartment that they would beat Silicon Valley and go public in three years — is more audacious than any of the myths.
Part III: The Pitch, the Skeptic, and the Retreat That Saved It
The thesis was B2B: a marketplace connecting China’s small manufacturers to global buyers. The F-1’s corporate history opens with it: “In 1999, we founded Alibaba.com and Alibaba.com.cn, the predecessor of 1688.com” — the export site and its domestic wholesale twin. Listings were free; growth came through community and trader word of mouth. By the end of 1999 the site had more than 40,000 users.
The money came in two rounds, both since hardened into legend.
First, Goldman Sachs. On October 27, 1999, a syndicate led by Shirley Lin of Goldman’s Asia principal-investment area signed: $5 million for 50 percent of the company. Ma had wanted to sell 10 percent for the same money; Lin got half the company plus board control and anti-dilution rights. Goldman’s own committee then cut its participation to roughly $3.3 million — about 33 percent — syndicating the rest; retellings blur the round’s half into Goldman’s third.
What Goldman did next is the era’s skeptic exhibit: it marked the position down 50 percent in the bust, then in early 2004 sold out entirely for roughly $22 million — a 7x return, months before the Yahoo deal implied ten times more, a decade before the stake would have been worth tens of billions. Lin’s own verdict: “Ten thousand times is very much an understatement.” One of the most expensive early sells in venture history — and a quiet mercy for Ma: the most control-laden position on the cap table left cheaply, just before the war’s decisive financing.
Second, SoftBank. In January 2000, Masayoshi Son led a $20 million investment for roughly 30 percent — sources say 30 to 34 percent; SoftBank’s 34.1 percent at the 2014 IPO includes later rounds. The deal is fact; the framing is theater from both sides. In a 2008 speech Ma said the first meeting lasted six minutes and the two “fell in love at first sight”; Son told Bloomberg in 2017 he decided on instinct, citing Ma’s “strong, shining eyes,” at a company with “no business plan and zero revenue.” The myth omits that this was no cold pitch: Son wanted to invest $40 million, Ma turned half down, and Alibaba already had Goldman’s money and a live site.
Then the crash came, and Ma’s filmed prediction — a painful price within three to five years — arrived on schedule. By 2001 the company had roughly $10 million in the bank and was burning $2 million a month: five months of runway. Savio Kwan, an ex-GE executive who joined as president and COO in January 2001, cut the burn to roughly $500,000 a month. US executives were let go, the US office shut, and the company retreated to Hangzhou under a slogan that repurposed an industry acronym: “B2C” — Back to China.
The retreat came with a revenue model — the least glamorous, most load-bearing fact in this story. Alibaba did not charge transaction fees. It sold memberships: China TrustPass (June 2001 international, March 2002 domestic) and Gold Supplier, an annual premium subscription for exporters. WTO accession turned exporters into a growth market exactly as Alibaba started charging them. In December 2002 the company reached positive cash flow; Ma has said the 2002 goal was to earn one dollar of profit — his telling, but the milestone is solid.
The structural fact: by early 2003 Alibaba had real, growing subscription revenue from businesses — and no consumer product at all. Everything that follows is financed by that fact.
Part IV: The Growth Machine
Taobao’s climb from zero to market leadership in three years decomposes into five machines, each dated and documented: a price wedge, a trust mechanism, a chat window, a community, and a marketing asymmetry. None was an accident.
A war started on purpose
Taobao was a preemptive strike, not a response to decline — the B2B business was healthy when Ma set the engineers loose in the apartment. The strategic model was explicit and contemporaneous: Son had beaten eBay in Japan, and Ma told Forbes in April 2005, “Son kicked Ebay out of Japan. I have the same chance in China.” The team worked in secrecy — even colleagues at headquarters didn’t know, by most accounts — registered the domain April 21, 2003, and shipped on May 10 into the teeth of the quarantine. The company later designated May 10 “Aliday.”
