CS183C Session 8: Eric Schmidt
Eric Emerson Schmidt was born in Washington, D.C. and spent his youth in its Northern Virginia suburbs. He was one of…en.wikipedia.org
Online shopping from a great selection at Books Store.www.amazon.com
For today’s class session, Reid Hoffman interviewed our special guest, Eric Schmidt, the chairman of Alphabet (and formerly the chairman and CEO of Google). You can see the video of the class here.
Q: You had a substantive career before Google, both at Sun and Novell. What did you learn that helped at Google.
A: I went to visit my old boss at Xerox PARC, Bob Taylor. He was the person who funded the ARPANET and defined the personal computer. As a young manager, the rule is that you absorb everything. My first experience at Sun colored the next 30 years of development.
The next 5 years are when you’ll learn all the little things about leadership.
Sun was tumultuous and political. Companies can reorganize prematurely, become religious. There was a meeting where we realized we could not match the cost of PCs. That was the signal that we needed to change.
I went to Novell under the mistaken goal of being a CEO. I didn’t do the due diligence, and if I had, I wouldn’t have gone. Our basic goal was to get out with our professional reputations intact and not end up in jail. The books were cooked, and people were frauds. But it turns out you can overcome that, and the skills I developed helped at Google.
I understood the rule of cash, so we did everything we could to manage for revenue, and the rest is history.
Q: Is there anything you’d tell your younger self?
A: Do things sooner and make fewer mistakes. The question is, what causes me not to make those decisions quickly. Some people are quicker than others, and it’s not clear which actually need to be answered quickly. Hindsight is always that you make the important decisions more quickly.
Q: Talk about when you joined Google.
A: These two young men are brilliant, crazy, and unreliable, and we need a CEO that can manage them. Larry and Sergey’s vetting technique was to spend a weekend with the candidates, which meant that none of them passed. When we met, it was clear that we were compatible. I had a list of things they needed to do, and we went and did them.
Almost all small companies are full of energy and no process. My list was straightforward: internationalization plans, sales plans, product plans, accounting, etc. My first meeting at Google was like being at a graduate school full of interesting people with no deadlines or deliverables. Offices at Google had 4 people because Larry and Sergey’s office at Stanford had 4 people.
Q: How did the relationship with Larry and Sergey develop?
A: I was well aware of the John Sculley/Steve Jobs story. Jobs needed someone to be CEO, the board picked Sculley, Jobs and Sculley had a fight, and the board picked Sculley because you have to back your CEO. Steve left, founded NEXT, then came back and within a month, had gotten the CEO who brought him back fired, and replaced the entire board. It was a very Steve Jobs thing to do.
I knew that it was Larry and Sergey’s company, and I acted that way. For example, I never did any press. Right before the IPO, Larry and Sergey did an interview with Playboy — no pictures. It turns out that the interview was at the wrong time in the quiet period, and it put the IPO in jeopardy. “Did we screw up?” The correct answer is Yes. But the even more correct answer is no problem, we’ll fix this. From that moment on, they’ve never given an interview. That was 12 years ago.
When they wanted to do interviews, they did them. Once they didn’t want to do it anymore, I did them.
It’s their company. At Facebook, it’s worked extremely well, with Mark bringing in Sheryl.
Q: Google is at 50,000 people…
A: 60,000. I have a saying about this — it’s easy to double, but it’s very hard to quadruple every year. You can kind of see how doubling works, but quadrupling is much less clear.
Q: What are the key hacks that did and didn’t work.
A: If you didn’t understand the subtlety, you’d assume the answer was to grow everything as fast as possible everywhere. That doesn’t work. The way you build great products is small teams with strong leaders who make tradeoffs and work all night to build a product that just barely works. Look at the iPod. Look at the iPhone. No apps. But now it’s 70% of the revenue of the world’s most valuable company.
Travis at Uber said that he understood scaling, but the app, the product wasn’t ready.
Larry and Sergey would play tricks on me. I’d say we’re not going to take on Microsoft. We’re not going to do a browser or OS. So they hired someone to improve the performance of Firefox. 6 months later, they showed me Chrome. Those a — holes. I knew they were going around me.
So I told them, we can’t do an operating system. So they bought Android and told me it was just for smartphones.
Maybe the lesson is that I’m just wrong all the time. But the secret is that you have to have judgment on when these things scale.
We launched Google Wave, a complicated email product, with a lot of fanfare. Great products go up. Wave just kept declining. It took Eric, great CEO, 18 months to kill it.
You’ve got to have products that can scale. What’s new is that once you have that product, you can scale very quickly. Look at Uber.
