I've heard the speculation about Google since I left. That it's a monopoly. That it's tracking users. That it's in cahoots with the government. That it spies on people. That it's evil. Well, maybe it is all that. I haven't worked there in more than five years. Things change. But from what I know about my coworkers in the Plex—many of whom are still there, putting in long hours perfecting a product used by millions every day—I'd say that's highly unlikely.
Is Google secretive? No question. Arrogant? Maybe. Tone-deaf to the concerns of the very users it claims to serve? Occasionally. But evil? I don't think so.
I started my career working at ad agencies. It was fun, challenging, and potentially well paying. I quit because I didn't like the idea I might have to sell something I didn't believe in. I worked in public broadcasting and then newspapers, where I found coworkers who sacrificed material rewards to be part of something connected to the common good. I got that same sense at Google, but with greater intensity and urgency. And stock options. This was no institution continuing a long tradition of public service. This was a headlong rush to reshape the world in a generation. And therein lies the company's biggest flaw, in my estimation: impatience with those not quick enough to grasp the obvious truth of Google's vision.
"When were we ever wrong?" Larry asked me.
Not often. But "not often" is not never. If Google's leaders accepted that reality, they might understand why some people are unwilling to suspend skepticism and surrender to Google's assurances the company can be trusted.
After Google, I find myself impatient with the way the world works. Why is it so hard to schedule a recording on my DVR? Why aren't all the signal lights synched to keep traffic flowing at optimum speed? Why, if I punch in my account number when I call customer service, do I have to give it to them again when I get a live person? These are all solvable problems. Smart people, motivated to make things better, can do almost anything.
I feel lucky to have seen firsthand just how true that is.
Timeline of Google Events
11/29/99 My first day at Google
12/4/99 CableFest'99
12/13/99 Inktomi partners with MSN
1/27/00 Premium Ads (original) launches
1/30/00 First OKRs set
3/14/00 Google directory launches
3/22/00 Larry becomes chief ofm products; birth of product review
3/27/00 Affiliate program launches
4/1/00 MentalPlex April Fools' joke
6/26/00 Yahoo replaces Inktomi with Google
9/27/00 AdWords launches (do-it-yourself CPM ads)
12/10/00 Google toolbar launches with "Not the usual yada yada" warning
1/17/01 Wayne Rosing starts full-time as VP of engineering
2/10/01 Acquisition of Deja News
3/20/01 Eric Schmidt named Chairman of the Board
3/29/01 China blocks access to Google
7/5/01 Engineering reorganization
8/1/01 Chad bikes America
9/11/01 Response to September 11 attacks
10/8/01 GoTo renamed Overture
10/24/01 Trakken CRM system installed
11/13/01 Yahoo-Overture deal announced
11/20/01 "10 Things We've Found to Be True"
11/28/01 Launch calendar instituted
12/13/01 Google catalogs launches
1/24/02 "No pop-ups" linked from Google's homepage
2/4/02 Earthlink switches from Overture to Google for ads
2/19/02 AdWords Select launches (auction-based CPC pricing)
3/5/02 "Why we sell ads, not placement" on homepage
4/4/02 Overture files patent lawsuit against Google
4/25/02 Yahoo renews with Overture for three years
5/1/02 AOL drops Overture and Inktomi for Google
5/2/02 Sheryl Sandberg begins building AdWords team
7/23/02 Yahoo renews contract with Google for search results
8/31/02 China blocks access to Google results again
9/11/02 Content-targeting project starts
9/22/02 Google news launches
10/1/02 Expansion to Saladoplex
11/20/02 Project Ocean launches (digitizing offline content)
12/12/02 Froogle launches
12/13/02 Yahoo buys Inktomi
1/6/03 Milan office opens
2/14/03 Google acquires Blogger.com
2/25/03 Overture buys FAST
3/12/03 Microsoft calls Google "pathetic"
4/11/03 1,000 Googlers
4/24/03 Google acquires Applied Semantics (content-targeting firm)
6/6/03 Yahoo to spend $10 million on branding campaign
6/17/03 Google stock splits
6/18/03 AdSense launches (self-service content-targeted ads)
9/25/03 Leak of financial data
12/3/03 All logs data available in Sawmill system
1/16/04 Yahoo drops Google
1/22/04 orkut launches
1/28/04 Leaks to the New York Times
4/1/04 Gmail launches
4/29/04 S-1 filed, "quiet period" begins
7/8/04 Financial planning fairs for Googlers
8/9/04 Yahoo/Overture settles patent lawsuit with Google
8/10/04 Playboy interview with founders published; one-week IPO delay
8/19/04 IPO
10/21/04 Earnings call
2/1/05 Earnings call
2/3/05 Microsoft announces $200 million ad campaign for search
3/4/05 My last day at Google
Glossary
This book is written in English but contains some phrases in
Geekspeak. Here's what they mean.
AdSense: Google CPC ads that appeared on other websites next to content other than search results and were content targeted.