Free as a wedge, not a business
EachNet ran its American parent’s model: listing fees plus transaction fees. Taobao launched with neither. Then it made free a public commitment rather than a promotion — pledged and repeatedly extended, most consequentially in October 2005, when a freshly capitalized Taobao announced three more years of free. It was still policy eleven years later; the 2014 F-1 states, “The creation of storefronts and listings are free of charge to sellers,” with revenue coming instead from advertising and, later, Tmall commissions.
The mechanic is the asymmetry of pain. For Taobao, free cost the difference between no revenue and no revenue — B2B subscriptions paid the bills either way. For eBay, matching free meant burning the very model it had come to China to extend. A fee-dependent incumbent cannot cheaply follow a cross-subsidized attacker to zero. eBay held its fees for two and a half years; by the time it let go, the market had moved. Free also solved cold-start liquidity: an unlimited free shelf built selection, and selection pulled buyers.
Alipay: manufacturing trust
Free listings recruited sellers. What produced transactions was escrow. On October 18, 2003 — five months after launch — Taobao introduced an escrow payment feature: the buyer’s money is held by the platform and released only when the buyer confirms the goods arrived. In December 2004 this became a company; the F-1’s milestone language is unusually plain: “In 2004, we established Alipay to address the issue of trust between buyers and sellers online.”
In a market with few credit cards, no chargebacks, and weak consumer protection, escrow answered the who-sends-first problem structurally. A Knowledge@Wharton expert panel in 2007 noted consumers found eBay’s payment experience “nerve-wracking” next to Taobao’s escrow. This — not price — was the decisive product fact of the war; the proof arrives in January 2006, when eBay matched free and kept losing anyway.
Wangwang: haggling as a feature
In July 2004 Taobao shipped AliWangWang: instant messaging built into the marketplace — text and voice, buyer to seller, before purchase. In Chinese commerce, negotiation is not friction; it’s the trust ritual. Wangwang let buyers haggle and take the measure of a seller the way they would across a market stall.
eBay’s platform did close to the opposite, for a structural reason: a fee-charging marketplace must keep buyers and sellers from meeting off-platform, so contact was minimized by design. Taobao monetized nothing, so it could let people talk. The celebrated culture fit was downstream of the business model, not just superior anthropology.
The localization gap ran through the whole product, per the Forbes 2010 post-mortem and case literature: listings organized by shopper demographics rather than a rigid category tree; roughly 90 percent of listings fixed-price against eBay’s auction-centric model; fourteen-day auto-extending listings; period satisfaction surveys of 77 percent for Taobao against 62 for eBay EachNet.
Sellers as the sales force, forums as the channel
Taobao seeded its seller base the way Alibaba had seeded B2B: community. Staff-moderated message boards, sellers organized into groups, forum culture treated as product surface. Sellers taught each other, policed each other, and — because a free shelf pays sellers to bring their own demand — recruited buyers. The playbook was ported from the B2B site, which had reached 40,000 users the same way.
The marketing asymmetry
eBay spent like an incumbent: per Erisman, corroborated by the case literature, it signed exclusive advertising deals with Sina, Sohu, and NetEase — China’s three main portals — to lock Taobao out of mainstream online advertising. Taobao went around: thousands of smaller sites, BBS forums, and, critically, television, which reached the mobile-phone-owning consumers not yet on the portals. eBay bought the front doors of the Chinese internet; Taobao advertised to the people who hadn’t walked through them yet.
