This is the story of the first time Jeff Bezos got something spectacularly wrong. It’s the story of why he was wrong, what happened afterward, and what we can learn from it. It’s also the story of how Jack Ma tackled exactly the same problem a few years later, and got it right.

For this particular story, I don’t have all the details. I may get a few things slightly wrong. I wouldn’t treat this post as a historical record so much as a way to get some life lessons from guys like Jack and Jeff—without spending as much tuition money as they did.


The Birth of Mr. Tooth

I’m going to venture a guess that you’ve never heard of Mr. Tooth. He looks like this:

It’s pretty goddamn hard to find a picture of Mr. Tooth these days. A Google search for “amazon auctions mr tooth” will produce nothing. I had to use the Wayback Machine, and even that took quite some time because his life was so short.

Mr. Tooth wasn’t purged from the internet through some vengeful move by Amazon to hide all traces of their embarrassment. No, the simple truth is that Mr. Tooth was so ineffectual that nobody ever bothered to remember him.

When I started at Amazon back in December 1998, it was a crazy time. A historic time. There was an electricity in the air, and it wasn’t just the smoke from the bar below us, there in the old Columbia building on 3rd Avenue. Everyone could feel history being made, and at the center of the fire was Jeff himself. Even back then, in a company of maybe 100 people (sans ops), he was already a cult figure with protected access.

Start looking for another related market to enter, and you can then sneak up on the original market from the side.

At the time, Jeff had never been wrong about anything. Certainly not in any big, meaningful way. Everything he launched was pure gold. First he launched Books, and who the hell would have ever guessed that an internet bookstore would catch fire like that? Then, he successfully launched Music, taking aim at record stores, and then DVD/Video, taking aim at Blockbuster.

Back in 1998–1999, every single launch of a new tab (remember tabs?) was a huge event. Eric Benson’s dog, Rufus, always got to press the launch button, and we always launched at midnight so there would be a big press release in the morning (and also in case there were hiccups with the site), and it was just nuts how everything Jeff did looked like it was going to topple an established industry.

We launched the Amazon Japan site shortly after BMVD (books, music, video, DVD), and it was also a huge success, which was noteworthy because gosh, it was Japan. It was the first time we had to deal with localization, different regulations, different supply chain, different consumer patterns, and so on. Still a big success. Because Jeff was Always Right™.


Sometime in that first month or two, I heard about a secret project, discussed in hushed tones. And it really was a secret project, at least insofar as such a thing could be accomplished in a tiny office of 100 people on two floors.

The project was so secret that they had forked the code base for the retail site. The site’s web server was called Obidos, named for the town near the fastest part of the Amazon river, and leadership had forked (copied) it and fittingly named it Varzea, after a big swamp. They’d cherry-picked some of the top engineers to work on Varzea, also in total secrecy.

I’ll pause here and get Life Lesson #1 out of the way, which is: Secret projects are stupid. Forking your code base is also stupid.

The whispered secret project? Amazon was gonna kill eBay. The plan was to build an entire auctions site from scratch, fill it with some sweet seed content like rare collector baseball cards from Sotheby’s, offer an ironclad guarantee (“We’ll refund you if our sellers screw up”), and then pounce. Boom, bam, the tab launches, the eBay sellers all pack up their shops and leave eBay en masse, all the eBay customers stampede over as well, and eBay is left sitting there wondering what hit them.

Great plan, eh? Well, of course we can laugh in hindsight, especially after a few drinks. Honestly, even at the time it should have been obvious that it would never work, because it’s not as if eBay was the world’s first network effect.

But Jeff had never been wrong before. And nobody had truly believed that his little bookstore was going to be so big, until it grew like a Cat-5 storm and proved everyone wrong. In early 1995, everyone from my tiny graduating CS class at the University of Washington was looking for jobs at big tech companies like DEC, IBM, and upstart Microsoft. One of our classmates told us she was going to work for this internet bookstore, and I remember feeling so sad for her. We felt like she had decided to throw away all her fancy training and become a librarian or a nun. She was employee #11.

By 1998, post-IPO, Jeff had cleared up any doubts as to whether his bookstore would succeed. Every time he tried something crazy-ambitious, he wound up being right. So when Jeff said they were going to kill eBay (this was before the U.S. government sued Microsoft and everyone learned not to say you’re killing your competitors), people believed him. They just straight up had faith. Amazon had a cult-like culture at the time, a speak-no-dissent mentality, where discussion of failure simply wasn’t allowed—perhaps because people superstitiously believed saying it might make it come true. Google+ was like that, too.

I don’t know why Amazon Auctions had a mascot named Mr. Tooth. My hunch is that Jeff came up with it himself. There’s really nothing wrong with Mr. Tooth as a mascot. He wasn’t a bad idea. The bad idea was Amazon Auctions — because you can’t beat a network effect with a nearly identical network.

