Why does Jeff Bezos recommend scaling failures and watching science fiction

Jeff Bezos, Amazon CEO, wrote an annual letter to shareholders , and volunteers from the RUSSOL start-ups school and Y Combinator's initiatives translated it in Russian to answer these questions and inspire readers to create their own Amazon

Jeff Bezos, CEO of Amazon. Photo source: flickr.com

Jeff Bezos, April 11, 2019

Our shareholders.

Something strange and important has happened in the last 20 years. Take a look at these numbers:

1999 3%
2000 3%
2001 6%
2002 17%
2003 22%
2004 25%
2005 28%
2006 28%
2007 29%
2008 30%
2009 31%
2010 34%
2011 38%
2012 42%
2013 46%
2014 49%
2015 51%
2016 54%
2017 56%
2018 58%

Percentages show the share of gross merchandise sales on Amazon by independent third-party sellers. And these are mainly indicators of small and medium-sized businesses (the rest is from Amazon’s own retail sales). The sales of independent sellers rose from 3% to 58%.

Third-party sellers nehilo so kicked our ass, the main seller

This is a high bar because our primary business has grown substantially: from $ 1.6 billion in 1999 to $ 117 billion last year. The cumulative average annual growth rate (CAGR) of our business for this period was 25%. At the same time, third-party sales increased from $ 0.1 billion to $ 160 billion - in their case, the cumulative average annual growth rate was 52%. To understand the scale: over the same period on eBay, with an increase in gross sales from $ 2.8 to $ 95 billion, CAGR growth was only 20%.

Why did third-party sellers on Amazon have an order of magnitude better than eBay? And why were their growth rates much higher than those of Amazon itself, with its level of sales organization? There is no definitive answer to this question, but something will help us to get closer to it.

We helped independent salespeople compete with our own business by offering them the best sales tools we could create. These are opportunities for retailers to manage inventory, process payments, track shipments, generate reports, and sell overseas — and we complement them every year. Great value and in the complete portfolio (complex operations from order to delivery) and the Amazon Prime client program. Together, these two programs have significantly improved the quality of services received by the client when buying from third parties.

Now that these two programs have become normal practice, it is difficult for people to fully understand how revolutionary they were at the time of launch. By investing in them, we took great financial risks. All the time while we were testing different ideas, the project required investments. We could not predict with certainty what these programs would take, but we moved forward, relying on the intuition and prompting of the heart, helping us to realize these projects with our optimism.

Intuition, curiosity and the influence of wandering

Since the creation of Amazon, we knew that we wanted to create a culture of creators - a culture of curious people and researchers who like to invent. Even as experts, they are open to new experiences and think like newbies. They look at things as temporary. A creative worldview helps us to reach out to great and difficult opportunities to solve. And we are convinced that the path to success lies through iteration: invention, launch, rethinking, restart, repeat - again and again. Creators know that success can be achieved along a winding path.

Sometimes (or rather, often) in business, you know where you're going, in which case your actions will be effective. Develop a plan and follow it. Wandering in business is inefficient, but it is also not accidental. The search is guided by premonition, intuition, curiosity and feeds on the deep conviction that the benefits that customers will receive will be large enough to deviate a little and wander on its way. This wandering is an important counterbalance to efficiency. You need both. Large-scale discoveries, knocked out of the general "line" is likely to require such wanderings.

Among the millions of Amazon Web Services (AWS) customers, there are startups as well as corporations, state-owned companies, and non-profit organizations. Each of them wants to create the best solutions for their users. We spend a lot of time figuring out what these organizations and their employees want — developers, operations managers, digital directors, information security directors, etc.

AWS, created by us, is largely due to the wishes of customers. It is important to ask the customers what they need, listen carefully to their answers and plan for thoughtful and fast (speed is important in business!) Work. No company can succeed without making a client a priority. But this is not enough. Needs play a key role, but the customer does not know how to explain them. We must invent by putting ourselves in their place. Use your imagination to understand what customers may need.

The AWS service as a whole serves as an example. Nobody asked us to run it. No one. But it turned out that such a service turned out to be necessary and in demand, just nobody knew about it. We had guesses, we took financial risks and began to create - to process, experiment and try countless times.

In the case of AWS, the same scenario occurred several times. For example, we invented DynamoDB, a highly scalable key-value database with a low degree of delay, which is now used by thousands of service users. Carefully listening to customers, we heard that companies faced the limitations of their database providers and worked for decades with uncomfortable solutions: expensive, patented, rigidly linked to the supplier and penalties.