What the share numbers actually say
Every share figure from this war needs its source and basis attached: the series mix users, listings, and GMV estimates from different research houses; eBay disputed the unfavorable ones; user counts were inflated on both sides. The record, attributed:
| When | Basis (as reported) | eBay EachNet | Taobao | Source |
|---|---|---|---|---|
| Mid-2003 | Online-auction market, at eBay’s acquisition | ~70–85% | just launched | Case literature; Forbes |
| April 2005 | Online-auction sales; claimed registered users | 53%; 10M users | 41%; 4M users | Forbes, April 25, 2005 |
| Full-year 2005 | GMV estimates, local research houses | 36% (from 79% in ‘03) | 59% (from 8% in ‘03) | The Economist, Sept 21, 2006 |
| March 2006 | Users (Analysys International) | 29% | 67% | Forbes 2010 post-mortem |
| End 2006 | C2C share at eBay’s exit | ~20% | ~70%, 30M+ users | Knowledge@Wharton, Jan 2007 |
No single percentage above deserves full trust. Two things do: the direction — crossover in 2005, rout by 2006 — is unanimous; and the dollar anchor, via Wharton: 2005 transaction volume above $1.2 billion for Taobao against roughly $600 million for eBay China. Two years in, the free marketplace was doing double the incumbent’s volume.
Part V: The War as eBay Fought It
Buying the leader
eBay entered China by acquisition. Its FY2002 10-K records $30.0 million paid in March 2002 for a minority EachNet stake — the filing has been read as both 33 and 38 percent, so I lead with the solid number, the dollars. In mid-2003 it bought the rest: $144.9 million total, per the FY2003 10-K. With the further $100 million investment announced in 2005, eBay’s China commitment reached roughly $300 million.
The commitment was real at the top. In February 2005, Meg Whitman told analysts China was “a must-win” and “likely to be the defining measure of business success on the Net.” The pithier line history remembers — whoever wins China wins the world — appears only in insider and secondary retellings; no contemporaneous primary record exists, and it may be a paraphrase that hardened into a quote.
The migration
Then eBay committed the unforced error the war pivots on. In 2004 — publicized that September and October — it shut EachNet’s China-hosted platform and migrated Chinese users onto eBay’s US-based global infrastructure. Three layers of cost: page latency across the Pacific; the Great Firewall, under which suspicious-looking content could get a whole foreign-hosted server temporarily blocked; and lost local features and engineering autonomy, with changes queued through San Jose — Erisman describes waiting for a “train seat” in the global release schedule. Shao Yibo, EachNet’s founder, put the result in one sentence: “on the day of the move, traffic dropped by half.”
On top sat the leadership the post-mortems keep returning to: a German country manager and an American CTO, neither Chinese-speaking — “eBay’s biggest mistake,” per one Forbes retrospective. Decisions about a haggling, forum-driven, escrow-needing market were made a twelve-hour flight away, in a language it didn’t speak, on servers outside its firewall.
Sellers left for the free, local, escrow-backed alternative that loaded instantly. Taobao’s job in 2004–2005 was largely to be standing where they landed.
The crocodile
Ma’s doctrine was stated mid-fight, on the record. Forbes, April 25, 2005: “Ebay may be a shark in the ocean, but I am a crocodile in the Yangtze River. If we fight in the ocean, we lose — but if we fight in the river, we win.” The same interview carries the colder line the post-mortems would spend years confirming: “Taobao didn’t win the first battle, but Ebay lost it.”
The backstop
The war still had to be paid for. In February 2004, SoftBank led an $82 million round — then the largest private placement for a Chinese internet company — explicitly to fund Taobao. Per Erisman, eBay had begun making acquisition overtures toward Alibaba that same year; his account alone.
Then the decisive check. In August 2005, Yahoo paid $1.0 billion in cash and contributed its struggling Yahoo China business (plus $8 million in costs) for approximately 46 percent of Alibaba’s common stock — about 40 percent fully diluted, roughly 35 percent voting, per Yahoo’s own filings. Behind the deal sat a decade-old relationship: the founder Ma had once guided along the Great Wall was writing the largest strategic check in Chinese internet history, on terms that left Ma in control.
What the billion bought was time, publicly committed: the October 2005 pledge of three more years of free. It also took Alibaba off the table as an eBay acquisition.
The epitaph
eBay’s response is one of the few moments where a company wrote its own epitaph in a press statement. October 19, 2005: “‘Free’ is not a business model.” The extension, eBay added, “speaks volumes about the strength of eBay’s business in China.”
