How Amazon Failed: Notes From A Year in Internet Time by Peter Christopher July, 2000

INTRODUCTION

I worked at Amazon.com from May, 1998 to June, 1999. I was a software engineer. I worked hard, made a difference at Amazon.com, and ended up with some very valuable stock options. It is now 2002. I have decided to post these notes from my experience at Amazon on my website. I wrote these notes up in the spring of 2000. Partly, when I wrote them, I wanted to write an insightful book about my experience at Amazon.com. Partly, it was a 'poor me' story about how certain things didn't work out the way I would have liked. Partly, I just wanted to write down all the important information I could remember for my own benefit, so I could look back on it later and have something to refer to.

This is not good writing; in fact, it's crappy. It has not been proof-read, it has not been finished. I haven't rewritten anything. All I did was write this pretty much in one week, day After day. And all I've done to put it on the website is to spell-check it and change all the names.

I wrote this when I was still pretty emotionally attached to the negative and positive aspects of the experience, and of my role in it. So basically, all in all, there is some intersting stuff in here, written by a crazy egomaniacal personality trying to get over some experiences, trying to remember them and gradually reframe them, let them go, etc..

So don't take it too seriously. Especially if I've written about YOU in here. All the people I worked with at Amazon were amazing people. I am kind of hard on some of them in this set of notes, not because they were bad or incompetent people, but just because I was threatened by them and didn't know how to deal. Thank you all so much because (I don't remember whether this comes through or not in the notes) but it was in many ways a totally awesome year, a great group of people to work with, a real special opportunity which I'm glad I had, which I would do again if I had the chance.

Also, kind of retarded, what I wrote about most was the things that hurt me or pissed me off or some egotistical crap about myself, or maybe some egotistical analysis. There were a lot of other episodes, facts, etc., which I just didn't write down.

I don't think I'm ever going to edit this or put any more energy into it at any point. No warranty.

So if you insist, go ahead and read it. But don't forget it's a PRELIMINARY DRAFT. NOT FOR ATTRIBUTION OR PUBLICATION WITHOUT THE EXPRESS WRITTEN CONSENT OF THE AUTHOR!

Cheers,

Peter Christopher

East Randolph, VT

March, 2002

HERE BEGINS THE STORY

My trip as a Connecticut Yankee in King Arthur's Court was in 1998. I was a Vermont Radical at Amazon.com. Naturally, I had some uncommon challenges alongside the common. Amazon.com was a corporation, indeed a multinational corporation; it was a political beast; it was a computer job; and it was even my first full-time job. It was my first time living in a city, and the year I earned a million dollars. I think for many people this will be a very interesting read.

Like all human beings, I am deeply emotional. I am at times rational, bitter, nostalgic, or confused when I look back on my time at Amazon. I couldn't paint an objective or complete or typical picture of Amazon.com or even my experience; what I will do however is relate my personal experience, sharing my highly subjective experience of being associated with and part of Amazon.com.

It's a portrait also of the mechanics of what my programs did; because I was foremost a programmer.

And a story of how internal politics, personalities, the inability of individuals to deal with our own and each other's feelings, and greed, were the main impediment to progress an Amazon.com. This is a story about what can happen in your life, in your company.

As I'm writing, I ought to, and will, be careful to abide by the NDA (nondisclosure agreement) which I signed when I went to work for Amazon. According to that agreement, I've agreed to avoid providing certain kinds of specific details about what software and hardware they use, information which cold be used against them y competitors.

In December of 1997, I was 22 years old and headed west. I told myself that in a few months I would be returning back east to work with an educational non-profit. Yet I had all my important belongings in the back of my Geo Metro. It was not just a field trip. I was on the road.

I had interviewed for two computer jobs in New England before leaving Vermont. It was on the advice of one of the students who had been in a physics class I was teaching at Vermont Technical College. “Go back to computers, make some money,” he said, “It’s not that you're not a good physics professor, its just that there is big money in computers.” I got one offer out of the two interviews but refused it – working for Liberty Mutual Insurance in New Hampshire, boring!

I didn't know I would end up in Seattle. I didn't know I would end up working at Amazon.com, earning a million dollars in one year, or resigning from my job with millions of dollars of stock options still outstanding. And I certainly didn't know Id be writing this book about the whole adventure. But here it is, from start to finish, with the stories of my one year at Amazon.com.

I wasn't at Amazon.com for the first years, and considering the pace at which things are changing, Amazon is a different place already from the one I describe in this book. Given those limitations though, I still think it might just be an interesting read.

Seattle and the Job Search

By the time I got to Seattle on January 5, my trip was about three weeks old. I had been to New York, Michigan, Des Moines, the Grand Canyon, Los Angeles, and that's a long way in a little car like a Geo Metro. I felt like I was finally able to be my own person, separate from my parents’ money, house, and influence. Not that they had much money or influence in their own communities, but they did have them in my life. I checked into the youth hostel downtown near the Pike Place Market, and went right to sleep – until about 2am, when one of my roommates in the hostel room arrived, unpacked, lay down, and started snoring. I've been told that I snore too, but I hope it isn't the spooky, reverberating, Chinese-torture snoring that kept me up the rest of that night.

In the morning, I headed straight for the phone book and looked up career resources. Apparently I was set on staying in Seattle, getting a computer job (Microsoft, presumably?) and had no intention of returning back east. I made two appointments: with Gary Kanter of Bellevue, to work with me on my resume; and with the Career Improvement Group in Fremont, to attend their free introductory session on finding the right job. Time flies, and so does money. I had about four thousand dollars saved up when I left Vermont, and my goal became to get a motivating, high-paying computer job the instant my four thousand ran out. By the end of January, my four thousand was half gone – but to show for it, I had a great resume, a room in Wallingford, and a plan to find a job.

The process of coming up with that plan involved taking four 3-hour courses from the Career Improvement Group, at about $200 each, and tailor-fitting their plan to my own needs. It felt like an enormous expenditure to me, a total miser in my Dads tradition. But I was sure that just as extravagant as it felt to me, I suspected it was even more valuable as an investment. The trip was important to me because it presented opportunities like this for me to reframe the way I thought about the world, learn about and then challenge my assumptions. This assumption was that I could do everything best myself, and that I worked best independently; in reality, I am a lot better off acknowledging where my skills are lacking, and seeking out experts in those areas. (But boy isn't it hard!)

The job-search process taught at the Career Improvement Group is really interesting. “Networking is dead”, they said, “Classifieds/Human Resources have all the jobs nobody wants, and headhunters take a big cut out of your starting salary. We teach you how to be your own agent, to get the best job at the best salary, period.” Their process involves researching to find hundreds of potential employers in a broad geographical area, which could conceivably have positions in your area of interest/expertise (keeping your data on 3x5 cards). Then, you contact each company to find the name of the person who would be actually hiring you – in my case, I asked for the name of the “Director of Computer Engineering.” After hand-addressing hundreds of envelopes, you mail out your hundreds of resumes/cover letters and immediately start placing phone calls to the people who received your letters, asking them for a few minutes of their time. A few weeks and dozens of interviews later, you have several offers on the table, and use salary negotiation techniques to milk each of them to their maximum before making a choice.

You might think that I just saved you $500 to $2000 dollars, but actually I didn't. The Career Improvement Group people would tell you the same things for free in their 3-hour introductory seminar. What's more important are the details of how you do each of those things. How do you call on the phone and get the name of the Director of Computer Engineering? (The receptionist wont necessarily want to volunteer the information if you just ask – especially because more often than not, there is no exact match for that job title at the company.) How do you ask for a few minutes of someone's time, and actually get it, even if the person is really busy? And how in the world do you negotiate a salary? The details of these activities are what you really pay for in this kind of a situation, and in retrospect, the value I have gotten from those few sessions was tens of thousands of dollars more than I paid for them.

So, as you might guess, I spent the next several months researching companies, making phone calls, addressing envelopes, and interviewing. In all, I sent out 300 resumes, far fewer than was recommended in any job search, and only a third of what they recommended for a position earning 50,000 – 60,000, which was my goal. I ended up with only four interviews but yet four job offers. Despite breaking dozens of the Career Improvement Groups rules, they provided so many brilliant tactics that only employing half of them, it was hard not to get job offers. My worst offense was only calling about a third of the people to whom Id sent cover letter/resume packets. But it was my first run through the process. I cant judge myself too harshly for my failures, particularly considering the results.

Job Offers

I ended up with these interviews: with Expeditors International, Skylight Software, Illuminet, and Amazon.com. According to the strategy, there were supposed to be more interviews, but as I mentioned, I didn't do a very good job following up my letters and resumes. And according to the strategy, they were supposed to happen at roughly the same time. Mine didn't.

The first interview was with Expeditors International. Expeditors is in the business of coordinating transactions between import/export vendors and customs. I learned that their software is written in their own language, known only by three people. Yuck. Now there's a skill that wont transfer. My phone calls were confirming again and again that I needed to be competent in microsoft flavored programming to get a job in Seattle. This proprietary language was obviously going to keep me right at Expeditors. Add a formal dress code, and an offer of $35,000 to start (with $3000 bonus potential) and I had little choice but to refuse.