Amazon Auctions wasn’t fundamentally different from eBay. Sure, it looked a little cleaner, which is unsurprising given that eBay’s site always looked like someone threw up all over it. And Amazon had established enough brand trust that the refund guarantee was legit. I have no doubt everyone believed it.

But Amazon Auctions was a direct competitor to eBay, which meant that in order for it to be successful, people were going to have to leave eBay.

And that, dear friends, did not happen. Amazon Auctions was a spectacular flop. Put in modern terms, it was about as successful as Google+. And for the same exact reason! The problem is that certain ecosystems exhibit what’s called a “network effect,” in which the system naturally reinforces itself through a feedback loop, causing an almost magnetic attraction back into the system.

Even the best marketing isn’t strong enough to beat an entrenched network effect.

In eBay’s network, the buyers go where the sellers are (for variety), and the sellers go where the buyers are (for reach), and it becomes self-reinforcing. Buyers are busy people, so they would visit Amazon’s Auctions site, not find what they wanted, and head straight back to eBay. Sellers have more time on their hands, and some of them tried Amazon Auctions. But they weren’t getting bites, so they started decorating their sites to point back to eBay, where all the buyers were.

In practice, the only way to get everyone moved over to an identical network would be to do it all at once, and of course that’s logistically impossible.

So, Life Lesson #2 (and this one is pretty goddamn important) is: Don’t try to beat a network by making a clone with improvements. It ain’t gonna work. There is too much gravitational inertia in the original network; nobody is incentivized to leave it.

There are some approaches that will work, and we will explore two of them: Jeff’s and Jack’s. Both require you to get into an entirely different market and build a network there. You can’t beat a strong network head on, but you can flank it.


Amazon Launches zShops

If you’re trying to choose a direction for your business and you’re up against a network effect, you should consider that market to be locked up. Instead, start looking for another related market to enter. Then, you can then sneak up on the original market from the side. And there is always a related market.

As a self-serving but relevant segue, I’ll mention that in 2011, when I wrote my Platforms Rant, it was actually Part 2 of an 11-part series I’d planned, every part being a different way to crap all over Google+ because it was a bad idea in at least 10 unique dimensions. (The 11th post was going to be the wrap-up where I begged them to just kill the damn thing.)

The first post in that series, which I managed to keep internal because I used corp Blogger, was about Real Names being a huge obstacle to adoption. Platforms was second, and then I had one planned about comment threads: how reddit was able to have 80,000 comments in a thread that grandmas were reading, whereas Google engineers would start complaining when a Google+ thread reached 50 comments. And so on. Eight more rants lost to the ages, because I got a little too drunk and couldn’t figure out how to use G+ and somehow posted to my external account.

Basically, I had been gearing up to argue that Google needed to go into a completely separate market to be able to compete with Facebook. Reddit was looking like a nice, ripe market because they were small. Plus, they were content-first rather than people-first, which meant they were fundamentally different from Facebook. And reddit never really figured out how to turn themselves into a social network. I wanted Google to do a social version of reddit.

That approach wouldn’t work today; reddit’s network of content creators and readers has grown too large. Would it have worked back in 2011? Who knows? But it sure as hell would have had a better chance than Google+ did. G+ was directly competing with Facebook’s massive network, whereas reddit was a related market that Google could have entered. Reddit was clearly doing something right, but they were also small enough that someone of Google’s size might have competed with them successfully.


Going after related markets was Bezos’ next step, after it had become clear that Amazon Auctions was going nowhere. He called the first sequel “Amazon zShops.” I’m not sure what the zShops meant. My hunch is that Jeff likes to equate Amazon with “A through Z” (meaning “we have everything”), hence the orange A→Z arrow in their logo. So the Z in zShops may have referred to the product being the long tail of pocket lint that winds up being sold on consumer-to-consumer (C2C) marketplaces.

Jeff puts a lot of thought into his product designs and marketing. That stuff matters. Unfortunately, even the best marketing isn’t strong enough to beat an entrenched network effect. Little Mr. Tooth had gone up against Mr. 800-pound Gorilla, lost the battle, and was soon to be yanked out.

Here is what Amazon’s tabs looked like after zShops launched, circa 2001:

Note that Auctions is still there, so it hasn’t been killed off yet. And there are a bunch of other tabs, stuff they launched while they were iterating on the auction problem. (Side note: Tabs are always a dumb idea. You hear that, web browsers and IDEs? Tabs don’t scale! Lists are how you scale user interfaces. M-x list-buffers!)

In any case, what exactly was zShops? How was it different from Auctions? Well, I’m going off 20-year-old memories here, but they got rid of bidding and developed a way for sellers to create storefront pages, with the idea being that sellers would be like miniature Amazon tabs.

