How Google Works

by Eric Schmidt

Knowledge Graph in search is based on organizing the Internet’s vast amount of unstructured data about a particular person, place, or thing into structured data, and presenting it in an easy-to-consume format.

Hangouts (live video chat with one or more people) transcodes various video formats in the cloud rather than at the device level, making it one-click easy to conduct a global video conference from any device.

Product leaders create product plans, but those product plans often (usually!) lack the most important component: What is

the technical insight upon which those new features, products, or platforms will be built? A technical insight is a new way of applying technology or design that either drives down the cost or increases the functions and usability of the product by a significant factor. The res...

The best thing about market research consultants? They are easy to blame and fire when they are wrong.

market research can’t tell you about solving problems that customers can’t conceive are solvable.

There’s nothing wrong with continuous improvement and smart business tactics, but the tail is wagging the dog when market research becomes more important than technical innovation.

The best products had achieved their success based on technical factors, not business ones, whereas the less stellar ones lacked technical distinction.

We are entering what lead Google economist Hal Varian calls a new period of “combinatorial innovation.” This occurs when there is a great availability of different component parts that can be combined or recombined to create new inventions.

For example, in the 1800s, the standardization of design of mechanical devices such as gears, pulleys, chains, and cams led to a manufacturing boom. In the 1900s, the gasoline engine led to innovations in automobiles, motorcycles, and airplanes. By the 1950s, it was the integrated circuit proliferating in numerous applications. In each of these cases, the development of complementary components led to a wave of inventions.

Another potential source of technical insights is to start with a solution to a narrow problem and look for ways to broaden its scope.

New technologies tend to come into the world in a very primitive condition, often designed for very specific problems. The steam engine was used as a nifty way to pump water out of mines long before it found its calling powering locomotives.64

When you base your product strategy on technical insights, you avoid me-too products

that simply deliver what customers are asking for. (Henry Ford: “If I had listened to customers, I would have gone out looking for faster horses.”)

Back then, scaling meant reaching millions: It took the global phone network eighty-nine years to reach 150 million phones.

At the risk of stating the obvious, though, a successful foundation must provide a good basis for revenue generation.

Many smart twentieth-century companies ran the numbers and found that for much of what they wanted to get done, Coase was right: The internal management costs were lower than the transaction costs of outsourcing. This led them to do as much as they could within the organization, and, when they did go outside their four walls, they worked with a small group of tightly controlled partners.

So the twentieth century was dominated by corporations that were large hierarchies—or, at their most expansive, closed networks.

This is the difference between twenty-first-

first- and twentieth-century economies. Whereas the twentieth century was dominated by monolithic, closed networks, the twenty-first will be driven by global, open ones.

Default to open, not closed

Vint Cerf and

Robert Kahn76 developed TCP/IP (Transmission Control Protocol / Internet Protocol),

Another example is the IBM PC, which launched in 1981 with an architecture that allowed software developers and manufacturers to build applications and add-on components, and even their own “clone” PCs, without paying IBM licensing fees. This decision helped establish the IBM PC as the definitive standard in the emerging “microcomputer” market, giving a huge

boost to a couple of small companies called

Microsoft and Intel.78 It also drew flocks of applications, accessories, and competitive manufacturers into the ecosystem, and ultimately created the dominant computing platform for the next twenty-five years. None of thi...

“Open” can be a rather Rashomon-like term—different companies will define it in different ways to meet their own objectives. But generally it means sharing more intellectual property such as software code or research results, adhering to open standards rather than creating your own, and giving customers the freedom to easily exit your platform. This can seem heretical to traditional, MBA-style thinking, which dictates that you build up a sustainable competitive advantage over rivals and then close the fortress and defend it with boiling oil and flaming arrows.

“no matter who you are, most of the smartest people work for someone else.”

If you focus on your competition, you will never deliver anything truly innovative.

Larry again:

“Obviously we think about competition to

some extent. But I feel my job is mostly getting people not to think about our competition. In general I think there’s a tendency for people to think about the things that exist. Our job is to think of the thing you haven’t thought of yet that you really need. And by definition, if ou...

As Nietzsche wrote in Thus Spake Zarathustra: “You must be proud of your enemy; then your enemy’s successes are also your successes.”87 Be proud of your competitors. Just don’t follow them.

The herd effect

Passionate people don’t use the word

A fine marker of smart creatives is passion.

Psychologist Carol Dweck has another term for it. She calls it a “growth mindset.”95 If you believe that the qualities defining you are carved in stone, you will be stuck trying to prove them over and over again, regardless of the circumstances. But if you have a growth mindset, you believe the qualities that define you can be modified and cultivated through effort.

You can change yourself; you can adapt; in fact, you are more comfortable and do better when you are forced to do so.

Finding learning animals can be a challenge. Jonathan’s modus operandi is to ask candidates to reflect on a past mistake. In the early 2000s, he used to ask candidates “What big trend did you miss about the Internet in 1996? What did you get right, and what did you get wrong?”

“I have no special talents. I am only passionately curious.”

We institutionalized the LAX test by making “Googleyness” one of four standard sections—along with general cognitive ability, role-related knowledge, and leadership experience—on our interview feedback form.

The same goes for managing people once they join you. Just like hiring, managing performance should be driven by data, with the sole objective of creating a meritocracy. You cannot be gender-, race-, and color-blind by fiat; you need to create empirical, objective methods to measure people. Then the best will thrive, regardless of where they’re from and what they look like.

Expand the aperture

The ideal candidate is out there. She has passion, intellect, integrity, and a unique perspective. Now, how do you find her and get her on board? There are four links in this critical chain...

Of course, sometimes we screw this up ourselves. Once, Salar Kamangar was impressed with one of our young marketing associates and wanted to transfer said young man into the APM program. Unfortunately, the APM program only accepted candidates with degrees in computer science, which this associate didn’t have. Although Salar argued that the young associate was a self-taught programmer and had a “history of working closely with engineers and shipping things,” several influential execs, including Jonathan, steadfastly refused to expand the aperture, and denied the transfer. The young marketing associate, Kevin Systrom, eventually left Google. He cofounded a company called Instagram, which he later sold to Facebook for a billion dollars.106 You’re welcome, Kevin!

Interviewing is the most important skill

Not only do most companies conduct overlong interviews, they conduct too many of them. One time, in our early days at Google, we interviewed a particular candidate over thirty times and we still couldn’t decide if we wanted to hire him. That’s just wrong. So we declared by fiat that a candidate couldn’t be interviewed more than thirty times. Then we did some research and discovered that each additional interviewer after the fourth

increased our “decision accuracy” by less than 1 percent. In other words, after four interviews the incremental cost of conducting additional interviews outweighs the value the additional feedback contributes to the ultimate hiring decision. So we lowered the maximum to five, a number with the added benefit (at least for computer scientists) of being prime.

The unit of currency in this system is the hiring packet, a document containing all the known information about a candidate who has progressed through the interview process. A hiring packet needs to be both comprehensive and standardized, so that all members of the hiring committee get the exact same information, and that information comprises a complete picture of the candidate. The Google hiring packet was designed by engineers with this objective in mind (as well as the objective of keeping Larry happy, since he reviewed each one getting an offer). It is based on a template that is standardized across the entire company (functions, countries, levels), with room for flexibility.

The urgency of the role isn’t sufficiently important to compromise quality in hiring.