We spent several years building our Amazon Aurora database management engine, fully managed by MySQL, compatible with PostgreSQL. This is a service with stability and access conditions not worse or even better than that of commercial mechanisms, which, at the same time, was 10 times cheaper. No wonder our solution worked.

But we were also optimistic about specialized databases. Over the past 20–30 years, companies basically solve their problems using relational databases. Due to the fact that the developers are very familiar with relational databases, they were used everywhere: where it is possible and impossible. Although the database sizes were not optimal and the request processing speed was large, you could get them to work.

But today, many applications need to store huge amounts of data - terabytes and petabytes. Changed and requirements for applications. Now applications are developing in a short waiting time, processing in real time and the need to process millions of requests per second. And this is not only key-value storages like DynamoDB, but also in-memory databases like Amazon ElastiCache, time series databases like Amazon Timestream, registry solutions like Amazon Quantum Ledger Database. All of these are tools for efficient work, saving money and moving your product to the market faster.

We also teach companies to use machine learning technologies. We have been working on this for a long time, but our attempts to make some of the internal machine learning tools available have not been successful. It took years to search - we experimented, tried different versions, refined, listened to valuable insights from our customers before we received SageMaker, which we launched a year and a half ago. SageMaker at all levels struggles with congestion, complexity and movement at random - this is a democratic version of artificial intelligence.

Today, thousands of our customers are building machine learning models in AWS using the SageMaker tool. We continue to improve service - including by adding new learning opportunities. Optimized learning implies a steep learning curve with a large number of moving components. Until recently, such a tool was available exclusively to technically advanced organizations with good funding. Changes were made possible by a culture of curiosity and a willingness to try new things when we put ourselves in the place of our customers. Customer-oriented searches have received a market response - AWS is now estimated at $ 30 billion a year and continues to grow rapidly.

Introducing the unimaginable

Amazon today is still a small player in the field of world retail. We occupy less than 10% of the retail market and in each country where we operate, there are retailers with a large part of the market. And, basically, this is due to the fact that 90% of sales are still ordinary offline stores. For many years we thought that we could have done for people in regular stores, but it seemed to us that we had to first invent something that could really delight consumers. And with Amazon Go, we had a clear idea: to get rid of the worst that is in offline retail - from the queue at the cashier. No one likes to stand in line. Instead, we imagined a shop where you can go, take what you need, and leave.

It was not easy to come to this. Technically difficult. This required the efforts of hundreds of talented, motivated computer specialists and engineers around the world. We needed to design and produce our own cameras and shelves, and create new computer vision algorithms, including the ability to form an image, combining data from hundreds of connected cameras. And it had to be carried out in such a way that the technology moved away into the shadow, became invisible. The reward for this was the feedback from customers who described the Amazon Go shopping experience as a “miracle”. We currently have 10 stores in Chicago, San Francisco and Seattle, and we look forward with enthusiasm.

Failures must scale too.

With the growth of the company, everything should scale, including your unsuccessful experiments. If the scale of your failures does not increase, you will not be able to create something weighty enough to move forward. For a company of the size of Amazon, experimenting at the right scale means periodically suffering multi-billion dollar losses. Of course, we will not experiment so uncontrollably. Naturally, we will work hard to make our bets play, but in the end not all bets are winning. Taking on such large-scale risks is a part of the services we provide to customers and the public. The good news for shareholders is that even one such successful bid covers the cost of failure.

Shot from movie Star Trek. Photo source: redshirtsalwaysdie.com

The development of the Fire phone and Echo began at about the same time. And although the Fire phone was a failure, we were able to use our experience (and programmers) and strengthen the development of Echo and Alexa. On Echo and Alexa we were inspired by a computer from Star Trek. The idea also originates from two other areas where we have been building and experimenting for many years: machine learning and cloud services. From the very beginning of Amazon, machine learning was a key component of our recommendations to users, and AWS gave us the opportunity to be at the forefront of cloud usage. After many years of development in 2014, we released Echo with integrated Alexa, who lives in the AWS cloud.

None of the clients asked Echo. This was an example of our “wandering”. Market research will not help here. If you came to a client in 2013 and asked: “You would like to have in your kitchen always a working black column the size of a jar of Pringles chips that you can talk to, ask her questions, and that can turn on the lights and play music ? ”I’m sure they would have looked at you strangely and replied,“ No, thanks, no need ”.