Next was Skylight Software. Now here was a company I was totally psyched for. Two owners, both around 30, had each quit Microsoft to start this company together. They were very talented, fun, and creative. The interviews at Skylight were informal, comfortable, and challenging. I loved it. A few days later, I had an offer to start the following Monday. I was there in a flash. Skylight had flexible hours – you could work as much or little as you liked, and would be paid a going rate hourly. My going rate was $20 per hour to start. Oddly, I learned after two weeks that most of my hours worked weren't counted towards my pay. Certainly it was understandable, in that I didn't accomplish much which the company could directly bill to clients (the main client being Microsoft): I was just getting my computer set up and trying to remember how to write code again. Three years was a long time, and Id forgotten a lot. On top of that, I had never worked in the Microsoft programming environment anyway.

After two weeks, I got a raise to $22 a hour and a promise that hours worked would heretofore be hours paid. I guess I was pretty good at figuring out a few computer problems and they wanted to keep me around. Nonetheless, I had another interview scheduled, and I went. It was with a telecommunications company, Illuminet, and programming in the unix environment, with which I was definitely more comfortable than with the Microsoft tools. The interviews there went well, and I got an offer for $45,000, about the same as I was earning at Skylight. But I just didn't think it would be as enjoyable as Skylight, so I turned down their offer.

At the time, I was reading about Microsoft's twenty-year development into the computer powerhouse it is today. And I was struck by how it seemed to me that the era of making money by writing computer programs per se or operating systems per se was probably waning; that you could make a good salary, but you couldn't get rich off it anymore. The next phase of millionaires would come from the businesses that are most able to make use of that information infrastructure effectively – whatever that might mean.

Finally, I got a call from Amazon. Amazon was one company where I had had a terrible time figuring out who to send my resume to and had ended up sending it to “Strategic Growth”, the equivalent of Human Resources at most companies. I had a phone interview with Nicholas Lovejoy, who I believe at that time was the Director of Finance Information Technology, which I was sure I blew. For some reason, they invited me in for interviews anyway. (I later learned that Nicholas is one of those people who manages to seem unimpressed by anything, so that all candidates who were screened by him were sure they had blown it.) I went.

I definitely enjoyed the interviews. And my few weeks of work at Skylight with C++ came very much in handy. Although software engineers at Amazon weren't supposed to be programming in C++ (only C or perl), they were certainly careful to test out the C++ knowledge of any potential employees. My first interview was with a programmer from the catalog/search team, and then two with finance programmers, and one from the distribution center team. I loved talking to the one from catalog most of all. Believe it or not, I made the mistake of telling this to one of the finance programmers. “If I had my choosing, I would be most inclined to work with the search group,” I had said. Two days later I was still waiting for the call.

But the call came. “The company definitely wants you, but we just donut think that was the right team for you,” Amazons Strategic Growth guy said. “Would you be willing to come in for another interview?” I was. This interview was with another Director in the company, but I didn't know why I was being interviewed by him. “If I had my choosing, I would be most inclined to work with the distribution center team or some kind of Organizational Procurement Paradigms team,” I told him. Then I found out he was from the Catalog group, “I'm sorry, but I donut think weave got the right position for you with our team.”

One more time, the call came, “Would you be willing to come in for a round of interviews with the Organizational Procurement Paradigms team?” I was. By this time I had it figured out. “So, you're the Organizational Procurement Paradigms team?” I asked, “Well, that's exactly what I'm hoping to work with, Organizational Procurement Paradigms.”

I got pretty cocky when I talked to the Strategic Growth specialist about salary range. “Not including the options, I'd expect to be at the 60,000 level,” I said. He grimaced and choked. I had meant that I'd expect to be at 60,000 if the job didn't offer any options, but clearly he had heard that I meant 60,000 plus options. He explained a brief primer on how the salaries at Amazon.com were below industry standard, but that the total compensation potential was far above industry norm.

Another week later, the call came. I was offered a position with the Organizational Procurement Paradigms group, with a $5,000 signing bonus, $54,000 starting salary, and 6000 options. I was pretty psyched.

write about first day of work - no computer, no desk, no phone for 2 weeks; first day was an All-hands meeting in an auditorium. The Germany, uk, and planetall acquisitions had just been announced, and Joy also gave a spiel on the status of the company, including exponential graphs, etc., which she'd been sing on her fundraising tour raising 300 million of capital. I met her on the elevator that day. Then I met Jeff Bezos in the shitter.

We knew Angus Franks would come on board in the middle of the summer. He was a programmer for Sun Microsystems who was moving to Seattle. He would be working with me on the procurement tools suite. I felt it was part of my job to train him although Hal hadn't talked with me about his arrival at all during the month before Angus showed up.

In fact, the day Angus did arrive at Amazon took me completely by surprise. As always, I came to work in the late morning and cruised into my office to inspect the carnage of the morning, contained in the dozens of emails waiting for me.

I didn't make it to my computer, though. Some young, handsome gentlemen was standing in the office. Starting from the bottom, he wore dressy (yet casual) black suede shoes; a clean pair of jeans with a dressy (yet casual belt; a well-ironed fashionable (yet casual) long-sleeved cotton shirt; and professional (yet casual) spectacles. There was no question he had mastered the fashionable (yet casual) ideal which is ordinarily only within the domain of the modern woman. Of course, perhaps it was after all just his girlfriend's doing.

"Hi, I'm Angus Franks," he said, "I've been meeting with some folks from around the office." Why should I give a shit, I thought.

"You are Peter Christopher, aren't you?" he then asked. I nodded. "Well, I'm Angus Franks," he repeated, "I'll be working with you on the procurement tools."

Aha. "You're that Angus Franks!" I said. "Sorry to be so dazed; you'd think I would have had some idea to expect you today, but nope, I had no idea it was today, or this week, or even this month. Well, welcome."

We sat down. As he told me about some of the work he had been doing. I found out that Angus was definitely a talented programmer as well as a talented dresser. He seemed like a nice guy, and I was glad to have the opportunity to work with him.

I took my role training Angus very seriously - partly because one of my talents is teaching, and partly because I felt this was my opportunity to begin developing my own team at Amazon.com.

Over the following months, I took Angus with me to an assortment of meetings so he could meet all the people I thought were important to our work. I deliberately told him my opinions about how each person worked and how I came to know them. I shared with Angus what my goals were for myself and for the procurement tools, and how I was going about achieving them - as well as areas where I disagreed with our boss Hal Jeffries, or anything else about the company.

Our relationship also went beyond work. Angus's girlfriend Jill was a graduate student in Seattle, and the three of us went out to dinner once every week. We traded off who paid from one week to the next and went to a different restaurant each week.

My initial impression of Angus's prospective programming ability was verified fairly quickly. I had a list of tasks which Angus and I would be working on to further our general goals with the procurement tools, and with the two of us working together we made substantial progress on the international project as well as scaling initiatives.

Over the six months before Angus started, Ed Fantango ad I had been primarily responsible for support of our tools at al hours. It was a really stressful period as far as support goes, due in large part to the quantity of programs we maintained and the rapidity we added and modified functionality. Over those six months, I was often woken up by emergency problems every other night. Ed and I brought Angus into the fold of support, and he began answering midnight pages with the rest of us - for a brief period during international launch, even taking the brunt of the support calls.

At one point, Angus began coming to work visibly perturbed and began expressing his frustration with the development environment, which he felt was responsible for the midnight calls. I disagreed - I felt the midnight problems were due to the rapid pace of change in the uses of our programs and the regular rollout of new functionality. I felt that very few of the emergencies could have been avoided by using a different programming language, no matter how ideal.

Although I did disagree with Angus, I certainly would not have minded being proven wrong. If we could achieve more rapid development with fewer bugs by switching to the java programming language, I was all fr it. I also felt it was important to give Angus respect by trusting his approach to one of our next problems. So I supported Angus's desire to develop a java interface into several of our main automated ordering tools, while I worked on improvements to the manual ordering tools in the C and perl programming languages.

As we entered December, all corporate Amazon employees received news that we would all be assigned to a warehouse team for the holiday rush five days per week, and that we would only work one day per week on our normal jobs. Technically, this included about 90% of the "downtown" corporate employees. The last 10% were considered to be working on such mission-critical operations that we could not be sacrificed to help in the warehouse. Ed Fantango and I were deemed mission critical because of our particular projects (by Hal Jeffries) and Angus was not. Since Angus had to go to the warehouse, he was not able to continue his project.

The tools I was working on at the time were a new generation of manual ordering tools for several different internal groups. At that time at Amazon.com, there were two very distinct groups who were placing manual orders with publishers: one group called predictive buying which pre-ordered large quantities of popular books, and one group called special orders which placed orders for individual, uncommon books only after a customer had ordered em from Amazon.com and Amazon.com had been unable to order them electronically from a distributor. Typically, books ordered by predictive buying are always in stock and able to ship in 24 hours, whereas books ordered by special orders take 4-6 weeks for Amazon to ship. Obviously, there was a commonality in what these two groups did: they both manually contacted publishers to order books for delivery to Amazon.com warehouses. In each group, there were lists of publishers for each buyer, and each week the buyers would contact the publishers on their list to place orders for a set of books for that publisher. If you are noticing the duplication of effort inherent in this process, you are understanding the tremendous inefficiency of this situation. Every week, two different Amazon buyers contacted the account representatives of each publisher and placed purchase orders for each Amazon warehouse. Each of those purchase orders would then be received separately at each warehouse, and each purchase order would have to go separately through Amazon's accounts payable to pay the publisher after the books arrived.