In January 2006, eBay eliminated seller listing fees in China. InfoWorld’s headline completed the thought: “EBay China decides ‘free’ is a business model.” It didn’t help — the most diagnostic fact of the war, because it isolates the variable: the advantage by then was escrow trust, local product, community, and liquidity, not price.
On December 20, 2006 — fourteen months after the “free” statement — eBay announced it was shutting eBay EachNet’s site and folding the business into a joint venture with Tom Online: eBay contributed EachNet plus $40 million for 49 percent; Tom Online put in $20 million for 51. Whitman flew to Shanghai for the announcement. A minority stake in someone else’s company is how a platform giant says it lost — and everyone read it that way.
Part VI: The Receipts
Monetization arrived on Alibaba’s own schedule: the Alimama advertising platform in 2007, Tmall — the brand-mall sibling that does charge commissions of 0.3 to 5 percent — in 2008, pay-for-performance search ads throughout. The model charges sellers for visibility, not for listing; by the December 2013 quarter the blended monetization rate across Alibaba’s China retail marketplaces was 3.05 percent of GMV. Free listings had been the top of the funnel all along.
One honest asterisk, because the governance concentration that won the war also produced the company’s ugliest episode. Between 2009 and 2010 — disclosed in May 2011 — Alipay’s ownership was transferred out of Alibaba Group into a company controlled by Jack Ma, framed as necessary for a payments license under June 2010 People’s Bank of China rules on foreign-invested payment companies; the F-1 calls it “our management’s response to regulations issued in June 2010 by the PBOC.” Yahoo said it had not been informed. Analysts noted an asset argued to be worth over $1 billion had moved for around $50 million. The July 2011 settlement gave Alibaba Group 37.5 percent of Alipay’s equity value at a future liquidity event — $2 billion floor, $6 billion cap. Whatever weight you give the regulatory rationale, it is the permanent exhibit on what founder control means when interests diverge.
The end-state numbers are from the F-1, not estimates. For the twelve months ended June 30, 2014, Alibaba’s China retail marketplaces processed RMB 1,833 billion — US$296 billion — in GMV (fiscal 2014, ended March 31, was $270 billion; sources routinely conflate the two), from 279 million annual active buyers and 8.5 million active sellers. Fiscal 2014 revenue was US$8.46 billion; net income US$3.77 billion. Alibaba’s marketplaces accounted for 54 percent of China’s 11.3 billion parcel deliveries. Pre-IPO ownership: SoftBank 34.1 percent, Yahoo 22.4 percent, Jack Ma 8.8 percent, Joe Tsai 3.6 percent.
On September 19, 2014, Alibaba listed on the NYSE at $68 per ADS and opened around $92.70. After the greenshoe on September 22, the offering totaled roughly $25 billion — the largest IPO in history at the time, past Agricultural Bank of China’s $22.1 billion — at a market value of about $231 billion. Son’s $20 million, held through everything, was worth on the order of $79 billion at the IPO valuation — arithmetic on the F-1 stake, not a reported figure.
The filmed 1999 target had been an IPO in 2002; what 2002 delivered instead was positive cash flow. The IPO took twelve more years and set a world record.
Thinking Tools Before the Verdict
As in every piece in this series, frameworks divide into thinking tools and a judging tool: these organize the story; the scorecard does the scoring.
Porter’s Five Forces, run from eBay’s chair in 2003, said the industry was won: dominant share, network effects, the strongest brand and balance sheet in e-commerce. Run from Hangzhou, the position is thinner. Buyer and seller power was enormous — no contracts, no lock-in, and sellers could multihome the moment a rival was free, so the network effect was rented, not owned. Entry was gated by liquidity, but a subsidized attacker’s entry cost was server bills. The substitute — the offline bazaar — had trained expectations (haggling, inspect-before-you-pay) the incumbent’s imported product fought. Rivalry was asymmetric in the deciding way: fee-dependent incumbent against cross-subsidized attacker. Porter wouldn’t have named the winner in 2003; it would have flagged that eBay’s fees and platform quality were the whole position — and eBay sacrificed the second defending the first.