My first concern with Google was that everything was a sham. They were using Quickbooks, and I was sure there were errors. So I made them show me the bank account, and they had money. We were making money from ads, but none from Europe. So I told Omid, go to Europe and don’t come back until you set up a sales office. Today, those markets represent 60% of our revenue.
Take Uber. Do you believe that the Uber model scales well outside the US? Ask the French. The people love it, the governments hate it. Now you need two scalabilities— product scalability and government scalability.
Great products happen when people build a product for themselves. Larry and Sergey created Google for themselves. Andy Bechtolsheim wrote them a $100K check, and Larry kept it in his pocket for a month until they set up a bank account.
It is tempting to believe a product works before it works. It’s especially hard for non-technical people, because they believe the engineers when they say that the product works, and then they prematurely scale.
I know that if I have a product that works, I can hire a marketing and sales and distribution team, and I can do it in a week.
You have to time it right. How do you know? First, people use it. Nobody at Google used Wave. How many people do you see walking around Stanford using Google Glass? Maybe we need just a small change, and then boom. You can pick hundreds of examples at Google, and thank goodness you can’t remember any of them.
Q: How do you build really fast for global opportunities. What techniques did you invent?
A: Jonathan and I wrote “How Google Works.” 1/3 of that book is about hiring. Sheryl set up most of those processes. Bob Taylor said it best, “Sell the dream.”
The way he funded the ARPANET was he called people up and described the project. If the person didn’t say yes right away, he moved on. You need people who get it quick.
You say to hire generalists. I’d say, we need people with this kind of experience. Larry and Sergey would say, that’s the stupidest thing I’ve ever heard. Okay, so who do you want to hire? Incredibly intelligent people. They’d figure out that what you want them to work on is stupid, and they work on the right things. It was all in good fun.
Their point was that the industry overvalues experience, and undervalues strategic and tactical flexibility.
In all my issues at Google, I knew I had no idea what to do, but I knew that I had the best team ever assembled to figure out what to do. Take AdWords. We were selling on an “as-sold” basis. And this young guy, Sal, said, we’re going to switch to all auctions. I was so worried that I implemented a cash restriction program. The day we turned on the auctions, revenue tripled.
You need ideas like that, that scale so fast. Let me put in a pitch for my current idea, machine learning. In the future, machine learning will be how you get leverage. Machine learning will produce immensely larger companies.
A: One of our rules was, we don’t want to hire your friends. Another rule was not to hire people from “lesser” universities. Another rule was to only hire people with good GPAs. It was frustrating, but it meant that we ended up with a lot of really smart people from great universities, and that served us well.
At Novell, we had what I called “glue” people. They sit at the boundaries, and help everyone work together. I tried to get rid of them because you don’t “need” them, but they kept coming back. I was telling Larry and Sergey about this, and they said, “Why don’t we just review all the hiring?” After that, we reviewed all the hiring packets and threw out all the glue people. One time, we had a guy that a founder didn’t want to hire because his last name was too humorous. We hired him.
Because we don’t fire people, it’s really important who you hire. The people that you hire make your culture. We’d hire people who were special in some way. You don’t hire generic people — you hire people who have had stress and achievement. The best people to hire are CFOs who’ve gone bankrupt, because they’ve been through the wars.
Once we decided to review all the tactics, we put in a scoring system. Sergey said, the problem is, these scores are biased. So we looked at the individual scorers, and how well the employees did after we hired them. We found, for example, that there was an inverse correlation between how we rated female candidates and their performance. So we had to change our entire hiring process to correct the unconscious biases that were affecting our hiring.
We also had the problem of people being interviewed. I still remember this packet of someone who had been interviewed 16 times. They decided not to hire them. So I said, you just can’t do this to people. You’re not allowed to interview them more than 8 times. Now it’s 5 for engineers and 4 for non-engineers. These are simple techniques that wouldn’t occur to a regular company. The people we hired applied an analytical bias.
During this heyday, we made sure that our marketing was very disordered. We’d show people the lava lamps, and then tell our shareholders, we have another half of the company that is completely ordered.
Q: What works for organizations?
A: Larry and Sergey and I would run things, so we just tried to go as quickly as possible. Every Monday, we’d have our 60 minute meeting, which took 2 hours. The goal was to get people off their computers for one hour. It didn’t work.
On Tuesdays, we had Google Product Strategy meetings. We were launching products, and we had no idea. The engineers were launching products without talking to the support people or the lawyers. For example, one engineer created a product that could geolocate you and your friends and predict where you were going to be. My face went ashen. Larry and Sergey loved it. They were screaming about how wonderful it was and how stupid I must be not to see it. I told them, it was great, but it is illegal, and we couldn’t afford to respond to all the subpoenas. Of course, Larry and Sergey were doing this for fun, to make a point. The solution was to allow the users to lie about their location. Then we launched Google Lattitude.