Not a bad idea, right? Well, it actually was a bad idea for the same reason Amazon Auctions failed. It was a bad idea because eBay already had storefronts by then — or, at worst, they could just add it as a feature. In contrast, zShops had no defensible differentiators from eBay.

But before we continue the saga, let’s stop and get Life Lesson #3 out of the way: Don’t give up on a good idea. Despite the swift and embarrassing failure of Amazon Auctions, Jeff Bezos never gave up. This is probably the most important thing I ever learned from him, and I learned a lot from him.

It didn’t matter that zShops was doomed, because at a higher level, Bezos was finally innovating. Amazon Auctions? Definitely not innovation. Amazon Auctions was smug taillight chasing, a strategy that remains popular at many companies today. But zShops was the beginnings of Bezos trying to do something unique. And although it wasn’t unique enough, it was an important step in the evolution of the idea that eBay hadn’t gotten it 100 percent right for everyone, and that surely there must be some part of the C2C market that was being underserved. Through exploratory intervention, Jeff set out to find it.

Contrast this approach with Google’s, which went something like this: Try something quickly, see if it succeeds, and if not, abandon it forever and move on. This strategy started with Eric Schmidt and “let 1,000 flowers bloom,” a philosophy that Google ran with during Eric’s tenure as CEO. The core idea was that you need to generate luck by doing a whole bunch of stuff, and eventually some of it will get lucky. A key problem with Eric’s approach, however, is that it needed more scale than even Google could generate. In order to generate luck, you don’t need a thousand ideas; you need a million.

Eric used to tell us that he was the worst person to come up with ideas, because being an old rich white guy (note: I’m paraphrasing), he was completely unrepresentative of what trends might take off with young people. Fair enough! But another key problem with Eric’s approach to innovation is that Google engineers weren’t representative, either. It turns out that coming up with a brand-new billion-dollar product idea is pretty hard, and it’s easier to get there by acquiring successful startups.

Jeff Bezos and Steve Jobs had something in common—something different from Eric. They were confident enough to believe that if they had a great idea and they got it wrong the first time, the idea was still good. Their implementation might have been wrong, but they kept hammering at it, and changing its shape, until it was right.

Just be aware that you should expect to wind up in a completely different market by the time you’ve found a successful implementation. To beat a network effect, you need a different product in a related market that hasn’t been locked up yet.


Bezos’ C2C Triumph

I’ve buried the lede pretty deep at this point, but it’s time for the big reveal: Jeff eventually succeeded wildly in C2C. It just didn’t look anything like Amazon Auctions.

The leading signal that you’ve finally got your innovation right is that it shows continuous incremental growth without much effort.

My memory is hazy, but there may have been another round of innovation after zShops before Jeff hit on something that worked. That eventual thing was Single Detail Page, which is when you look at an item on Amazon’s retail site and it says “new and used,” giving you the option of buying from a third-party seller. Heck, it’s not even C2C anymore, although that line blurs even in eBay, with big sellers being real businesses in their own right, making them more B2C than C2C.

zShops had failed because it was still trying to compete with eBay, inasmuch as they had sellers, storefronts, transaction fees, and the same basic customer experience as eBay. So zShops sellers would just decorate their pages with a bunch of eBay logos and links to their storefronts over on eBay. This overall similarity caused no end of customer confusion. Many customers thought they were buying from eBay, and that eBay and Amazon were the same company. This was a huge struggle for Amazon’s customer service.

But Single Detail Page (SDP) was another beast entirely. SDP wasn’t going after eBay customers; it was going after Amazon customers. This was an ideal market: a large, captive (or at least receptive) audience of people who have expressed their purchasing intent by navigating to exactly what they want to buy. This was a new audience for C2C sellers, and there was no way for eBay to copy Amazon’s idea. SDP was a defensible innovation, sometimes called a moat, which leveraged Amazon’s own network effect in the retail-goods related market.

It’s important to note that SDP was able to grow incrementally. This had not been possible for Auctions or zShops, because they needed everyone (buyers and sellers) to move over together in order to generate enough value to keep customers around. SDP had borrowed value from Amazon’s retail network, so it saw strong, steady growth. The leading signal that you’ve finally got your innovation right is that it shows continuous incremental growth without much effort. This seems obvious, but big companies always seem to want to make up for innovation deficiencies through marketing and advertising before they’re ready to try something else.

Did Jeff win? Well, eBay still does a lot of business, and they seem to have much greater long-tail variety than Amazon. So Jeff hasn’t beaten eBay yet. But he was able to invent a new market for third-party sellers on SDP, giving him a foothold in the original market (eBay), one that he has gradually been growing into a stronger position.

And that, folks, is the only proven way to tackle a network effect. Make a new market and win in that one. If you’re lucky, your network might even be able to muscle in on the other one. But before you can start that fight, you first need your own fully entrenched network.