Rogers’ Diffusion of Innovations explains the adoption speed. Relative advantage: free plus escrow, against fees plus a payment experience Wharton’s panel called “nerve-wracking.” Compatibility: haggling and fixed prices matched how China already shopped; auction mechanics and no-contact design did not. Complexity: local language, local navigation. Trialability: free — trying Taobao risked nothing. Observability: television, forums, and seller communities made adoption visible to precisely the consumers not yet reading portals. Taobao sits near the ceiling on all five factors; the diffusion profile alone predicts the crossover.
VRIO disciplines the moat question. Free is a price, not a resource — imitable by definition, and eBay imitated it in January 2006. What was valuable, rare, costly to imitate, and organized by 2006 was the stack underneath: escrow trust at scale, two-sided liquidity, community, a Hangzhou product organization shipping for this market. The January 2006 fee cut is the clean experiment: when the copyable part was copied and the outcome didn’t change, what remained was, by construction, the moat.
The Evidence Scorecard: Would You Have Funded Alibaba?
This is the series’ judging instrument — the gated, weighted scorecard described on the series hub, scored from evidence available at the time. Snapshot one: May 2003, the week Taobao launched — a zero-revenue product against an incumbent holding 70-plus percent, inside a company earning only B2B subscriptions. Snapshot two: December 2006, the month eBay exited.
The gates, at May 2003
Pain. B2B: unambiguous — exporters had been paying real subscription money for two years. C2C: a flag — consumers in early 2003 were not clamoring to shop online; the trust gap suppressed the very behavior that would have expressed the pain. Real but latent, activated partly by SARS itself. Pass, the C2C half graded on faith plus one epidemic.
Buyer. B2B: nameable and paying — Gold Supplier exporters. Taobao: deliberately nobody. A standalone Taobao pitch fails this gate outright; it clears only as what it was — a cross-subsidized land grab funded by the paying business next door. That distinction drives the whole card.
Market. Venture-scale beyond argument: tens of millions online and compounding, WTO at the exporters’ backs, the largest consumer market on earth behind the trust unlock. Pass.
Behavior change. The heaviest gate in the story: asking consumers with no credit cards and no buyer protection to mail money to strangers. Escrow and the chat window were aimed directly at it. Pass, flagged as the top risk.
The weighted card
| Category | Weight | May 2003 | Dec 2006 |
|---|---|---|---|
| Product-market fit | 30 | 17 | 26 |
| Distribution | 25 | 12 | 22 |
| Unit economics | 20 | 8 | 10 |
| Market quality | 10 | 9 | 10 |
| Team / founder-market fit | 10 | 8 | 9 |
| Moat / defensibility | 5 | 1 | 4 |
| Total | 100 | 55 | 81 |
Product-market fit (17 → 26). In May 2003 the behavioral evidence was all B2B: paying subscribers, renewals, cash-flow positive since December. By December 2006, Taobao had 30-plus million registered users, majority share on every published measure, and the strongest evidence a marketplace can produce: buyers and sellers who stayed after the incumbent matched the price. The honest wrinkle keeps 2006 off the ceiling — willingness to pay on Taobao itself was still untested, because nothing on Taobao cost anything.
Distribution (12 → 22). At launch, locked out of mainstream online advertising by eBay’s portal exclusives, betting on unproven channels. By 2006 the TV-plus-forums-plus-community machine was demonstrably repeatable — a market won under an ad blockade, sellers themselves the acquisition channel.
Unit economics (8 → 10). The card refuses to be romantic here. In 2003 Taobao was designed to earn nothing; in December 2006 — at the moment of total victory — it was still earning nothing, on purpose, monetization an intention rather than a demonstrated fact. The B2B engine keeps the score off the floor at both snapshots; the winning product itself had never charged anyone.
Market quality (9 → 10). The “why now” stack — internet adoption, WTO, SARS — was about as strong as the category allows.
