When you focus on the things people use every day, you solve big problems.
iGoogle gadgets are a new form of advertising.
The orginary and the every day.
Occam’s razor for logic: the simplest answer is probably right.
The Google homepage. Sergey created the original homepage. What inspired him to make it plain?
“We didn’t have a webmaster. And I don’t do HTML.” Sergey Brin
Search is what we focus on.
In the mid-90s, we did a study at Stanford: Find out which country won the most medals in the 1992 Olympics. Two users trying to find the answer. They took about 15 seconds to think of something to type. 30 seconds go by. 45 seconds go by. Why aren’t they talking to each other? What are you waiting for?
“I’m waiting for the rest of it”.
There is a copyright at the bottom of the page, and it’s not for any legal reasons. It acts as a period.
People would ask, “Is this a real company?”
“How many people work there?”
80 people is the answer, said Mayer. But she hesitated to tell people that.
Search for “beautiful code”. It would bounce through a few datacenters until it accepts a query. Then it is read by a bunch of machine, hit the load balancer. Then to a mixer. Then queried to Google web server, then it hits about 300-400 backends (AdWords).
Another Google web server.
In an average query, you will have hit about 700-1000 servers to return 5 million results. In milliseconds.
1.3 searchs per user.
It’s important to know where you are and where you want to be. Know them well enough to guess. Know the state of the art. How much does disk cost? How close are we to maxing out?
Split AB testing.
Design is becoming a science. You can mathematically find different things. How does whitespace affect our user happiness metric. How does the color yellow increase metrics?
Understand users better than they understand themselves
“If I had asked people what they wanted, they would ask for faster horses” – Henry Ford
How many results are best on a search results page? 10, 20, 25, 30 results? After about 6 weeks, Google searches dropped 20%. Also 20% fewer first-page searches. It takes a little bit longer to pull that many results. Milliseconds. Latency caused this decline.
As the web speeds up, people search more. Speed really matters.
Same thing on maps. An Ajax application that weighs 110k isn’t bad. Can we put Maps on a diet?
We pulled 30-50k out, and as a result 30% more searches.
At Google video, it would say “Please wait 24-48 hours…”. YouTube came along and said, watch it now. The urgent can drown out the important, though.
How do you take searchers and make them better? You could make technology better, but we need something that would teach users how to be better searchers.
Because the feedback loop is so fast, searchers go from novice to expert faster. Think 10 years out.
Search will be a lot more personalized in ten years. We will have a lot more content to index. When we think about how to build search, it’s important to look into the future.
How do you scale retrieval from each index?
We like search, because it’s really hard. Google 411 is about a real, flexible technology: voice-to-text. If it works, then we can move this technology to our other services.
1% of the web is Arabic, whereas 50% of the web is English. If we can build a great automated translation engine, we can type in Arabic, traslate the query into English, get more data, and then fold them into an Arabic results set. Imagine broadening this technology out.
“A healthy disrespect for the impossible” – Larry Page
The things we get excited about are usually impossible. Search for “Dr Zhivago”. Dr means doctor. “Rodeo Dr Burton Green” –> Dr means drive.
“New York Times Square”. But add two words, “the circle”, you could be talking about an article on NYT.
How can you actually correct queries and find user intent
Google Health – an attempt to put users medical records in their hands, in their control. Having prescriptions, lab results, etc is really exciting. We realize it’s an impossible problem. It’s impossible
Trying to work on a really hard problem, will still make a difference in someone’s life.
Google Book Search… impossible. There are so many books, you can’t get them all.
Be scrappy. Revel in constraints.
Around the world, Google has about 140 domains for search. When we first did our internationalization, things weren’t that smooth. We figured out that letting users write the other versions of the site. “If you don’t see your native language, help us build it”
Google bork, bork, bork uses the UNIX chef command for a Sweedish chefs.
Imagination is a muscle
Google Ride Finder, “what were you thinking”. It uses GPS signals from taxies and other rides. It was interesting for the engineers to figure out if we could get this data, and update it in real time. People don’t use it for finding rides, but it’s an experiment to see if this technology exists and works. It’s useful because it exercises the imagination muscle.
What are the different state of the arts? What new things can you do? It keeps you thinking about what’s possible.
Use divergent thinking to get the creativity flowing.
Google Street View – is camera technology good enough?
20% time. Google News, Google Alerts, Orkut, Google Wireless, Google Labs, Google Scholar. 50% of our features started as 20% time projects. It’s really about letting people build things they want.
The orginary and the everyday becomes the extraordinary. The simplest design is probably right. Numbers help you understand your users better than they can understand themselves. Building real and flexible technologies.