The problem was partly historical, because the two manual ordering groups had two different sets of tools (programs) maintained by two different software teams. Each set of tools was also out of date in its own right, but that's a separate issue. The special order tools had been written and maintained primarily by Ed Fantango in the Organizational Procurement Paradigms group, whereas the predictive buying tools had been written and maintained by an assortment of non-supply-chain programmers.

The reason the two buying groups had remained separate until that time was that the programming teams had not been working towards a solution where the programs could be combined. Most programmers weren't even aware of what the problems were to this progress, so as they worked on their own programs, they weren't able to make progress towards combining them.

The technical problem which I am talking about is that a common framework for various kinds of pending orders had not been adopted. That's it. Now you know one problem which (probably until this day) costs Amazon.com about a million dollars a month.

The reason I became familiar with these issues was because after the international project launched, I made a deliberate effort to interview buyers from each of the buying groups, and the programmers responsible for each set of programs. For each group, there was pending demand (books which had to be ordered) and pending orders (which had been mailed or faxed to the publishers but for which the confirmation had not been received by the publishers).

I learned that the pending demand for the predictive buying groups was stored in excel spreadsheet files, and the pending orders were stored in the oracle database in the confirmed orders area (a data integrity problem which gave me goose bumps). In contrast, the demand for the special orders programs was stored in the Oracle database as special order demand, and replicated in a non-oracle database. The special orders pending orders were stored in a combination of the non-oracle database and in the memories of the special order buyers. Interestingly, the reason why the special orders data was stored in a non-oracle ("Berkeley") database is because it was written by the head of special orders, who was not invited to participate in Oracle Database development because he was not officially a software engineer. Since then, he (Ed Fantango, who I've mentioned) had moved into the Organizational Procurement Paradigms group, in software, and is one of the most dedicated and productive programmers. At that time, however, his special orders tools were still based on the Berkeley database, despite inherent problems with the Berkeley database and incompatibility with the company-wide standard Oracle database.

Just for completeness, I should also indicate that there were several other sets of buyers with other sets of programs as well, including the Amazon Advantage program (described on the Amazon website), out of print ordering, and music special orders. I had done some work with these tools as well, and my goal was to eventually eliminate as much duplicity as reasonable from these programs eventually. The bulk of the publishers orders, however, were through predictive buying and special orders, and that was the source of the primary inefficiency as well. To be honest, these goals were not only mine - many individuals who used any of these tools were aware of the inefficiency of duplicity and hoped to eliminate it. Through my investigations, I became the only person aware of how to actually solve these problems in the near-term, and as I began working on them, I was the only one who as making any progress.

Although I had a good understanding of this problem, I was not successful at solving it due to political blunders. When I showed my boss what I was working on and explained it to him, he became extremely frustrated at how I was wasting my problem on this problem, canned me from the promotion he had been tempting me with, and removed me from working on the procurement tools without consulting me.

WORK

The all-hands meeting with the Junglee Acquisition, "shop the web", "boom chakalalaka... spread the jam on amazon" "First we thought shop the web would launch in Dec; but all we got was a link." We who worked at Amazon thought it might be really major; would be when people would begin to use Amazon for all shopping; for some reason, it didn't work well. Partly this was because of secrecy perhaps and the dedication of many SE to auctions....

I should probably explain some background before I describe the circumstances of that terrible interaction with my boss.

My boss Hal Jeffries ad done an amazing job recruiting brilliant programmers to work on our team. He had a great ability to get the best candidates really excited about working on his team. He also worked extremely hard to keep recruiting, making our team the fastest-growing team in software.

One problem, though, was that his spiel about what his team would be working on was unrealistic. He painted a picture with each candidate of our team working on interesting architectural issues, the stuff that brings tears to the eyes of software engineers: developing a state-of-the-art event-driven procurement system, optimizing programming languages to deal with salability issues, and writing object-oriented code.

He was smart enough to see that this was what the candidates wanted to hear, and he also believed those were potential issues which our group should deal with. I learned through my interactions with Hal that he generally didn't tell outright lies. But this is because in his perspective, there are no outright lies since most everything is partly rue; and if there's anyone who is capable of arguing an unlikely but possible point, it's Hal Jeffries. Incidentally, after several glasses of wine at a party, Hal confided to me that this quality of being able to effectively present the partially true as the entirely true was something he admired about Amazon.com CEO Jeff Bezos whom he knew personally. When I asked him whether he was proud of his own ability to do the same, he said absolutely. Does this kind of opportunism contribute or detract from an individual's leadership? Perhaps it adds effectiveness in some was and takes it away in others.

So the result of this was that our team, OPP, consisted of a vast number of programmers who were excited about architectural issues and expecting to spend their time and energy on architectural issues. Hal would continue to emphasize architectural issues in our group's priorities, and about half of the group had very little interaction with the main functions of our group. Hence, for insurance, one subgroup was working on the distributor returns project for many months. This project was designing a program to aid the warehouses in identifying overstocked books to return to the original supplies. Since it was pilot project written in the java programming language, the team spent a tremendous amount of time working on ironing out basic architectural issues. Indeed, I think they were working primarily on architectural issues and then on the distributor returns project as a prototype using their new architecture.

I certainly think there's some value in the R&D-type work which was going n; and Hal had been effective at building a large enough team so that we could have people working on those kinds of more strategic work. As we go on here, you'll have to decide for yourselves whether the ways these goals were prioritized was ultimately a problem or not.

The Organizational Procurement Paradigms team had been tasked in the second half of 1998 with purchasing a demand-planning software system, and by December, that software was ready to roll. Chris Jenkins was the team lead for the demand planning group, and in November, he and I had each come up with the need for new tools for the predictive buying groups. His goal was for their software to deal effectively with the new demand-planning forecasts his demand-planning software would be generating in December. I felt strongly that the new software should deal with pending orders by using he proper tables in the Oracle database, and so I began working on the project myself in late November, with the help of Sandra from predictive buying.

By mid-December, my software was almost ready for initial live testing. Sandra and I hoped that she and I would be able to do initial live tests of my ordering tools in mid-December, but it turned out that what happened instead was that in mid-December, Sandra's entire group was called down to work in the warehouse in the Christmas rush and we didn't put the software into production until early January.

In mid-December, I spent some time at the warehouse helping out the Christmas, as well as time at the office finishing my work on the new predictive buying tools. I was getting very little sleep and stressing out in the final days, finishing software I'd promised for the date I'd promised Sandra. One night about 2am, I was about ready to call it a night and walk back home when the phone rang. It was my boss, Hal, who wanted to check in with me. I was about to have the worst phone call of my life.

I was really proud of my progress on my tools. I explained them to him, and he tried them himself over the internet. But whereas everyone else I showed them to was excited and pleased, Hal was annoyed.

"I think this is just the wrong time and the wrong technology for this, Peter. These kinds of tools shouldn't be built over HTML, but rather in Java on an enterprise services architecture," he said.

"And why is that?" I asked.

"Because ultimately it's much more easily configurable."

"Well, I can see your point that we should consider these issues. So, as for configurability, what kinds of configurability are you talking about?" I asked.

"For instance, you might some day want to have the requirement that purchase orders of a certain size be forwarded to a manager for approval."

"OK, that's a great example for us to consider," I said. "Now, management approval sounds like the kid of data which which requires sufficient integrity that we would want to store it in the Oracle database, right? And when we're talking about modifying the database, we're talking about an extended development process involving the database modifications committee, the database administrators; the software development whether built in HTML or java is the smaller part of the greater process work, so at least in this example, it would seem that the advantage of modifiability resulting from a different programming strategy wouldn't necessarily be obviously huge."

"It sounds like you ust don't have the experience to be able to address this issue and that you just have to trust my greater experience in situations like this," he said. At this point I was feeling rather upset, and unfortunately I had no idea how to deal with this conflict. One bright idea would have been to suggest we sleep on it and talk tomorrow instead of when I had hardly slept for several days in the midst of the Christmas rush. However, I selected this alternate tactic:

"Hal, we have thus far had one significant time when we disagreed about something like this. That was when you aid that you thought Brendan would have a message-passing architecture ready in time so that we could design the international version of the procurement tools around it. I disagreed at the time; you said I should trust you, and I agreed to. However, it turned out that my original analysis, perhaps due to my more recent discussions with Brendan, was correct. So it makes me uncomfortable when you say I should trust you based on your greater experience again. We still don't have that architecture in place; perhaps I deserve some room to be trusted here as well."

Hal changed the subject. "Why aren't the eo-1 and eo-2 machines in use yet?" Hal asked. eo-1 and eo-2 were two new production machines to replace the hopelessly overworked computer called 'carajas' (which I preferred to refer to as crackass).

"Basically, you gave the responsibility for that task to Freddy Martin, and he hasn't got it done yet. He hasn't even got any directories mounted on it yet so we can't possibly migrate to it. Frankly, I'm not even sure if he knows how to run the most elementary commands in the unix operating system." Freddy Martin was Hal's chosen left-hand management man who had oriingally bee hired as a software developer before I came on, but who actually didn't know how to use or program in the "unix" operating system, which was the only one we used at Amazon.

"I've been really frustrated by how slow he is being getting in the technical stuff as well," Hal said.

