Jeff Bezos, Jack Ma, and the Quest to Kill eBay

Twenty years on, Steve tells the story of the first time Jeff Bezos was wrong. The setting is the old Columbia Building on 3rd Avenue in late 1998 — Amazon at maybe a hundred people, Bezos still a cult figure, Rufus the dog hitting the launch button at midnight. A hushed project starts: a forked codebase (Obidos copied to Varzea, named for an Amazon-river swamp), a secret team, a mascot named Mr. Tooth, and a plan to kill eBay by launching an auction site that would peel the sellers off in a single weekend. Sellers decorate their Amazon storefronts with links back to eBay where the buyers are; buyers come, find nothing, leave. Mr. Tooth is quietly retired. The lesson the failure forces: you can't beat a network with a near-identical clone — the gravity is too strong. zShops tries again and fails for the same reason. What finally works is Single Detail Page: third-party "new and used" listings grafted onto Amazon's existing retail network, growing incrementally because they borrow value from a network Amazon already owns. Three rules drop out: secret forks are stupid, don't clone a network, don't give up on a good idea — plus the signal Steve still uses, that "the leading signal that you've finally got your innovation right is that it shows continuous incremental growth without much effort." The second half is Jack Ma: in 2004 eBay has 85% of the Chinese market; by 2007 they've left, beaten by TaoBao's instant-messaging chat between buyers and sellers — the one feature eBay structurally couldn't copy without killing its own transaction fees. TaoBao starts as an eBay clone and ends as the Chinese Craigslist, then makes its money a few innovations later in Tmall.

You can't fight a network head-on, but you can flank it through a related market — and you may not know which related market until you've tried and failed two or three times. The Single Detail Page anecdote still gets quoted in startup decks more than a decade later.