Bits of Books - Books by Title


In The Plex

Steven Levy





Early VC guy wanted Page and Brin to hire a real CEO. At first totally refused, so Doerr took them on a tour to meet all the high tech royalty - Steve Jobs, Andy Grove (Intel), Scott Cook (Intuit) jeff Bezos (Amazon). OK we'll do it - as long as it's Steve Jobs. (At the time Jobs CEO of both Aplle and Pixar.) Finally Doerr found them Eric Scmidt, who was running Novell, and who had big advantage of being a computer engineer that they admired. But the final decider was that he was the only candidate who'd been to Burning Man.

Bill Goss fecund innovator with rep for coming up with clever tech tricks. GoTo ran auctions for advertisers to put ads on pages depending on words used in a search. And he started idea of advertisers paying for clicks on their banner ads rather than the number of page views. But he made a fundamental mistake. They missed the one year deadline for patenting the two breakthroughs.

Google didn't like idea of straight auction - if you bid 50cents a click and next best was 40cents, you were stuck with paying 9cents too much. This led to companies springing up to automate the cat and mouse game of winners shaving down and losers shaving up, and eventually winding up with a substantially lower bid. Google changed system so the winner just paid 1cent more than the next best bid, which put the incentive back to bidding higher. (They found out later that it was a well-known system called the Vickery second-bid auction, which had won its innovator a Nobel Prize)

Google's adwords had a further refinement. The placing was determined only partly by the amount advertiser bid. Rest was by how many times it got clicked. Google wanted ads that were useful to users. Forces the advertiser to do all the work; to make sure ad keywords relevant, and not pestering visitors with anything at all like spam.

When launched AdSense, at first bought ad space on pages like New York Times and the matched ads with content. But a few teething probs. A gory story in New York Post about a body chopped up and stuffed inside a garbage bag was matched with a Google ad for plastic bags. A story about a food poisoning outbreak at Los Angeles Olive Garden restaurant was accompanied by a coupon offering "free dinner for two at Olive Garden".

Customers had problem switching to an auction system for placing ads as opposed to just buying the space for a fixed price. But Google salespeople no longer selling space, they were providing data to help guy sell more, using tools which G provided to not just assess the ad, but to rethink the way company worked.

Old way of sellimg ads was over. New way more like computer dating - matching advertisers and users.

Both Page and Brin were Montessori kids, so ingrained from start to question, to do things their own way.

When G IPO, obviously a lot of vested staff got very rich. But head of engineering, Wayne Rosing addressed a full staff meeting waving a baseball bat. He told them that if he saw any new BMWs or Porsches in the parking lot he wd use the bat to smash their windscreens. They were also told not to check the stock prices during the day. When didn't get full compliance of this policy, it was revamped: if you saw someone checking the stock ticker, all you had to do was walk over and tap them on the shoulder. Then that person wd have to buy you a share of stock.

Google did things differently when hiring project managers. Whereas conventional firms wd hire top business grads from Stanford or Harvard, Google wanted engineers who were smart enough to understand engineers. So they hired ambitious computer science grads. But they didn't have intrinsic authority to order their teams of (older) CS engineers around. They had to persuade them, and cd only do it with data. This was an advantage for Google, bc meant product development had to make sense to everyone, not just stuff that the manager loved.And the data wasn't hard to get - they cd run a '1% A/B expt' where 1 in 100 users gets a modified version of program, and then measure how that affected average page views or ad clicks.

John Doerr put $12.5m into Google in 1999 and joined board. But he also introduced Objectives and Key Results, or OKR, which he'd got from Andy Grove at Intel.Google guys loved it, bc forced everyone to put a number on their projects. Couldn't say "I'm going to make gmail a success" but "I will launch gmail in September and have a million users by November". And the numbers were public - the OKRs appeared on everyone's biographical info on G's internal website.