Then I lost my composure and all my frustrations and pain came out in a loud sob. After a few moments I was able to talk again. "I feel like Angus, Ed, and I are working on these important tactical projects but aren't being recognized in the group in any significant ay for our contributions; all our group meetings focus on strategic projects and strategic accomplishments and that's important, but it's only half the situation here." It was hard to ge that stuff out - I was still having to hold back my tears. I'm sure I was making Hal uncomfortable and giving him that much more evidence of my imperfection.

"I really should make it more clear how important tactical work is in our group," he said.

"How about this, can we agree that I will get eo-1 and eo-2 in production myself, and you will take increased devotion to respecting the tactical work in OPP alongside the strategic."

"That sounds great, Peter," he said, "You get some sleep, OK guy?"

With that conversation, I realized that I had come face to face with the monumental choice of whether either to work politically for my ego, to be impersonal and respectful and effective, or else to be open about my feelings and who I was. I couldn't remember having made the choice; I could only remember the events. My desire to connect with the people I spend time around at a more personal level had become so great that I just unloaded. My fantasy about being the ultimate success at Amazon, nourished by earlier interactions with Hal and Brendan, and Rick Dalzell, and through my successes in the company, were gone. I was just a guy with lots of imperfections who happened to be in my shoes then.

At the time, I had about five months of work at Amazon before my first year of stock options vested and I could quit. The stock price at the time was at an all-time high. If it stayed there, then on the upcoming May 4, I would be worth $800,000 after already paying $500,000 of taxes. Life became a miserable, ironic, disgusting, insane waiting game. I had less than five months to go.

In a Chinese restaurant, I got a fortune cookie which said, "Sometimes it is best to keep your intentions to yourself." I taped it to my computer monitor and kept my mouth shut.

WORK

That christmas (98):

planetall to Delaware

away team

corvette

times to be at work

record shipping

meals

jobs

temps/friends

Toys/CE

50% off DVDs (one other company ordered from us)

I've always had a taste for impatience in my achievements. Maybe it is because my parents were set on proving their worth through my precociousness. Maybe it is because I am just naturally an impatient person. For better or worse though, I fantasize about accumulating achievements a the youngest age possible. Then I set myself on track to make those fantasies real.

Ultimately, I sometimes succeed and sometimes fail. As you may suspect, in the pursuit of my Amazon advancement I was going to fail; and in pursuit of my hoard of cash I was going to succeed.

By about the 15^th of December or so, the order flow had slowed down enough so that most of Amazon's downtown employees were able to return to their normal jobs - albeit a majority of them with sniffly-sneezy headcolds from the 8-12 hours of graveyard or swing shift. Hal had Freddy Martin organize a meeting of selected OPP team members to work on the "new" services-based architecture of OPP. This was the same project which mos of OPP ha supposedly been recruited to work on, over the past year. Although progress was being made on what was called 'backend software rearchitecture', it was far slower than had originally been planned. Many of the people working on the project, including Hal Jeffries and Red Smith, were in a perennial state of euphoria supposedly because we were "finally" about to really make serious progress on the project. At first, I was excited for the team and the company and the "new" project. But after about six months with little progress, I was getting sick of the unrealistic sense of necessity and arrival that Hal continued to preach and others continued to mimic. So at the time, this was about the sixth month of these meetings, drawing increasing amount of manpower and going over endless repeating scenarios of how Amazon's systems might work some day. Due in part to my terrible phone call with Hal, I was not invited to these meetings. And I felt the effectiveness of the meetings was overrated so I didn't make it a point to attend. But I told Angus about them.

Angus was inherently more interested in the architectural kinds of issues, like how to best design a message-passing layer between programs, so he was a more natural choice to be responsible for that aspect of the purchasing tools. He was also hot to trot his ideas and experience before the team. When I told Angusa about the meetings, he stood up from his computer, said "I'm going" to me, and tookoff down the hall without another word. As if to timed to perfectly coincide with my disastrous conversation with Hal, Angus was ready to take my most-favored-fellow spot and the new management position Hal had invented for me. So that's how it happens. I thought to myself. Well, at least I knew for future reference.

Two days later I took a week off for Christmas, which I'd been planning.

WORK

Hal's indications:

meeting with Dalzell,pushed out...

said i was become leader

Brendan Reid conversation (and perhaps Brendan's history w/ mgt)

international post-mortem

data integrity team

Rick Dalzell's comments in bathroom - great work, hearing about me, etc.

On January 1, a sign appeared on top of the PEMCO financial center: 365 days until 2000. It didn't take me long to calculate my own special date. May 4 was my anniversary, so I had 123 days to go. Every time I drove by PEMCO on I-5, I checked to se how close I was to "242 days to 2000". Sometimes I wondered how many other paper-rich internet nutties had their own option maturity dates vesting according to the calibration signs atop PEMCO in downtown Seattle.

It would have ben nice if my friendship with Angus had been deeper. Actually, I thought it had been. I had told him about my aspiration to lad our small team, as well as some of the indications Hal had made to me about my potential leadership role. It didn't become clear to me until Angus curtly charged out of the room, that he shared my aspirations for leadership as well. I suppose also, I never asked either.

On the first day of work in 1999, Rick Dalzell stepped into out office and asked Angus to step into Hal's office for a minute. Ten minutes later, Angus returned and sat back down at his des without a word, stiff and uncomfortable with his baseball cap pulled down over his eyes.

"Feel like a new man?" I asked.

"Yep," he replied, without turning around.

A few days later, Hal asked me to sop by for a meeting.

"Peter, I wanted to talk to you about some changes we're headed towards over the next few weeks. As I've mentioned before, there will be subgroups within OPP - three strategic projects and oe tactical team. Since you've been shouldering a great deal of tactical responsibility there over the past year, I'd like to move you off of the work you're doing, and one one of the strategic teams I'm announcing tomorrow."

"I guess we've had some disagreements over the past month, haven't we - " I prompted.

"Year, well, this might be a good change of pace for you. We're going to have a tactical team within OPP now, with a tactical manager, and that tactical manager is going to be Angus Franks. It would have to be either you or Angus, and I've been really impressed with the maturity he has shown recently."

"Will I be able to complete the work I'm doing on the manual ordering tools?" I asked.

"That will be up to your new team lead, Ben Saran."

"I'm uncomfortable working for Ben, Hal. I really don't feel like he has any technical grasp whatsoever."

"I don't agree with you at all , Peter. What grounds do you have to possibly contend with Ben's technical competence?"

"I'll tell you. For two months, he was supposedly working on these 'workflow tools' and didn't make any progress - never began actually producing any single line of code or even a demonstration model, nothing whatsoever. That made start to wonder. Then when I started working on the workflow tools he agreed to be responsible for putting together a simple system for importing configuration data about the publishers into a format we could use.

"Well, after two weeks of working on the project, Ben came in on a Saturday with his wife, and she helped him get the data out of the Access database, and he wrote a short program which was supposedly going to manipulate the data. Funny thing is, his program didn't even run, much less have any desirable effect. It was clear that he had tried to make it work, but also clear that he had failed miserably at everything from conceptualizing the problem to coding.

"I know that if I'm working for Ben, he won't be competent at his job, and the only way the project can succeed is for me to do both my work and his."

I suppose telling your boss that his personal choices for team leads are incompetent may not be the best way to build favor. But I was pretty hopeless.

"Peter, look," Hal said, "Ben was highly successful in installing the fax program for special orders. He went through the process of selecting a software package, and installing it."

"Hal," I replied, "Listen to what you're saying - hi installed a fax program; this is only slightly more complicated than installing Windows. And from my interviews with special orders people concerning the manual ordering tools which I'm working on, I learned that many of them don't use his fax system because it's not reliable."

"Listen Peter, this is getting ridiculous. If it were most anyone on this team, I would change his assignment. But I just think you're far above this. I think you really need to concentrate on figuring out how to make this relationship with Ben work. He is a highly competent contributor to OPP, and I just don't see where you're coming from with your accusations."

"I'm sorry. If you feel strongly that this is all in my head, I definitely should take more time for this and give it a chance. Sorry to be such a pain in the butt about this, Hal."

And on that note, we parted.

I finally realized that Hal Jeffries was a human being. He had seemed so puffed up. He was superhuman. And yet, here I finally had to confront that my head really was making me see things. There was at least one chink in his armor, at least one blind spot in my vision: Hal had difficulty differentiating between a real thing (a successful project, a realistic timeline) and an imaginary, illusory version presented as if i were realistic or successful. Maybe it was because he had become a manager. Maybe it was just his nature as a human being. I became clear to me that this inability permeated his entire relationship to Amazon.com; his entire being and probably every relationship he had with anyone in the world at that time. Hal, I realized, should e in Marketing, not Software Development, and it was tearing up our team and the company to have such a massively intelligent, hard-working man in the wrong job.

HERE

Despite the severely rainy weather, January had its moments. At the insistence of the direct buying groups, my ordering tools (which they called "Workflow Tools") were made the top tactical priority. I was asked to finish them. I said I could all the bugs out of the no-frills version and have it in use by the full predictive buying team in two week. I was given two weeks of time from my 'new assignment' to finish the ordering tools. At my request, I was also given an additional two weeks to document the most important sections of code I had written over the previous six months.

There were two parts of the new annual ordering tools I was finishing for the predictive buyers: sending of purchase orders, and confirmation of purchase orders. I was extremely proud of the way my new tools worked, so I'm going to related the whole process to you.