Team / founder-market fit (8 → 9, capped). Scored on documents and behavior, not aura: the burn cut, the subscription pivot to cash-flow positive mid-bust, a secret product shipped from an apartment during a quarantine, a free pledge extended precisely when the capital landed. The cap exists because hindsight halo is strongest on exactly this company — the card weights what Ma did, not what he says he said.
Moat (1 → 4). Nothing in 2003. By 2006: two-sided liquidity, escrow trust, community — validated by the only test that matters, an incumbent that matched the price, spent $300 million, and went home.
Reading the card
Fifty-five is the top edge of “interesting but unproven” — the correct grade for May 2003, and the reason most rational investors would have passed. Eighty-one is “promising with one major risk,” and the risk has a name: a victorious product with deliberately unproven economics, inside governance concentrated enough that five years later an Alipay-sized asset could leave the company before the second-largest shareholder heard about it.
The non-compensatory rule, applied. Standalone Taobao in May 2003 fails fatally: no buyer, no revenue design, no moat, incumbent at 70-plus percent. No team or market score compensates — the rule exists to say so. What rescued the bet was not a category score but the structure around it: the fatal weakness was underwritten in-house by Gold Supplier cash flow. As a standalone Series A pitch, then, Taobao-2003 is a pass — a trap by the bands, unless the evidence was about to change. As the bet the actual investors faced — SoftBank in February 2004, Yahoo in August 2005, crossover already visible — it was a trajectory purchase: momentum and team, swallowing the unit-economics blank. Both swallowed it; one turned $20 million into the best venture outcome ever recorded.
Kill criteria, stated as of May 2003
- If escrow (or something like it) fails to make strangers transact, Taobao is a free classifieds board. → Fired and was answered: escrow shipped October 18, 2003; transactions followed.
- If B2B subscription growth stalls, the free strategy is unfundable. → Never fired; WTO-era export demand kept compounding.
- If eBay matches free in 2003–04, while Taobao is subscale, and keeps its platform local and fast, the wedge disappears. → The scariest one; eBay did the opposite twice — held fees until January 2006, moved its platform offshore in 2004.
- If no one refinances a zero-revenue war, the pledge breaks before the incumbent does. → Answered twice: $82 million in February 2004, $1 billion in August 2005.
- If a distressed or heavy-handed shareholder forces liquidity or a sale, the war ends by ownership, not market share. → Half-fired: Goldman, holding 1999-vintage control terms, sold everything in early 2004 — luckily to no strategic effect; SoftBank, post-crash, held.
The kill list reads as the table of contents of the war: trust, fuel, the incumbent’s response, capital, control.
Hidden Forces
The cross-subsidy was the business model. The most load-bearing structure in the story is the least discussed: a boring, profitable subscription business paying for a spectacular, free consumer war. Strip out the B2B engine and the entire Taobao strategy is a bluff eBay could have called by waiting.
The Great Firewall was a silent combatant. When eBay moved Chinese users onto US-hosted infrastructure, the state’s internet architecture became Taobao’s ally: trans-Pacific latency as a permanent tax, collateral server blocking as random outages. Nobody at Alibaba lifted a finger. The same state’s 2010 payment rules later detonated the Alipay crisis — the house takes a cut both ways.
A tour guide’s relationship became a billion-dollar term sheet. The Yahoo deal — that size, on those management-friendly terms — was available from exactly one company on earth: the one whose co-founder Ma had walked along the Great Wall years before Alibaba needed a war chest.
eBay’s fee model designed its product defeat. A transaction-fee marketplace must keep buyers and sellers from meeting off-platform, so eBay minimized contact. A zero-fee marketplace has nothing to protect, so Taobao built a chat window and let China haggle. The culture fit was a degree of freedom purchased by the revenue model — a more transferable lesson than “localize better.”
Founder control cut both ways. Concentrated governance made decade-scale patience and unmonetized victory possible — no earnings call ever forced Taobao to charge early. The same concentration produced 2011. The war-winning trait and the governance scandal are one trait.