When G released gmail, it had 2 unusual features. First it showed ads that were related to content of yr email. Second, G boasted that with gmail you cd keep yr emails forever. G's engineer culture thought both these features were cool, and were stunned when users complained - they were upset that G seemed to be looking at their emails to get context ads, and they weren't happy that their stuff wd be stored by G forever.

More books on Computers

Inevitable showdown over privacy. It wasn't Google that put info onto webpages and online databases, but it was G that dislodged it all. The problem wasn't so much that people published, it was that G was relentless in finding out everything about them, whether it was contact details that used to be hidden in a database, or twenty year old criminal charge that may or may not have resulted in a conviction.

Controlling Your Online Identity

It affected people at Google. Search exec Susan Wojcicki second entry was post from a Silicon Valley gossip blog which inaccurately claimed she'd stolen some of the credit for developing AdWords. CEO Eric Schmidt objected to information about himself and a political donation, and wanted it removed from the indexes. But his staff refused, on the basis that the whole purpose of G was making all info accessible.

Google needed astronomical number of servers to run its data farms. The key to its strategy was buying low quality equipment at dirt cheap prices, knowing that a percentage of them wd fail, but designing in smart redundancy so that it didn't matter.

But main thing they worried about was speed - how long pages took to load. Pages Law: software guys can't resist temptation to keep loading their web pages with every new gimmick, so every 18 months, software becomes twice as slow. G had data logs to show how important speed was. When the web slows down, people consciously or unconsciously decide not to do extra searches. They may not even realize they're doing it, but on aggregate, G can tell. They boosted performance of Picassa, G's photo sharing site, and though they made no official announcement, traffic on the site increased 40% the first day it was implemented.

Microsoft found the same thing when they were testing Bing. When results were delayed, users responded with their own latency, taking longer to click on links after a search was completed. The presumption was that the longer the results delayed, the more time users have to start thinking about something else, and have to refocus before they decide what result to click on.

Measured things in terms of human lives. A 70 year life span is about 2 billion seconds. If a product has 100m users and it wastes 4 seconds of a users time every day, then that's more than 100 people killed in a year.

When G set up servers on east Coast of US for first time, they had to shift terabytes of data across the country. And in 2000, broadband was expensive. At a Gb per sec, it was going to cost 1/4 million dollars a month. So G exploited a loophole in the billing system, where the provider measured activity in pipe, but discarded unusual spikes of activity. They figured if they used zero bandwidth all month except for the 30 hours they needed to push data, they would be under 5% usage, and the spike would be dropped from the bill. So for 3 nights a month G pumped huge volumes of traffic, and the bill came out to be nothing.

Then G found it cd buy up fibre at rock bottom prices from all the optical fibre companies that fell over in the dotcom crash. By the end of 2001, G owned more fiber than anyone on the planet. They also realized that they had to start building their own data centers. The first one was on the Columbia River out past Portland. The 200,000 sq ft building cost $50m, but the servers - over 100,000 of them -inside were worth nearly a billion dollars.

G guy (Dan Siroka) worked on Barak Obama election campaign. Conventional wisdom was that you got people to donate with a crafty emotional pitch to engage their idealism or their politics. Siroka ran a lot of A/B tests and found by far best tactic was to offer swag - T-shirts or coffee mugs. Even found message on nav buttons was important: LEARN MORE was significantly better than JOIN US NOW or SIGN UP NOW.

When G started its Universal Library project, couldn't find a machine that cd turn pages gently enough, so used army of humans, and so you see a huge number of scanned fingers and thumbs.G originally calculated that 33 million books had been published, and they wanted them all. Later, using a wider definition of what a book was, and in all languages, estimate went out to 129,864,880 books as of August 2010.

Facebook - A 'That' company: "The company that shows up once in a very long while the Google of yesterday, the Microsoft of long ago ... that company that's on the cusp of Changing The World ... where you know you'll kick yourself in three years if you don't jump on the bandwagon now, even after someone told you it was rolling towards the promised land."

Books by Title

Books by Author

Books by Topic

Bits of Books To Impress

Reputation Control .........................................................................................Client