Jack Ma Does It Differently

Alibaba’s story is interesting, too, because a few years later, Jack Ma did beat eBay at their own game. I wasn’t there for it, so I don’t have colorful anecdotes—just the basic facts. Also, unlike Mr. Tooth, Alibaba vs. eBay is a story that has been told plenty of times. Search for “eBay in China” to find some great write-ups of how eBay lost there.

In 2004, eBay had an 85 percent market share in China, which amounted to an impregnable network effect. Alibaba was worried that eBay would impinge on their B2B business, and started (what appeared to be) a direct eBay competitor called TaoBao. Within two or three years they had pretty much driven eBay out of China, and in 2007 eBay finally threw in the towel and closed their doors in the country.

At first, Jack tried a bunch of innovations to try to differentiate his company from eBay, just as Jeff had done. eBay hadn’t tailored their offering well for the Chinese consumer, so TaoBao felt more familiar to Chinese buyers. All of Jack’s innovations were clever, and all would become helpful once the momentum started to shift away from eBay. But none of his little improvements over eBay would have been effective at starting a momentum shift because of our Golden Rule: You can’t beat an existing network with a similar offering.

The core problem is that network-effect leaders keep a keen eye on their competition, so it’s easy for them to copy innovations, which in turn helps protect their network. Facebook immediately responded to Google+’s big so-called innovation (Circles) by adding more or less equivalent functionality and, boom, Google+ was no longer different.

Of course, copying your competitors is normally a bad idea because it has the side effect of removing differentiation, forcing everyone to compete on price. You can play that game, but it’s no fun. However, if you have a lock on your market through a network effect, copying would-be attackers is just strengthening your incumbency.

But it turns out that network effects have a weak spot in their armor: differentiators that they can’t (or won’t) copy.

Jack Ma exploited that weak spot with a feature that eBay really didn’t want to introduce: instant messaging chat between buyers and sellers. eBay’s entire business model hinges on keeping buyers and sellers as separate as possible so they don’t collude to conduct their transaction in a separate channel and bypass the transaction fees, which is how eBay makes money.

The consensus seems to be that TaoBao’s chat feature killed eBay in China. Price-sensitive buyers and sellers flocked to TaoBao to avoid transaction fees. This gouged eBay’s network, but victory was pyrrhic: Alibaba couldn’t actually make any money off TaoBao. Put simply, they had started as eBay and finished as Craigslist, which is a completely different market — a related market!

The similarity to Craigslist didn’t end with removing transaction fees. Instant messaging also allowed haggling between the buyer and seller, which is another hallmark of Craigslist transactions. The only difference is that on Craigslist, the haggling happens in person. TaoBao had found success through the creation of a fully online equivalent of Craigslist, which (at least in China) didn’t exist at the time.

So TaoBao started as an eBay competitor and wound up in different market. They found that consumer rules were different in this new market. When TaoBao tried to start monetizing — initially by introducing sponsored placements via keyword bidding — their customers freaked out, and they had to back away from it. TaoBao had beaten eBay, but had they really won? In a way, yes. They had prevented eBay from hurting their B2B business, which was Jack Ma’s original goal.

The story ends with Alibaba making money a few innovation rounds later—not with TaoBao, but in a different, related market called The Mall, or (nowadays) Tmall, which was a market for more upscale brands. Ironically, this is what Amazon’s zShops aspired to be. Without eBay as a competitor, Alibaba had an easier time getting sellers to sign up. But TaoBao’s continued dominance in the Chinese auctions space has required TaoBao to forgo monetization because they won the space by training their customers that it should be free.

The lesson here is the same one we learned from Jeff Bezos: To beat a network effect, you may start at point A, but you’re going to be successful only by steering to some distant point B, which will always be a related market. And you may not know exactly where point B is until you’ve gone through several rounds of innovation. So if you’re sure you want to be in a market space, don’t give up after your first failure.


Networks Can Still Be Beaten

The overall takeaway here is that even though you can’t compete directly with a network effect, you can still beat them by flanking. The mighty eBay was toppled in China. Facebook has failed in many countries, usually because someone else beat them to market during the expansion phase. And for decades, Microsoft had a three-way network effect between Windows (the OS), Windows app developers, and OEMs bundling Windows. But Linux has snuck in through the related server-side computing market, and has made enough headway that some 40 percent of Azure customers run Linux.

It doesn’t matter how far out your competitor is or how much of a lock it seems to have on the market. You can get at them through related markets. It might take a few tries to get it right. Don’t give up. Keep innovating. Eventually you’ll hit a formula that works.

Simple, right? Then why do so many companies still try to tackle networks head on?

You tell me.


Disclaimer: I do not speak for my employer (Grab); these opinions are my own. Also, the timing of this post right before re:invent is coincidence. I like AWS and will be attending!