A couple times a week, a forecast of popular books would be produced by certain software (maintained by the demand planning team in OPP). The forecast would predict upcoming sales predictions for all Amazon's books. Then, based on the inventory and books on order for each warehouse, a file was made of book to be ordered. The predictive buying group would then make any final changes to the order quantities (for instance, potentially increasing the order quantity for books which would be featured soon in the media).

Here's where my programs begin to come in. Once the file of books to order had been finalized, it would be run through my sending program. The sending program split up books by publisher and made one purchase order for each publisher. Each publisher would then receive an automatic email, fax, or EDI file for each purchase order. (EDI stands for electronic data interchange and is the quickest and easiest way to send an order, because it is completely automated and electronic. Of course, a publisher has to have an EDI computer system of their own set up, and only the largest publishers had EDI systems. So a huge chunk of orders had to be sent out through the emails an faxes.)

The emails and faxes were each based on templates (determined by the predictive buying group) In the process of making each purchase order to send, the genAngus items in the template (like PUBLISHER_ADDRESS or TOTAL_SHIPPING_COST) would be replaced by the correct data for that publisher.

The process of sending ot hundreds of these purchase orders literally took only ten minutes. The system in existence before I arrived took several people several days to complete. The time savings alone freed up a substantial chunk of time for buyers and allowed any more purchase orders to be sent to many more publishers. This was a major win for the company and impacted the bottom line profitability by at least a half-million dollars per year. This margin was due to improved discounts when ordering directory from publishers than from distributors (generally the cost was about 20% lower).

Then it was time to get confirmations back from the vendors. The confirmations would usually come back within a couple days. Each confirmation would contain quantity shipped, quantity backordered, list price, and discount for each item. Normally the publisher would just confirm they would send the entire quantity at the discount which we had printed in the purchase orders, but sometimes there would be a change in quantity or discount.

There were then two more important tools: one for checking the status of pending orders, and one for confirming orders. Both these tools worked as cgi web-pages. Pending orders could be searched based on publisher, dates, status, type of order (email, fax, or EDI), or by the Amazon buyer. Thus, any Amazon buyer could (for instance) search for all pending email orders over a week old which had been sent to one of their publishers (each Amazon buyer was responsible for a group of publishers). This turned out to be much easier than trying to remember the status of all outstanding orders in a buyer's head or keep track of them on scrap paper or in a spreadsheet. At any given time, it was usual for each buyer to have dozens upon dozens of orders outstanding.

The final tool was for confirming orders. When a response had been received from a publisher, either over the phone, on email, or in a fax, the buyer then used the confirmation web page to adjust quantities or ordering status and confirm the order in Amazon's database.

Nine months after leaving Amazon, I talked with one of the buyers from the predictive buying group whether they were still using the software I had written for the, or whether they had been replaced by newer, java-based tools (as was supposedly going to happen shortly after I left).

"Nobody else has replaced those tools. They revolutionized the was we do business, Peter," she said. "In fact, I'm flying to England to teach the Amazon.co.uk how to use them now."

That felt nice to hear; one year after I had finished working on the they still worked.

"Do they still come up with a different background color for each PO you pull up? I remember how you requested we do something for a background other than the blue screen of death!"

"It still has a new color every time. And I still chuckle about it, too."

My new project was supposed to be the definitive inventory super tool. For the first few week of 1999, the other three members of the new team began working on it as I finished up the work I described in the previous chapter.

The whole company had taken on a different feel. A corporate, impersonal, highy structured and secretive, militaristic feel. There were traces of this feeling since I first began working at Amazon.com in may of 98 but it didn't really feel inevitable or definite. In 1998, many employees, including myself, were intrigued to learn about projects throughout the company, and most everyone knew about all internal projects. Corporate buyouts were another story, but usually these were announced and internal projects began, to incorporate them into Amazon.com.

All that changed at the end of 1998. Rumors leaked out that we (Amazon) hd a significant majority stake in Drugstore.com. (I first learned about this from a friend of mine at Drugstore.com who said the news has leaked - which I hadn't known yet - and that Jeff Bezos had been extremely irate about the leak. The next week, the news was in Business Week.)

At the same time, internally in the company many secret projects were under way. One secret project was headed by a new employee, and never came to fruition. Another secret project consisted of 15 of Amazon's best developers and wa headed by two seasoned Amazonians. This group all moved to one corner in the fourth floor of the Columbia Building. In three months, they completed the entire system for Amazon.com auction site from scratch, including purchasing enough hardware so Amazon could handle all internet Auction traffic and designing a site which was better in several ways than the other auction sites available. Amazingly, the secret project's nature never leaked until about a week before final launch, even with the company.

I think the secret projects were demoralizing. Since I haven't any other corporate experience besides Amazon, all I can compare are Amazon when I arrive in the spring of 998 versus Amazon in the spring of 1999. What it felt like was that what little community we had, around a meaningful purpose, were completely relegated and obliviated by the new corporate purpose: to maximize our wealth. I didn't like it. I wasn't the only one, either. There was never any kind of formal announcement or indication within the company that our mission had changed, because it hadn't. There was never any kind of formal indication that our tactics had changed either; but the lack of communication was truly half of the change in tactics in the first place.

The mission of Amazon had always been: to get rich by being the most customer-centric-feeling company in the world, where you can buy anything online. When no significant competitors had existed for Amazon in the online consumer shopping space, it wasn't a problem within the company to b open about projects and goal. However, when competitors began to surface - ebay, prieline, Yahoo, AOL (sometimes because they introduced competing products, sometimes because we entered their markets), then Amazon changed tactics nd became secretive. According to rumor, this had more to do with knocking down the market capitalization to reduce the efficacy of secondary offerings by competitors, and less to do with new products the competitors could launch. But since the company was so secretive, I didn't know what the real reasons had been why the management had made the decisions they did.

Once secret projects began to really et traction, the leaders of my team wondered whether their project was a secret project.

I guess many companies have to have some degree of secrecy in their work - particularly in their R&D work. Since the nature of the game is cutthroat competition, they cant leave themselves open to disadvantage. On the other hand, in the way it feels to the employees and in the efficient of their work, it may create some problems. Ultimately, is that secrecy important to getting the shareholders rich? (And all employees are shareholders at Amazon). Yes, it probably is Yet, does it make it a less pleasant place for many people to work, and in some ways, less efficient. ? Yes again. Whether it is possible to openly run a large, profitable company, I don't know, but it would probably require some tradeoff of money versus enjoyment of life, and Amazon was never interested in making a tradeoff.

- financing and profitability

This chapter is more about my impressions of the finances of internet companies and less about my experience per se. since I realize some readers may be less interested in this subject, I've tried to concentrate the financial discussion here in this chapter. Feel free to skip ahead to the next chapter.

In early 1999, Amazon's stock valuation was rather speculative: the cash on hand and tangible assets of the company (less than $200 million) were minuscule compared to the market capitalization of $30 Billion; the company had gross margins of only about 20% and no profitability, yet a price to sales ratio in the stock market round 50-to-1. These numbers don't even take into account the expected dilution of Amazon's stock due to stock-options and acquisitions.

Speculative mania like this makes it tempting to grow the company through stock acquisitions and secondary stock offerings than through cash flow. Of course, this must also present an important dilemma for stockholders. If the company's prospects are really such that the stock is worth so much, then isn't it important to limit the dilution of the company's stock? With speculative stocks like Amazon's, the current assets and profitability are really irrelevant when compared to the future potential of the company. Amazon's valuation is based on the potential that Amazon can someday sell a huge quantity of a huge variety of products profitable. And there really isn't much risk built into Amazon's $30Billion valuation. To reframe the question for shareholders again - if Amazon issues new stock worth $250M in connection with an acquisition, will that increase Amazon's value by at least $250M beyond the already-optimistic $30B? After an acquisition worth $250M, the new market capitalization of the company would be $30.25B (assuming the stock price did not change), so the company had better have increased its future potential by that .25B ($250 M).

The same kind of reasoning applies to stock options in employees' compensation packages. When employees are given stock options, those dilute the stock as much as an equal amount of stock issued in an acquisition. Depending on whether the stock ever becomes more valuable, however, stock options may or may not be exercised and may or may not be dilutive therefore. I'll get into that more shortly.

One other kind of dilution is a stock offering of some kind. A direct stock offering is a reasonable and prudent thing for a company to do when the stock is fairly valued, and if the company can invest the added capital more profitably than the expected return on the stock price. The immediate result of a stock offering is that the number of shares outstanding increases. So if the stock price stays the same, the market capitalization increases the exact same amount as the amount of case received. So if a company had a $2B stock offering, it would then have $2B of additional cash to invest to justify the additional $2B of market capitalization.

If a company's stock is overvalued and the management expects the price to fall, it can be in the best interest of the company to raise capital through a stock offering even if the added cash could only be invested with a minimal turn.

Conversely, it is generally a real mistake for a company to raise money through a stock offering when management believes the stock is undervalued. This is because when the price of the stock goes up to a reasonable level, the company will have less invested capital thanks to the stock offering, than the market capitalization will have increased due to the stock dilution.

