YC
A Conversation with Werner Vogels
@ Y Combinator
11/12/18
Takeaways
WV
Werner Vogels
11/12/18
@ Y Combinator
I was an academic before I joined Amazon, working as a research scientist at Cornell for 10 years, focusing on large-scale distributed systems and doing startups on the side.
WV
Werner Vogels
11/12/18
@ Y Combinator
I realized that Amazon is not just a retailer; it's a massive technology operation, operating at a scale I had never seen before.
WV
Werner Vogels
11/12/18
@ Y Combinator
When I joined Amazon, I wanted to bring more academic rigor to scaling, as we needed to revisit our processes and technologies to handle significant growth.
WV
Werner Vogels
11/12/18
@ Y Combinator
At Amazon, we focus on performance and measurement, understanding that improving page latencies can significantly increase conversion rates.
WV
Werner Vogels
11/12/18
@ Y Combinator
Building independent teams is crucial for fast-moving companies; we aim to reduce hierarchy and empower individuals to take ownership of their work.
WV
Werner Vogels
11/12/18
@ Y Combinator
Amazon's culture is driven by 14 leadership principles that emphasize customer obsession, ownership, and a deep dive into data.
WV
Werner Vogels
11/12/18
@ Y Combinator
Hiring for culture fit is essential at Amazon; a bad hire can disrupt small teams significantly, so we focus on small teams of 10 to 12 people.
WV
Werner Vogels
11/12/18
@ Y Combinator
Managing teams is a completely different discipline than focusing solely on technology; a VP of Engineering prioritizes team effectiveness, while a CTO focuses on the right technologies and tools.
WV
Werner Vogels
11/12/18
@ Y Combinator
In the early days of Amazon, we faced failures with companies integrating due to infrastructure issues, which led to the creation of AWS.
WV
Werner Vogels
11/12/18
@ Y Combinator
Amazon's architectural changes in the late 90s aimed to scale quickly, but resulted in a brittle monolithic structure that couldn't sustain growth.
WV
Werner Vogels
11/12/18
@ Y Combinator
Transitioning to a service-oriented architecture involved breaking down the monolith into independent services, which took two to three years to implement due to the lack of historical precedents.
WV
Werner Vogels
11/12/18
@ Y Combinator
The effectiveness of engineers at Amazon dropped as deployments slowed, primarily because database management was centralized and controlled by a conservative group of DBAs.
WV
Werner Vogels
11/12/18
@ Y Combinator
To improve innovation, we moved to a microservices architecture, allowing teams to manage their own data stores and reducing reliance on centralized resources.
WV
Werner Vogels
11/12/18
@ Y Combinator
Despite the benefits of microservices, we noticed a decline in team effectiveness as they became bogged down with shared infrastructure management.
WV
Werner Vogels
11/12/18
@ Y Combinator
AWS's rapid growth surprised us; we anticipated it would be significant, but the speed of adoption exceeded our expectations.
WV
Werner Vogels
11/12/18
@ Y Combinator
At Amazon, innovation is driven by teams responding to customer requests, with about 95% of features developed based on direct customer feedback.
WV
Werner Vogels
11/12/18
@ Y Combinator
When launching new products, we focus on a minimal feature set to ensure reliability, then iterate based on customer usage and feedback.
WV
Werner Vogels
11/12/18
@ Y Combinator
With AWS Lambda, we introduced a serverless environment that allows developers to focus on code without managing servers, changing the development landscape significantly.
WV
Werner Vogels
11/12/18
@ Y Combinator
Customer feedback has reshaped our product roadmap, emphasizing the importance of access management over previously anticipated features like secondary indices.
WV
Werner Vogels
11/12/18
@ Y Combinator
We can't treat a minimum viable product (MVP) like an MVP because businesses will depend on it, which requires a different culture and structure.
WV
Werner Vogels
11/12/18
@ Y Combinator
Last year, we launched 1,400 new features and services, and as the number of teams grows, the pace of development accelerates.
WV
Werner Vogels
11/12/18
@ Y Combinator
Each team at Amazon has its own customer set and is expected to understand their needs, which fosters a fast-moving environment for software development.
WV
Werner Vogels
11/12/18
@ Y Combinator
The 'working backwards' process at Amazon starts with writing a press release for yourself to clarify what you're going to build, followed by answering the 20 most frequent questions about the product.
WV
Werner Vogels
11/12/18
@ Y Combinator
At Amazon, we focus on developing only those things that truly matter for our customers, avoiding the risk of engineers taking charge and building technology instead of products.
WV
Werner Vogels
11/12/18
@ Y Combinator
In meetings, we avoid using slides and instead read a six-page narrative document in silence for the first 30 minutes to ensure everyone is on the same page.
WV
Werner Vogels
11/12/18
@ Y Combinator
Security must become a priority for everyone in a company, from the CEO to engineers, as protecting customer data is essential for business success.
WV
Werner Vogels
11/12/18
@ Y Combinator
Startups often make the mistake of treating AWS like a traditional data center, which limits their ability to leverage AWS's higher services and security features.
WV
Werner Vogels
11/12/18
@ Y Combinator
There are two types of startups: those that aim for rapid growth and acquisition, and those that focus on building sustainable businesses over the long term.
WV
Werner Vogels
11/12/18
@ Y Combinator
It's crucial for startups to understand their identity and approach, whether they are mercenaries seeking profit or missionaries passionate about their product.