The Luck Audit
Specific breaks, and the skill that made each exploitable — no “right place, right time.”
Lucky: WTO accession, December 11, 2001. Alibaba had zero agency over China’s trade integration, and the export boom was the rising tide under the entire subscription business. The skill: TrustPass and Gold Supplier were already on the shelf — built during the retreat, months before the wave — so the tide lifted a boat that existed.
Lucky: SARS accelerated e-commerce adoption in the exact month Taobao launched. Duncan Clark writes that SARS “came to represent the turning point when the Internet emerged as a truly mass medium in China” — and Taobao shipped into precisely that month by coincidence; the skunkworks predated the epidemic’s arrival in Hangzhou. The skill: the company ran through a thirteen-day quarantine without dropping the B2B business — record lead volume on day one of working from home — and shipped anyway. Said plainly, too: Alibaba manufactured its own exposure by sending an employee to the Canton Fair mid-outbreak, and was lucky the human cost stopped where it did.
Lucky: eBay’s unforced errors. The migration, the non-Chinese-speaking leadership, the portal exclusives aimed at the wrong audience, the fee stubbornness until January 2006 — Alibaba controlled none of it. A competent, locally autonomous eBay China is the strongest “Taobao doesn’t win by 2006” counterfactual. The skill: a localized, escrow-backed, free product standing exactly where every defecting seller would land, and a stated doctrine of fighting only on that terrain.
Lucky: Son’s $20 million survived Son’s own catastrophe. SoftBank lost most of its market value in the crash — retrospectives say 97 to 99 percent, with something like $70 billion of Son’s paper wealth erased — yet the Alibaba stake was never sold. Alibaba’s independence depended on its lead investor not needing liquidity at the worst moment. That was Son’s conviction; it was Alibaba’s luck.
Lucky: Yahoo’s singular motivation in August 2005. A buyer with a failing China unit, a founder-level relationship, and strategic fear of rivals — willing to write $1 billion on terms that left management in control — existed exactly once, exactly then. The skill: Ma had kept the Yang relationship warm for eight years, then used the money not to diversify but to double the bet publicly within two months — the pledge that broke the incumbent in fourteen months.
What This Actually Means
I’ll resist the founder-lessons format, as ever. But four patterns survive the evidence.
Terrain selection is a strategy, not a slogan. The crocodile doctrine was stated to Forbes in April 2005, mid-war — not retrofitted by a winner. Everything documented about Taobao’s design follows from it: fight where local knowledge, local hosting, local habits, and a subsidy the incumbent can’t match convert smallness into advantage — then still be standing, on home terrain, when the incumbent defeats itself.
Free is a tactic with a fuel requirement. eBay’s “‘Free’ is not a business model” became history’s punchline, but read carefully, the statement was true. Free was not Taobao’s business model; Gold Supplier subscriptions were, and later advertising was. The transferable pattern is the pairing — a paying business underwriting a free wedge into an adjacent market — never the free wedge alone.
In low-trust markets, trust infrastructure precedes marketplace growth. The cleanest natural experiment in the story is January 2006: eBay matched the price and kept losing, which means price was never the binding constraint. Escrow was. Alipay’s own filing language — established “to address the issue of trust between buyers and sellers online” — is the rare corporate sentence that is also the analytical conclusion.
The record beats the legend, in both directions. Ma’s self-mythology — KFC, Harvard, six minutes — is unverifiable stagecraft, labeled as such throughout. But the documented record outclasses the myths: a filmed 1999 speech predicting the crash’s pain and naming Silicon Valley as the competition; a strategy stated on the record mid-war, then executed; a product shipped to the day under quarantine. Discounting the legend doesn’t shrink the founder — the verifiable version was already extraordinary, and the mythmaking was a growth channel of its own, aimed at the one marketplace this piece can’t score: attention.