Unlike Amazon.com, many companies actually earn money. With their earnings, they can make investments in further facilities, R&D, etc., to increase their future profitability. This is a very appropriate business model for many companies. However, this is simply an inappropriate model for many other companies. One obvious example of such a company is a medical company doing research in a very specific field. These medical companies might not even have any sales whatsoever, much less profit. The whole point of these companies isn't to be immediately profitable, of course; rather, the goal is to be profitable after five to ten years, when medical research pays off in direct sales by the company or license fees licensing a drug to a major pharmaceutical firm. These are reasonable business models for the business these companies are in. Because there is public capital interested in these companies, the public can be reasonably engaged in investing in these riskier investments, which are likely to have a higher long-term return.

This basic business model has been adopted by hundreds of internet companies, because there has been a tremendous amount of capital made available by investors confident in the internet revolution.

Because of the rapid pace of growth in the popularity and use of the internet, and in the availability of speculative capital in the private and public markets, it is altogether fitting and proper that such business models be employed by the internet companies.

So the question on everyone's mind is of course whether the internet stocks they are buying in direct offerings or on the market are overvalued, undervalued,or fairly valued. Let's look at some interesting perspectives on that question.

After one of my interviews at Amazon.com, I had picked up a copy of the 1997 annual report from the receptionist on the way out. It had been printed in around April of 1998 and I was reading it in May. There were some amazing features of this report - annual revenue growth around 600%, ongoing operating losses, and most interestingly, a $100M loan, convertible into Amazon's stock at a split-adjusted price around $5. At that time, Amazon's cash flow was nowhere near $1M/day, and together with X M already on hand, the $100 M loan seemed like way more than enough to cover operating needs for the foreseeable future. It seemed to me that Amazon had some serious strategic ambition and had armed itself with a warchest for expensive new projects.

That intuition proved correct - sooner than I had imagined. A few days later, Amazon made me an offer to join their Organizational Procurement Paradigms team, and then the next day Amazon announced its first three major acquisitions: Bookpages, telebuch, and planetall.com. Bookpages and Telebuch were leading online bookstores in England and Germany; planetall was a calendar service.

Shortly thereafter, Amazon took out $350M of bonds around 8%, and immediately paid back the $100M convertible loan. Apparently, Amazon management believed their stock was worth more than $5/sh. The stock market agreed - at the time it was trading around $10/sh.

The $350M helped Amazon through numerous expansion projects like music, international launch, and video launch

In late 98, Amazon's stock price reacted favorably to Oppenheimer's $67 price target, trading at times u around $100.

The market optimism presented an opportune time for management to issue some sort of security into the public market. A convertible bond was selected - convertible bonds function much like normal corporate bonds, having a coupon rate payable at regular intervals, and a maturity date when the face value is paid back to the holder. But if the company's stock rises significantly, the bonds can also be converted into common stock and sold at market value by the owners. Convertible bonds tend to have a lower coupon (interest) rate than standard corporate bonds, and are generally issued with a conversion price 20%-40% higher than the current market price of the stock at the ti of the bond issuance.

Amazon's bonds had a coupon rate around 4% and a conversion price around $75 although they were issued when the stock was near $60. Originally, Amazon had sought to issue $500M of these convertible bonds but due to high interest in the issue, increased that amount to $1.25 Billion on the day of issue. If The conversation privileges for Amazon's stock were all exercised, this would increase the number of shares outstanding by about 5-10% but would also have the effect that Amazon would not have to repay the bonds, or pay any interest on them whatsoever.

Judging by this offering, one would think that Amazon's management did not believe its stock was significantly undervalued at $67/share.

At the all-hands meeting following the bond issuance, a substantial presentation was made to all Amazon employees concerning the issuance. We were told that the issuance was in case the public markets 'went south on us' (in Bezos's words) - although he never specified whether that might be due to a general market correction or simply an Amazon correction.

In my personal opinion, Amazon's stock at $60 was t the highend of a reasonable value, butt not completely off the scale of reason. I actually believe this convertible offering was great for the shareholders and company but terrible for the bond-buyers. I think it was great for the company because it was issued at a high valuation for the company, so he new potential shareholders weren't getting a disproportionate amount of the company for their investment, in particular because the conversion premium reduced the value received for their investment even further. On the other hand, for the bond buyers, it was a terrible investment because despite these drawbacks, their stability of principal is far from guaranteed: the ability of Amazon to pay back loans in the millions and billions of dollars is only likely if the speculation about Amazon's potential massive future profitability is true. If Amazon cannot be hugely profitable in the future, there is no way it could pay back the $400M of outstanding senior debt, much less the $1.25 B of convertible bonds. And if Amazon is really going to be so hugely profitable, then the investors should have bought the straight stock in the first place, rather than sacrificing the conversion premium. For God's sake, at the time this $1.5 B of bonds was issued, Amazon had still not sold $1.25B in products over its entire life history in business.

When it comes to the revenue growth and profitability of Amazon.com as well as other internet companies, investors have to put some thought into hard questions. Do stock options, stock offerings, and acquisition dilutions of stock increase future profitability by a greater proportion than the number of the shares outstanding has been increased? Or did the speculative stock market valuation of a company already include the assumption that the company would expand in certain directions, and therefore is the dilution not increasing the profitability more than the number of shares outstanding has been increased?

An investor might decide to treat some forms of dilution differently from others. The companies themselves certainly will. One interesting feature of stock options, for instance, is that the company actually gets some cash when they are exercised. Typically, of course, no employee would exercise stock options unless the market price of the stock is well above the exercise price (which is how much the company will get at the time of exercise.)

In my case, my first-year salary plus signing bonus was $60,000. I was granted 1,200 options exercisable per year of employment (not adjusted for splits). So, interestingly, the company knew that if I ever exercised those options, it would receive more cash (about $115,000) at the time of exercise than it had paid me the entire year in salary and benefits, and even office expenses. In a sense, then, that's definitely better for the company than just plain having that stock diluted without getting the cash. On the other hand, I sold that same stock on the market for around $1M; if the company had sold the stock directly, it could have gotten that entire $1M for cash flow.

I think it's a real joke any time internet companies report earnings. Partly it's ridiculous because of the way stock option stock dilution is related to earnings. At many growth companies like Amazon.com, Microsoft, and Yahoo, employee work for far lower salaries than they would earn elsewhere. Investors should probably consider what "earnings" would be if companies paid employees in real dollars showing up as business expenses rather than hiding the expenses in diluted stock, which does not count against earnings.

Another convenient device employed by Amazon.com and probably others is investments in other unprofitable internet companies. We may never know to what extent it was considered by Jeff Bezos, but one effect of these investments is hiding R&D expenditures from Amazon's reported losses. Like coca-cola, Amazon does not report consolidated losses (including its share of subsidiary losses). So although Amazon's ownership of drugstore.com loses many millions of dollars of earnings and dilution, it shows up on Amazon's balance sheet as an unrecognized gain now that drugstore.com has gone public.

When it comes down to it, I'm not particularly concerned that internet companies are losing tremendous amounts of money or that they under-report their losses. I just think investors should be aware of it. Many people have gotten rich off the internet companies. I've done fairly well. And many more will. But those people may not be primarily the holders of common stock.

When I consider investing in internet companies, I carefully consider what is the likely market share, market size, and profit margin in ten years, compute the reasonable valuation then, estimate the likely dilution per ar, and add the dilution to the desired rate of return to find the price I'd pay for a stock.

For Amazon, I estimate the market share in 2010 as $25B of 10%-net-margin products (goods sold by Amazon directly), $100B of 2%-net-margin products (goods sold by others through Amazon, and $2B of net 70%-margin-products (selling advertising and customer information). That's $6Billion. Given a P/E of 15, that's $90B. I want a 20% return; plus I assume 5% dilution average per year. So 90 x (.75)^10 = $5B. That's $17 per share. Since great companies like Amazon may be overvalued eternally and may sell for a P/E in the 25 range; ad since I don't need a 20% return on my whole portfolio, I might invest 1% of my portfolio in Amazon stock but I wouldn't invest 5% until it hits 17 or lower.

One of the final amusing developments of late has been Amazon's recent announcements about adding tabs to its homepage in exchange for money-losing internet companies. It's altogether reasonable that Amazon could consider opportunities like this but I do find it amusing how they conveniently are framed to show up as earnings on Amazon's balance sheet. In my mind, the quality of those earnings are rather of a different character than advertising dollars from a cash-flow-positive, established company with actual earnings.

"Paris" background

When I began working at Amazon, the company was finishing the development of its music store and beginning its international development. This chapter is about the continuation of the further multi-product development at Amazon.com and how it precipitated my last project at Amazon, which for the purposes of this book I'm calling Paris.

Amazon has always had the wisdom to launch new projects gradually. Its first new product, music, was one such launch. First, amazon prepared by getting its distribution centers set up for CDs and with new Organizational Procurement Paradigms software, its Distribution Centers (DCs) stocked.

Once the DCs were ready to begin shipping, Amazon added a search box to its home page. This wasn't the full-blown music store; it was just a search box which could be used for album titles. Amazon could then have a month to perfect the distribution center process and stock up on popular titles while the website software and content teams were finishing the official music store. When everyone was ready, we launched the official music store. It was an exciting time at Amazon.com. There was a short party on the first floor of the Columbia Building; the stock price went completely ballistic; and the excitement about Amazon sent demand for books and music right through the roof. Everybody with any free time working at Amazon spent it down at the warehouse picking, packing, and shipping books and Cds. The total customer demand jumped around 50% including all products, as did the stock price. Neither one came back down. Interestingly, the three biggest bumps in the customer demand at Amazon, as well as the stock price, were the IPO, the Christmas rush, and music launch. It seemed that the popularity of Amazon.com drove people both to buy the stock and to buy books from the company. Us software engineers sometimes made charts comparing the daily stock close with the daily revenue from the website. Eery match!