AdWords: The second Google advertising program, after Google Original Ads. AdWords gave advertisers the ability to create and manage their own CPM ad campaigns using an online tool and a credit card. The ads contained only text, no images, and would automatically appear next to Google.com search results when a user searched for the keyword(s) targeted by the advertiser.
AdWords Select: The successor to AdWords. The system that introduced auction-based CPC pricing for Google ads.
Affiliate program: A program by which one website pays other websites to send it traffic. The sites sending the traffic are called "affiliates." For example, Amazon pays a commission to affiliates that promote specific books it offers for sale. The affiliate gets paid only if one of its users clicks on a specially encoded link pointing to Amazon, and then actually purchases the book there. Google's program paid affiliates if their users conducted searches using special Google search boxes that could be added to almost any website.
Algorithm: A set of instructions (such as a computer program) for solving a specific problem, like finding a particular bit of information in a database. Also, a stiff, rocking motion modeled by a former vice-president.
Canonical: When engineers don't want to say "the preferred" or "the usual" way of doing something, they talk about the canonical example, which is pretty much the same thing but conveys a more exalted sense of correctness to the practice in question. The canonical way to avoid spending an hour in traffic is to go to work at noon.
Clickthrough rate (CTR): The percentage of people seeing an ad who actually click on it. A high CTR means an ad is effective at driving traffic to the advertiser's website. Anything better than 1 or 2 percent is considered high for most types of ads, but CTRs improve the more targeted an ad is to the interests of the person viewing it.
Content targeting: The placement of ads relevant to the text of the page on which they appear. This requires analyzing the page's content and then matching ads from a database that are likely to appeal to someone reading that content. For example, a news story about vacationing in Florida might include ads about theme parks there, airlines, and hotels.
CPC: "Cost per click"—a method of paying for online advertising in which the advertiser pays on
ly when a user actually clicks on the ad and is transferred to the advertiser's website. This is a form of "pay-for-performance" advertising.
CPM: "Cost per thousand impressions"-a method of paying for online advertising in which advertisers pay a set rate to have their ads displayed a thousand times, regardless of whether anyone actually clicks on them. The M in CPM comes from the Latin word "mille," which means one thousand.
CPU: Central processing unit—the part of a computer that does the actual "thinking" or processing of code to execute instructions.
Crawler: An automated computer program that follows links from website to website, collecting information about the pages it finds and the URLs at which they live. See Spider.
Cruft: Cruft is bad. Like the stuff that grows under ungroomed toenails. Like barnacles on a speedboat. It usually refers to old code or dead links on a web page, but it can be applied to any unwanted material that accumulates anywhere. The men's locker room in the Googleplex was filled with cruft, much of it unwashed and hockey-related.
Database: Information collected and stored in a computer in a structured way, so that it can be easily accessed and searched. Like a dictionary that contains words in alphabetical order.
Google Original Ads: The first ad system developed and deployed by Google, it offered CPM ads that were sold by sales representatives. These ads, which appeared above the search results, became known as "premium Ads" when AdWords ads were introduced.
Googler: A Google employee who is neither still a Noogler nor yet a Xoogler.
GoTo (later renamed Overture): An advertising network founded in Pasadena, California, that pioneered the idea of a pay-for-placement search engine, in which companies bid for the most prominent position in search results for specific terms. GoTo was Google's key competitor in supplying ads to major Internet sites like AOL, Yahoo, and MSN.
GWS: Google Web Server (pronounced: "gwiss"). The software that interacts with users when they enter searches and when Google sends results back to them. It doesn't determine the results, it just delivers them. It also controls the look of the pages on Google's website, so when those pages need to change, a new GWS needs to be pushed out (sent) to the servers.
Hardware: Any part of a computer you can touch with your hands. Anything with wires, disk drives, cables, or a power cord coming out of it.
Incremental index: An index that is continually refreshed with new data and is integrated with another index that changes less frequently. For example, an incremental index might include daily-updated pages from news websites to be mixed in with infrequently changing pages that form a much larger index.
Index: An organized list of web pages that can be searched much more quickly than the original pages listed within it. There are many ways to organize an index, and the more efficient the technique used, the faster an algorithm can find and retrieve a specific piece of information.
Inktomi: A search-technology company from Berkeley, California, that supplied search results to the majority of web portals, such as Yahoo, AOL, and MSN, before Google entered the market.
Intellectual property (IP): The output of a creative effort that can be legally protected. It often refers to patented technology, such as search algorithms, but can also refer to music, movies, artwork, and so on.
Internet Service Provider (ISP): A company that provides a connection to the Internet, whether by phone line, cable modem, or wireless network.
Keyword (also query or search term): The word or words that users type into a search box—that is, the thing they are trying to find. Advertisers specify (target) keywords when they purchase ads. When a user searches for one of the targeted keywords, the results page may display the targeted ad.
Machine: A generic term for a computer or a web server. Also a box, a PC, a server.