After we launched Chrome, we had a debate about HTML5 and the exclusivity of browsers. We had 20 people in the room, and it was clear that Sergey was not going to give up. He disagreed with me, and Larry agreed with both of us. I told them to go back to their office and figure it out, and if they didn’t have a decision by noon tomorrow, I’d make the call. So the next day, they had another solution that was better than any of the original three.
We had an engineer who built a free WiFi system. We were asking business questions, and it was clear he had no answers. The wrong thing would be to slam him for not having the answers. The right thing, which I did, was to tell him that free WiFi is a great idea, and you need to figure out these issues and come back to us. 6 months later, we launched the product to great success.
We did Deals on Wednesdays. Thursdays were no-meeting days, which of course meant that they were filled with meetings.
Wayne Rosing came to me and said, Larry is on the warpath. He was reading the “Snippets” and comparing them to what the managers were saying. So Larry turned the five engineering managers into individual contributors, and had 120 engineers report to one person, Bill Coughran. We ran that way for two years.
Q: What was the role of the CEO?
A: My role was to manage the chaos. You need to have someone to run fast and have a good product sense. That was Larry and Sergey. My job was to organize the world around them.
Larry told me, “We don’t need you now, but we’ll need you in the future,” and I think they were right.
Mark Zuckerberg is a great technical founder who allows the key stakeholders (Sheryl and the CFO) to run large parts of the company.
Q: What were key things not to do in order to move fast?
A: Surprisingly, not many. Because the ambition was so broad, my only lever was to slow a few things down. I don’t agree that you should narrow your focus.
We would refuse to do exclusive deals, but I would tell the partners, we only have capacity for one partner, and you’re it.
All success starts from doing one thing really well, but you’ll recruit better with a broader vision so you can sell the dream.
Q: 20% time?
A: There were two things, Don’t Be Evil and 20% time. I thought Don’t Be Evil was a joke. There was a discussion about ad targeting, and this engineer named Ron pounded the table and said, “That would be evil.” We didn’t do the change. The analogy is the Kanban system — anyone can stop the line. But it only works if you have a strong set of shared values.
If engineers were passionate about something, they could work on it in 20% time. Many of our best features came from 20% time. Marissa would watch for these things. She maintained a Top 100 list with 300 things on it, and would try to get these things to talk with each other.
Q: How did you grow managers?
A: I would argue that the APM program has produced more entrepreneurs than any other program. Marissa had the idea that you would hire people right out of college, with a technical degree, who did not want to be programmers. She would train them — take them on trips for weeks — and it forged these incredibly tight bonds between people who were highly technical and could specify products. Sundar was one of these people.
The program was originally invented by Microsoft, though Marissa didn’t know that. That model is scalable.
We hired Shona Brown from McKinsey. We hired her and put her in charge of operations. She said, we’ll run our own McKinsey. We’ll hire young people who would otherwise work at McKinsey, but have them work for us. They’d work on issues, then go with the business unit where they worked. One of my Stanford students went into the program. Today, he runs half of YouTube.
Dennis Woodside came in through a similar program. We put him in charge of the emerging world. Then we put him in charge of all of Europe. Then all of Motorola. And now he’s #2 at Dropbox.
Jonathan said, “We need a chain gang — people who are waiting for work.” So we came up with projects that would take a month or less, and all of a sudden, we had all our questions being answered.
Q: What are counterintuitive signs of talent?
A: I’ll describe Noam. He went to Berkeley to study Mathematics, and got bored. He applied to Google, and was rejected. But someone knew him, and he showed us his spelling corrector. So we hired him. Larry asked, why didn’t we hire him? So we came up with a new rule — if someone is really, really smart, hire them anyway.
I was at dinner, and Noam came up to me and said, “I need 10,000 computers to solve general knowledge by this weekend.” So I told the guy in charge of machines, give him all the machines he needs.
We try to hire people who can co-exist with other human beings, but there are cases where we broke that rule too — we just keep them over there.
One of the things in our book is that you should hire divas. They are a pain in the ass. But the divas that believe will drive the culture. Steve Jobs was a diva. Bill Joy is a diva. If you find these people, you’re probably going to work for them someday, so treat them well.
Q: How important is it to have separate groups for innovation?
A: It’s a mess. Every great project has started with a graduate student and an assistant professor looking for tenure. Unix was two people. Gmail was one person, Paul Bucheit. Linux was Linus Torvalds.