It was extremely confidence-building for everyone at amazon how quickly Amazon dominated online music selling. By the end of our first full quarter in operation, we had beaten out CDNow and Music Boulevard and were the biggest music store on the net.

Throughout the spring and summer of 98, while most of Amazon was hard at work on making music launch happen, and international launch happen, our valiant CEO Bezos was trying to figure out what strategy to use to make Amazon.com the best place to buy anything online. For a few months of this time, I was sitting in one of the cubicles right outside of Bezos's office. Aside from his famous laugh, I also heard him discussing possible acquisitions with management or on the phone. None of the possibilities I heard him talking about actually happened. (Wouldn't it be funny if he were purposely trying to leak names through the employees in the hall!) Soon I was moved to an office elsewhere in the building. Nevertheless, it was quite clear that Bezos was seriously investigating a number of possible new product directions for Amazon.

By the fall, Bezos was clear with the company about future directions. At the all-hands meeting, he said we would definitely be moving into video, toys, and consumer electronics. Many of us already knew about video. Toys and consumer electronics (or CE as we came to know it) would be "1999 deliverables".

"Barnes and Noble is no longer even a competitor to us" he said. He also explained that we would be tackling toys before CE because there was already a major competitor in the toys space versus no major player of our caliber in CE. (It turned out that they were launched together.) Finally, he explained that amazon's strategic vision was to move towards selling the top products in every product area and provide links (for a fee) or host items (for a fee) for the rest of the items a customer might want. I as actually pleased with his impressive treatment of leading us to be THE ecommerce destination. "There are two possible routes which can be effective as we pursue these new products with business partners: either we can be fully-responsible for the customer's experience, and their contact to guarantee satisfaction;or we can make it extremely clear to the customer that we are NOT responsible for the business partners and are only hosting the link or product as a convenience. Anything in between these two clear possibilities becomes a real problem. We don't know yet which of these paths we may end up taking with which partners, or how any of this will shake out, but this is what I'm working on."

"As for competition," he said, "I believe there will be several huge players in the ecommerce space we're engaging in. We will be one of them. We will definitely not be the only one. And also note this: as bug as most people believe the internet will be, is a complete underestimation. It is going to be huge beyond what anyone on earth can imagine. In my estimation, it will ultimately represent somewhere in the low double digits of the percentage of all retail sales. And if you do some research to find out the size of the world retail market, you'll know what I'm talking about.

"When you think about Amazon.com, remember that we are first and foremost aiming to be the world's most customer-centric company. We aim to personalize our store entirely for every customer in ways we cannot imagine yet. W aren't planning to concentrate on primary books forever, we'll concentrate on whatever customers demand. Customers may end up leading us away from intellectual property, or they may continue to use us primarily for intellectual property.

"Let me tell you about one of my favorite stores. It's called FRY's ELECTRONICS. And if you're in the bay area, you should stop there and check it out. When you think of an electronics store, you probably think of a store with lots and lots of electronics. Only electronics. And Fry's has that. But it has a lot more too. It has rows and rows...

"Snack foods, like corn chips and sodas. And it has books about electronics and computers as well as the electronics and computer software and hardware. You see, Fry's isn't jest an electronics store but a store with everything that computer or electronics people would regularly need. You computer people like these chips and sodas! That's what Amazon.com can be for every customer - the most personalized store imaginable, the most customer-centric company on Earth!"

That was all he gave us at that time. But that was a lot.

But soon thereafter, he gave us the Junglee acquisition; though he didn't explain it per se, it was clear Amazon was definitely pursuing multiple products in multiple ways.

Junglee at the time was a private silicon-valley company developing comparison-shopping technology for the internet. They provided product search technologies for a variety of internet companies as well as a jobs database. After the acquisition, Amazon discontinued Junglee's outside contracts and sold its jobs database, putting all the Junglee employees who moved to Seattle to work on a new project. This was late in the summer of 98, at around the time international was preparing to launch. I had been at Amazon almost six months, but it felt like forever.

Although some of Junglee's computer programmers elected to stay in the Bay Area rather than move to Seattle, and therefore had to quit the jobs they were offered at Amazoncom in connection with the acquisitions, a majority of them moved to Seattle, and were put to work quietly on the fourth floor of the Columbia building. Their project was kept quiet by and large, though from time to time we Amazonians learned that the new project was some kind of multi-product shopping database.

As late summer wore into the fall of 1998, Amazon morphed into the secretive, strategic company I loved to hate. And unfortunately, I was moved into an office further from Jeff Bezos's so I no longer had the convenience of hearing his discussions when he left his door pen. What exactly were the Junglee folks working on, and how did it fit into Amazon's goal to be the best place to go to buy anything online?

At the next quarterly all-hands meeting in December 98, we learned the answers to some of our questions. One of the ex-Junglee launches presented their new project: "Shop the Web". "Shop the web" had the look and feel of Amazon.com: a "detail page" for each item; a search page; a search results page; a browsing page. It was a really impressive piece of work. Junglee had some very interesting technology employing pseudo-SQL over a virtual database including the internet. Their initial product launch for Amazon.com included merchandise from only a few dozen select merchants who uploaded link and produce data to the shop the web service on a regular basis.

With this original shop-the-web service, as presented that day, Amazon would receive a referral bonus for any products purchase at a partner's site through Amazon's shop-the-web. Interestingly, years before, Amazon had pioneered a web-based referrals program called Associates where Amazon paid associates a percentage for any sales at Amazon resulting from links at an associate's website. Now the tables were turned and Amazon was going to be receiving the sales commissions. One final twist about the launch of shop-the-web was that it used very different software and hardware from Amazon's current systems - and it wasn't entirely clear what the new kind of hardware and software could handle the website traffic Amazon experienced regularly.

In the second week of December 1998, after a week of internal beta testing, Shop-The-Web was launched. At first, there was simply a small link to Shop-The-Web, and the link appeared near the bottom of Amazon's main page. Soon, we hoped,it would move to a tab of its own in the main store as had been explained at the all-hands meeting. It was an impressive schedule for a group of new employees at Amazon, but due to how late in the year the Junglee acquisition had happened, the launch date was embarrassingly too late for a smashing Christmas season.

Also launched around this time was a 'gifts' store with its own tab and homepage, featuring a few select products like palm-pilots, plush toys, and GPS machines (all stocked through only one warehouse).

But for some reason, Shop-the-web never got its own ta. Maybe it was because all the software engineers and project managers were down at Amazon's Seattle warehouse packing books, music, and videos in he Christmas rush. Maybe it was because the software or hardware didn't perform adequately in some cases. Maybe it was because much of amazon's senior management was off-site, knocking heads on the possibility for adding a half-dozen new warehouses and several new product categories in 1999. maybe it was because Jeff Bezos was uncomfortable with the inability of shop-the-web to catalog the entire internet, or the inability of Amazon to host product pages for business partners or it was too vague what the relationship between Amazon and those other companies was. Maybe Jeff Bezos just didn't like the name. In any case, Shop-the-web never got its own tab at Amazon.com.

Amazon.com entered 1999 with a small shop-the-web link on its home page and a gifts tab in the navigation bar. I began to worry that the company was getting unfocused in its secrecy and stupid in its tactics. Meanwhile, Yahoo! Went shopping. Yahoo bought a web company whose software enabled anyone to set up its own product store on a centralized, hosted website. Together with Yahoo's auctions and massive customer base, this threat was obviously a huge obstacle and competitor to Amazon's ambitious goals. I would have thought that Amazon's management would have been imperative about the need to race to the forefront of this space. But all we heard from management was to make a calm insistence to software to make our software more mature; to fix up any things which were left hanging from Christmas, international launch,or (God forbid) music launch. This internal housekeeping effort, described by senior management at the December all-hands meeting as "Gohio" (Get Our House In Order) had at its heart a desire o easily add new products, warehouses, or countries to Amazon, by having mature, configurable software. But it was not an effort to launch a bold new emerchandising direction. As Yahoo's site became more and more impressive, I began to wonder whether Jeff Bezos had entirely lost his head. The rest of Amazon truly seemed to need his leadership; everyone was chanting the mantra "Gohio" but we were missing something. What in the world was the company doing at this stage - moping that sales weren't picking up as quickly after xmas as they had before?

In early February, my new software team had been working on a ubertool for inventory management for about a month. I was nearly finished with my work in the manual ordering tools and on documenting the last of the important programs I had written in the automated procurement tools suite. I gave a presentation to the members of the Organizational Procurement Paradigms team on the active procurement system and handed over the reigns to the new tactical team.

Then I heard form the co-leaders of my new team that our new project had been put on hold. We were immediately to begin work on a new, mission-critical project to be used in several strategic projects, the nature of which was yet unknown. "he project is direct from Jeff Bezos," my new Captain reported.