Hold both books at once. A company that failed a rational scorecard in May 2003 won everything by December 2006 because specific, nameable things changed — escrow shipped, capital arrived twice, and the incumbent spent three years making every error available to it. The scorecard names what changed. The luck audit names what nobody controlled. The honest account is that Alibaba needed both columns full, and got them.
Sources and Notes
Primary documents:
- Alibaba Group Holding Ltd., Form F-1/A (Amendment No. 6), SEC, September 5, 2014 — all 2014 scale metrics, pre-IPO ownership, corporate-history milestones, the June 28, 1999 legal founding, the “free of charge” listings language, the Alipay-divestment framing.
- eBay Inc., Form 10-K FY2002 (initial EachNet stake, $30.0 million) and FY2003 (completion, $144.9 million total).
- Yahoo Inc., Form 8-K (August 2005) and Form 10-K FY2005 (the 2005 deal terms); Form 8-K (July 2011) (the Alipay settlement terms).
- Tom Online SEC filings (2006 joint-venture structure).
Insider accounts: Porter Erisman, Alibaba’s World (2015) and Crocodile in the Yangtze (2012), which contains the filmed February 1999 founding speech; his eBay-war account via EcommerceBytes (June 2015). Duncan Clark, Alibaba: The House That Jack Ma Built (2016). Savio Kwan’s first-person crisis accounts (GGV podcast; YourStory).
Journalism and retrospectives: Forbes, “Standing Up To a Giant,” April 25, 2005 (the Ma quotes and April 2005 figures). InfoWorld (the October 2005 eBay statement; the January 2006 fee elimination). The Economist, September 21, 2006 (share series). Knowledge@Wharton, January 2007 (JV terms, total spend, end-2006 figures). Tech Monitor, February 2004 ($82 million round). The Globe and Mail, 2014 (Goldman, via Shirley Lin). TechNode, February 2020 (translated first-person SARS-quarantine accounts). Forbes/China Tracker, September 2010 (the eBay post-mortem). Fortune and Bloomberg, September 22, 2014 (IPO record).
Disputed or single-source details:
- eBay’s initial EachNet stake: the FY2002 10-K has been read as both 33 and 38 percent; the $30.0 million is unambiguous, so the text leads with dollars.
- The Goldman 1999 round: $5 million bought 50 percent; Goldman’s own participation was ~$3.3 million for ~33 percent. Retellings blur the two.
- SoftBank’s 2000 stake: reported between 30 and 34 percent; the 34.1 percent pre-IPO figure includes later investments.
- The “six minutes” Son meeting: self-reported by both principals; some retellings say five. The investment is fact; the stopwatch is legend.
- Taobao skunkworks headcount: six (TechNode’s first-person account) versus seven (Erisman-derived retellings).
- 2005–2006 market shares differ by basis (users, listings, GMV) and research house; eBay disputed the unfavorable series. Direction unanimous; each table figure carries its source and basis.
- Whitman’s “whoever wins China wins the world”: insider retellings only; her verifiable February 2005 phrasing was “a must-win.”
- Ma’s self-reported biography — the KFC rejection (24/23), ten Harvard rejections, ~30 job rejections, the “one dollar of profit” goal, the Seattle “beer” search, the 27-kilometer cycling — has no independent documentation. Presented as his tellings.
- Erisman-only: eBay’s acquisition overtures toward Alibaba from 2004.
- Clark-only: Joe Tsai’s compensation figures at joining.
- The Ma–Yang Great Wall meeting is usually dated 1997; sources vary 1997–98.
- Son’s dot-com losses (~97–99 percent of SoftBank’s value; ~$70 billion of paper wealth) are retrospective approximations. The ~$79 billion IPO-day stake value is arithmetic, not a reported number.
Analytical frameworks: the evidence scorecard is the series’ standard instrument, described in full on the series hub. Porter’s Five Forces and Jay Barney’s VRIO (Journal of Management, 1991) are thinking tools; Rogers’ Diffusion of Innovations is the adoption lens. Claims resting on inference are flagged in the text with “reportedly,” “in his telling,” or equivalent.