One day, I thought, I love this Google Docs product. Bring me the team. “They won’t fit.” They had 150 people.
The teams are far larger than they should be. It’s a failure of architecture — the programmers don’t have the right libraries. I hope that machine learning will fix that problem.
Well-done platform APIs are the key to everything. Our system is so powerful, but so commingled with itself that it is hard to partition. We have a huge product to move more functionality into Google Compute.
The vast majority of startups are likely to use AWS. What’s powerful is that they are programmable by mere mortals. The original Mac toolkit is another example. I worked on the Alto — it had a single book that showed all the APIs.
Q: What about Alphabet?
A: There is a size at which companies begin to fall in on themselves. Google has done well because the founders are so talented and brilliant. Google, Amazon, Facebook — they all have strong founders and leaders. Steve Jobs thought Google was doing too many things.
We spent two years arguing about whether we should change the logo. Our mission is to organize all the world’s information. The three of us visited Warren Buffett in Omaha. Alphabet is an attempt to build a holding company like Berkshire Hathaway out of an existing operating company. It’s never been done before.
We didn’t even know how many Alphabet companies there would be. Google Life Sciences will be one. Calico, Google Fiber.
We’re trying to push the Alphabet companies to be separate companies, not divisions. Warren Buffett loves to hire people who would run the companies whether he hired them or not.
John, who is going to run our cars business, wants to work on this stuff 24/7. He is going to be the CEO, and bear the downside and the upside, and that’s what he wants.
Q: What let Google win?
A: First, you have to have the right founders. Impressive, smart, passionate, committed. Two, you need to have some luck. In our case, search and ads turned out to be a gold mine of revenue. It just went like this [hockey stick]. The fact that we ended up in a high gross margin business so early gave us the flexibility to try so many things. Almost every other business has lower margins and less flexibility.
The ideal business is Microsoft — a monopoly business with hardware competitors who need you, in a growing industry.
Uber is a software company, for a very specific reason. It has very different economics than if they owned the assets.
Why do hotels not own their own buildings? It’s better to be the operator with a fixed revenue share, and let others ride the real estate cycle.
Q: Were there any “Valley of the Shadow of Death” moments?
A: A lot. 6 months after we went to auctions, we had to merge three different databases. I asked if I should be in the datacenter, and the engineers said, what are you talking about, we never go to the datacenter. So I flew to New York to take the weekend off. I was sitting next to one of employees, and the customer called up and yelled at her for 30 minutes. She told me, he calls every day. Is he unhappy? Yes, because he couldn’t get his report to show his boss how much product we got sold for him.
I went back to the office, and told people, we have to fix this. We met every day, then every hour. And we did.
Student Question: Can you walk us through your day?
A: In the morning, I went to the doctor for a routine exam. The results were fine. I drove to the office. Larry and Sergey got bored with the old office, and moved us to the Rails office — the old Mayfield building. Sergey decided that we were wasting money on interiors, so it’s a cold, concrete shell with power outlets. It’s filled with drones, self-driving cars, and kites. We had a meeting with the Chairman of Wipro. We’re trying to do a partnership to deploy Google products to help them with their business. Then we had our board meeting. Our rule is that you dress down. So everyone wears jeans, and Larry and Sergey typically arrive on rollerblades or bicycles. We have reports from subcommittees, and a report from Sundar, then demos and executive session. One of the things we’re arguing about is how to distribute software across the teams. Some people wanted to divide it between the teams. I argued that it should be a free good across all the companies, and I won. Then I went to dinner.
Student Question: How do you think about capital allocation for Alphabet?
A: We have about $70 billion in cash. Currently, it sits in Google, which is controlled by the parent, and at some point, we’ll move it around.
The process for funding is the same as always. There’s a competition for the best ideas. Take Google Fiber. It’s an incredibly good business, once it’s established. I’ve argued that we should be as aggressive as possible. It turns out that the rate of deployment depends on getting permissions and electricians. Companies like ours have so much cash that the main limit is opportunities to deploy it.
We never stop hiring engineers; properly deployed, they can always generate enormous returns above their salary.
Student Question: What emotional toll have you paid?
A: High-growth companies are a blast. This is success. Stop complaining. I was at Novell. I’ve seen the other side. This is good!
Student Question: Antitrust?
A: What I’ve always told our product teams is, I’ll do the lawyering, and you do the producting. Just build me incredible products.
Obviously, it’s a concern. But now that the Europeans are focused on all the American companies, it’s a little easier. European law is biased against large companies. There’s a commissioner in Europe who launched a study — he listed 32 companies, 31 of which were American.