The new project was going to be some kind of architectural backbone to be used by important new strategic advances at Amazon. Strangely, although my co-leaders were glad to explain that the new project was some kind of website-backbone-file-transfer thing, they were quite apprehensive at first about whether to tell me and the other project member any more details about the projects. Within a few days, that became ridiculous.

So that is the end of my notes I wrote two years ago (a year after leaving Amazon). There's then three months of torture which I didn't write about. My team was poorly-managed; and the whole company was pretty hosed. As a result, we were working on a stupid project which was not really necessary for the launch of the web-shopping site to compete with Yahoo's.

It turned out that there was a reason why there was so little direction. There was a major secret effort involving all top management and about 15 software people to launch a new site for Amazon: auctions.

I think it was a mistake for them to work on the auctions instead of the store-building software; but these things are always easier in retrospect. In any case, Yahoo kicked our butt to that new market, and Amazon never caught up. The auctions turned out to not give Amazon much of a market share in that space either, too bad - because actually our site was better than ebay's - we just didn't have the history.

I had a confrontation with one of the managers from that secret project and it really pissed me off. I was glad his "strike price" for his options was so high that never made as much from Amazon as I did.

And as for me, Auctions was good for my pocketbook. When Amazon launched the auction site, the stock price went up again. It did start to come down before my options vested and I was able to sell my stock, some. But I probably got an extra hundred grand thanks to the launch of auctions. Thanks guys.

I believe that I was able to impact Amazon's bottom line by at least $1 million per year due to my nw tools. Just that figure alone would have paid for my stock options and salary, which cost the company shareholders $900,000 of stock dilution. Considering the other achievements I had at Amazon, I was most likely worth the investment they made in me from a cost-benefit analysis. It was a revolutionary time in business, when it was possible for individuals to make a huge difference like that in the company. If I had been able to continue my work on the manual ordering processes, I probably would have been able to save the company at least an additional million dollars a year, with my next 6 months of work. Although I don't think impact on the bottom line of the company was less than average, I don't think it was too exceptional either. I think many software engineers, as well as other employees within the company, made the same magnitude of contribution. I was never the ideal employee in my manager's eyes; he was never my ideal manager either. Yet nonetheless it was possible to make substantial progress. On the other hand, the personal and political issues which I experienced and witnessed did, I believe, have a huge impact in Amazon's failure to be the best place to buy things on the internet.

Here's how it all came to a head. Finally, I reached my anniversary date. I sold half of my Amazon.com shares. It was a bit depressing because our quarterly conference call had been three days before, and it was the first one that wasn't heads-and-shoulders way above expectations. It was the first one where Jeff Bezos had sounded like he was trying to sell the company with good tidbits, rather than letting the conference call participants ask him for details. It was the first one where he maybe seemed scared and a bit defensive and seemed ready to try to make promises. So because of that, the market started dumping Amazon stock, and it went down about 30-40% before I (and others at Amazon) were allowed to sell any of their shares (there's a special trading window starting about 3 days after each quarterly conference call when for the next 3 weeks or so you can buy or sell the stock of the company you work for). I sold all the rest of my shares by the end of the month. All told, I ended up with somewhat over half a million to keep, plus about $300,000 which I would have to return to the federal government in income taxes.

After selling my shares, I kept working at Amazon for a bit. But I really didn't want to. I secretly hoped the stock would completely crater after I had sold, and that I would have an easier decision to quit. It did keep going down, but it didn't crater.

What I did was that I requested through Ben Saran that I be granted a 9-month leave of absense from Amazon. He reportedly asked up the chain to Hal Jeffries and possibly some people in the personnel departments, but it was impossible. 3-months was the longest permissible leave, and they woulnd't grant that to me in any case because that was only for special circumstances. So I turned in my letter of resignation to Ben, Hal, and Rick Dalzell.

I was one of the first people of the 'middle wave' to leave. There had certainly been founding-round employees who had left, many of them very rich, far richer than me.

I got calls from people all over the company who didn't want me to leave. Hal called; we had a pretty terrible conversation in which I really let rip with everything I was pissed about - I couldn't trust him, the whole team knew he couldn't be trusted; he hadn't given me the vacation he'd promised; etc. "Well I guess this conversation's over then," he said and hung up. It was the wrong way to end it there.

I got a call from Rick Dalzell; he didn't want me to leave - I suspect partly because he respected my work in particular and knew other people did too, and partly just because he didn't want to have the precedent of people leaving. I told him I had been working extremely hard for a long time, and needed a lot of time off, in particular, 9 months. He said "Peter, some people have taken leaves - but they end up coming back (and gave some examples) - and they all say that the decompression time can be much shorter than the compression time." We talked a bit about my frustrations with Hal, with the job in general. (Rick later emailed me with the message 'We wish you all the best that life has to offer.' - I have not fogotten that a bit, it is very much still appreciated.)

I also got offers from 3 other departments in the company to join their teams instead of leaving. The first one to call, I said that if they would give me a 3-month leave of absense immediately, I'd join their team then take it - they talked to Rick Dalzell, and gave me the offer, and I took it.

So officialy I took 3 months of leave. Again, I hoped the stock price would plummet to the ground so it would make my deision easier. I moved to Olympia, WA with my new girlfriend. The stock price didn't plummet.

So After the 3 months, I went back and met with that group. It turned out that Hal had emailed some kind of scathing letter to them, and said that he and other people in the group (including Angus, Brendan, and Chris - and maybe others, I don't know) thought I was not a good person to have onboard at Amazon. I was kind of surprised, but I guess he had been pretty hurt and wanted to hurt me back. Actually, the truth is, it did hurt me back emotionally. Everybody would rather be appreciated than disliked. They asked Brendan and Chris what the story was (I didn't think they would have agreed to what Hal had supposedly proposed) but supposedly they both agreed. So I guess I had pissed them off quite a lot too. (I have a hard time acknowledging the possibility that I was nota good employee for Amazon - and I guess half the reason I've written this whole diatribe of notes is a kind of poor-me book jutifying what a great guy I am and how much I did for the company and how these guys were wrong.) Later, at dinner with Angus, I asked him whether he knew about the list or agreed with Hal that he would be on the list; he said he had never known about it and disagreed with the idea that Amazon was better without me. I actually think he was telling me the truth. But whatever.

But the truth is, I know I still could have gotten to work for any of the other groups, returning within the 3-month leave period, and still got to keep my options. (The big deal with 3-months and the options was that if I returned before that time was out, then I would still be entitled to vesting options on my next anniversary - options which if the stock stayed where it was, would have been worth about another half-million to me After taxes.) I did go talk to one of the other groups, and I said to them that I was interested in working for them only if I could work 2 days/wk in Seattle and the other 3 days/wk from home in Olympia. They said it was impossible (which I knew it was). It was kind of my way of allowing my last little ego-hope for getting another half-million and also keeping my sanity. They did say that they were interested if I would go 5 days/wk in Seattle, and I'm actually sure I could have achieved the same thing with several other groups.

But the truth was, working at Amazon wasn't the right thing for me. I had earned a half-million dollars After all taxes. That was enough for me to probably live on forever, as long as I lived cheaply. And I figured that chances are, I would probably get enough investment income from the money that I would end up getting the whole million in the bank within about 5-10 years. So - what was better, a miserable year at Amazon, at the conclusion of which there was a CHANCE of having a million dollars (or the stock could have cratered After 11 months and left me with nothing) - OR - 5-10 years of pleasurable existence, at the end of which I would probably still have the million dollars? ESPECIALLY since I believed the stock was overvalued by about 400% and that it seemed like the management and stock market were figuring this out !! -

Well, I decided to quit. And not go back. So goodbye Amazon.com and it was on to a new life for me. Which I've now been living for a while. It took a lot of strange turns adjusting to my new life, but here I am. And although Amazon's stock did NOT crater in that year (it waited 18 months), I think I made the right choice.

By leaving when I did, the psychic patterns of iving for the future, of anger, resentment, secrecy, they did not build up to the point where they took over my psyche. Yes, they're part of me. But they're not all of me. And so in one sense I am still "doing a bit of work" for society to keep myself at peace in exchange for the investment income that experience continues to provide me.

Maybe six months later, I decided to call Brendan and Chris and ask their opinions concerning my performance at Amazon. Here's what they said.

Brendan Reid: he said that he felt I'd originally come on at a convenient time when Hal was able to invest in new developers but that now Amazon couldn't do that any more and instead it had to hire people with more experience. He confirmed that he wouldn't have recommend Amazon that it hire me at the time I took leave.

Chris Jenkins. I asked "were my issues severe enough so that you felt they outweighed my contributions, and that Amazon would be better off without me?" He aid it's a close call but now; he said I do however have a chip on my shoulder and have a hard time receiving feedback.

It reminded me of strange flaw in Amazon's internal review process. We were all supposed to be reviewed by several of our self-selected peers, as well as our managers. I did ask some people to review me (including Brendan, but not Chris or some other people I knew were threatened by me) - and then later Hal told me that I had gotten a lot of peer reviews. Perhaps Chris and other people had just decided to review me of their own accord.

In that whole experience, I learned a whole lot. I learned I have tremendous potential in a million directions. I learned a vast amount about computers, companies, working with people. I also learned that most people will be threatened by my successes, and that if I ever do anything to hurt anyone, they will work hard to hurt me back harder. And I learned that while I can choose any life I want in this world, and can achieve/do almost anything I want, I can never take things back.

So - onward.!