Since that first-generation Echo, customers have bought over 100 million devices supporting Alex. Last year, we improved Alex’s ability to understand queries and answer questions by 20%, adding billions of facts to improve Alex’s knowledge. The developers doubled the number of Alex's skills to more than 80,000, and customers talked to her tens of billions more in 2018 than in 2017. The number of devices with Alex in 2018 has doubled. Now more than 150 different devices are available on the market with Alex, from headphones and computers to machines and components of a smart home. And there will be even more!

And the last thing I want to say. As I said in my first letter to shareholders over 20 years ago, our focus is on attracting and retaining versatile and talented employees who think like owners. Investing in staff helps to achieve this, and, as in many ways at Amazon, not only statistics are needed here, but intuition and heart.

Last year, we raised our minimum hourly wage to $ 15 across the US (for all staff, hourly, temporary, and seasonal workers). This jump in pay has affected more than 250,000 Amazon employees, as well as more than 100,000 seasonal workers who have worked at Amazon centers throughout the country on past holidays. But it was not the main thing for us in making this decision. We have always offered competitive pay. But now we have decided that we need to be a leader - to offer payment without regard to competition. We did it because we think it is right.

Now I offer this challenge to all competitors in the retail sector (they themselves know who they are) to respond to our offer of employee remuneration and minimum wages. Let's! Make $ 16 and answer us. Such competition will benefit everyone.

Many other programs that we offered to employees were also from the heart, and not from the head. I mentioned earlier the Career Choice program when we pay 95% of the total costs for obtaining a certificate or diploma in selected areas so that our employees receive a demanded education, even if this education takes them away from the company. Over 16,000 employees took advantage of a program that continues to grow. Also, our Career Skills program allows hourly workers to gain knowledge in writing resumes, communications, and basic computer skills. Last October, we signed a Pledge to America's Workers, initiated by the President of the United States, and committed ourselves to improving the skills of our 50,000 workers.

Our investments are not only with our current employees. To prepare the future workforce, we invested $ 50 million, including through the Future Engineer Program to support education in the exact sciences and computer science in primary and secondary schools, as well as in universities, with a special focus on involving girls and minorities in data professions. We also continue to use the potential of our veterans. We are on track to fulfill the commitment to hire 25,000 veterans and wives of the military by 2021. With the help of a special program, we provide them with on-the-job training in a field such as computing in a cloud environment.

I would like to thank our customers for giving us the opportunity to work for them, while encouraging us to become better, our shareholders for their continued support and all our employees around the world for their hard work and leadership spirit. The Amazon team listens to customers and is looking for something new for them.

As always, I am attaching a copy of our first 1997 letter. It is still Day 1.

Respectfully,
Jeff Bezos

Translators and proofreaders: D. Demidova, CathBa, Yu. Yartsev, I. Zvyagin, A. Litvin

And something else. Here are 4 insights that I saw in the letter of Bezos:

1. Failures and mistakes are important. This is one of the surest ways to learn and grow. No need to worry that if I did not succeed, then I’m a penny price and it’s better not to do it. On the contrary, more systematic approaches with feils lead to higher chances of achieving success.
2. Fiction and especially science fiction should be read and watched by anyone who wants to become a technological entrepreneur or be part of a progressive one. In such literature, you can learn a lot and be inspired, as did the Amazon with a smart column Echo
3. The presence of an open API for any good product always leads to the growth of users. It worked for Amazon, VKontakte, Yandex, etc. In the DNA of the product being created, it is necessary to foresee the use of the platform by third-party developers.
4. If the minority Americans are Mexicans, then we have Tajiks and Uzbeks. Having got rid of prejudices and investing more in them we will get a different look at the problems to be solved and the products created. It is more profitable for countries to have hundreds of thousands of first-class IT specialists, rather than first-class “janitors”, who inherit the profession.

Note from the author of the publication: I, Yuri , the founder of the non-profit start-up school RUSSOL, is coordinating the initiative to translate the lecture course of the start-up school Y Combinator . Together with volunteers and freelancers, we translated and subtitled 36 hours of videos about creating startups, starting with the search for ideas and ending with investments and international sales. Materials of 2018 here , 2017 - there .

The mission of RUSSOL - we want the middle class to grow and develop in the post-Soviet space. And we help start-up entrepreneurs with open educational lectures, meetings, conferences where we tell how to find an idea, create a product, attract money and enter international markets. As part of the initiative, we are starting to translate letters from Jef Bezos to Amazon shareholders.

We are looking for you reader, so that you make a small contribution to the initiative and help translate lectures and letters, make notes, read texts, work on the site on the tilde and blog on Habré. We are also looking for sponsors who are close to the topic and who are ready to support our initiatives. Let's communicate to learn together, learn from experience, grow and benefit your home.