Nontrivial: A euphemism for "impossible." Since engineers are not going to admit anything is impossible, they use this word instead. When an engineer says something is "nontrivial," it's the equivalent of an airline pilot calmly saying you may encounter "just a bit of turbulence" as he flies you into a Category 5 hurricane. See also Trivial.
Noogler: A new Googler. The tag usually sticks until another Noogler is hired into the same group.
Orthogonal: Engineers often talk about things being orthogonal to each other. The first time I heard the term, I thought it meant something like "eleven-sided." It doesn't. It's some kind of technical way to say "unrelated." I still don't really get it. But that didn't stop me from casually dropping it into conversations with engineers: "Oh, yeah, that press release is totally orthogonal to the ads we're running on Yahoo."
Overture: The name assumed by the advertising network GoTo in October 2001.
PageRank: An algorithm used for analyzing the relative importance of pages on the web. Written by, and named for, Google's co-founder Larry Page. PageRank's breakthrough approach was to look at the sites linking to a particular page to determine how many other websites deemed that page authoritative or important.
Pay for inclusion: Some search engines accept payment from website owners to guarantee that their sites will be included in search results. These search engines don't necessarily guarantee the site prominent placement. Web-crawling software can take weeks or months to find new sites; pay for inclusion gives those sites a way to accelerate the process.
Pay for performance (PFP): A method of paying for online advertising in which the advertiser pays only when users actually perform an agreed-upon action, such as clicking on the ad or registering for an account on the advertiser's website.
Pay for placement: The practice of some search engines to accept payment from website owners to give their sites more prominence within search results, as opposed to in a separate, clearly marked advertising area.
Query: The words a user types into a search box, or a single search. See Keyword.
Server: A computer that has been configured to hold large amounts of information and provide it to other computers quickly across a network.
So ... : This all-purpose word is not a word at all. It's the sound of an engineer clearing his or her throat before beginning to speak. The first week I worked at Google, it seemed as if some linguistic virus had infected all the technical staff. Every sentence in every conversation began this way. So ..., eventually, I got used to it.
Software: The part of a computer you can't touch: the programming and applications that instruct it what to do. All the bits and bytes that are stored in a computer's memory.
Spam: As a noun, "spam" is any unrequested and unwanted electronic material sent by one person to another, whether it's junk email sent to thousands of people simultaneously or a chain letter sent by your very close friend who swears something bad will happen if you don't forward it to ten others. As a verb, "to spam" means to send spam and can also mean to try to obtain an advantage, such as a higher ranking in search results, through deceptive practices, including hidden text on pages or unnecessary repetition of certain words.
Spider: Web-crawling software that gathers data from websites that are the basis for an index. See Crawler.
Targeting: Matching an ad to a trigger that causes it to be displayed. The trigger may be a keyword a user enters for a search, or the content of a web page determined to be relevant to the subject of the ad. The better the targeting, the more relevant the ad is to the keyword or content that triggers it, and the more likely the user is to click on it.
TGIF: Short for "Thank God it's Friday," and the name of a weekly meeting where Googlers are updated on the week's events, usually while munching on chicken wings and drinking beer.
Trivial: As defined by Google engineer Georges Harik: "I say that a task is, 'trivial' if I think it is possible. It's, 'easy' if I can't see a way to do it but I'm certain it can be done. It's, 'hard' if several people have declared it impossible but I disagree with them. It's, 'impossible' if I am too tired to do it."
URL: Short for "uniform resource locator." A URL is the address of a
particular page on the web. Many URLs begin with "http://www." and end with ".com" (for example, http://www.google.com).
User interface (UI): The look and feel of the part of a website visible to visitors. The UI is the website's face and includes the graphics, the text, the forms, and any other elements with which a user can interact.
Web server: A computer that delivers content, such as a web page, when requested to do so by another computer on a network. Or the software that runs on the computer that delivers the content. See Server.
Xoogler (pronounced "zoogler"): A former Googler. I coined this term as the name for a blog I maintain at Xooglers.blogspot.com, a gathering place for former Googlers to reminisce about the company's early days.
Acknowledgments
I was lucky long before I'd ever heard of Google.
My parents Marvin and Helene Edwards raised all their children with love and fairness, and instilled in us a desire to be upright and do good in the world. Not once have they wavered in their commitment to our happiness or hesitated to sacrifice to bring it about. When they could ill afford it, they helped fund my dreams and then gave me the confidence to pursue them. I can never repay the debt I owe them for their guidance, their patience, and their understanding.
My wife Kristen experienced all the pressures and insanity of a Silicon Valley startup without the compensating perks that I enjoyed. She didn't divorce me. For more than twenty-five years she has encouraged me, supported me, engaged me, and endured me. She is a ruthless editor and a stickler for facts. She has, on occasion, been the only reason I've remained sane. She is my best friend. "Gratitude" is an inadequate word for all I feel for her, but the end pages of a book about search technology hardly seem the place to delve deeper.
I'm Feeling Lucky: The Confessions of Google Employee Number 59 Page 46