Archives for August 2007

Clearing out my tabs

I always end up with a ton of open tabs in my browser. Here’s some of the things I’ve enjoyed, but won’t do a full-scale blog post about. You might have missed these the first time around:

– Mike Grehan noticed a Google experiment to let users suggest urls to Google for specific searches. If you repeat the search, your suggestion will show up at #1 for you. Google is always running a bunch of experiments; I just like the idea of users contributing suggestions to Google.
– I noticed two good articles about using AdWords well. The first one is from Amy Konefal. She walks you through separate bids for content vs. search ads; the ability to not show ads to sites you choose to exclude; Google’s Placement Performance reports, and how to check the return-on-investment (ROI) for individual sites; how to ramp up advertising using site targeting for sites that perform especially well; and how to mine Google’s new Search Query reports to find new keywords to bid on, or poorly performing keywords to exclude by adding as a negative keyword. You should really go read the whole post though; it puts some of Google’s ad tools in a nice historical perspective. It also drives home that Google provides a lot of tools for the advertisers that are willing to invest the time. I believe that the more familiar you become with AdWords, the better your ROI will be.
– The other AdWords article I enjoyed was by a post by Brad Geddes. Geddes discusses some of the same ground as Konefal from a slightly different viewpoint. Geddes additionally mentions geotargeting to improve your ROI, the fact that you can exclude IP addresses from seeing your ads, the fact that Site Exclusion allows you to block an unlimited number of sites, and Google’s invalid clicks report. That report tells you how many ad clicks were discarded by Google so you didn’t have to pay for them. Geddes points out that the invalid clicks report can help reconcile your analytics results with your ad clicks. Together these two articles cover a lot of AdWords tools. There are still a few other things you can explore, such as auto-tagging your ad clicks so that things like page reloads and back-button navigation don’t cause confusion.
Walkscore takes an address and estimates how “walkable” that address is. It looks at things like where the nearest grocery store is, how far it is to the nearest bookstore, etc. It’s a pretty neat use of web APIs for maps.
– I’m normally not a sucker for Matt-baiting. I didn’t link to the Cartoon Matt doll (until now, I guess). But I have to say that I really enjoyed LOLCUTTS. Very creative, Michael. 🙂 I’m surprised that you didn’t take advantage of some of my sillier photos.
– People have figured out how to compile and load native apps on the iPhone. One recent thing I’ve seen is an NES emulator for the iPhone. It’s not something I’d run, but I still like the idea of being able to run my own programs on the iPhone. Now if homebrew iPhone apps could read the tilt sensor well, maybe I wouldn’t have to carry a pedometer in my pocket. 🙂

Google Dance!

I am writing this blog post looking out onto the Googleplex courtyard where webmasters are dancing and talking. We just finished the “Meet the Engineers” Q&A. Lots of people came inside building 43 to ask about crawling, duplicate content, and more. I wish I could easily summarize it, but it was just straight questions for two hours.

I think I’ll go down to dance and talk myself. If you’re at the conference, you’ve still got a couple days to catch me and chat more.

Example questions at SES: Universal Search

Whether you call it blended search, 3D search (Ask’s name), or universal search, it has the potential to surface as many relevant results as other hot search topics such as personalization. At SES yesterday, I sat in the back of the Universal Search session. There are good write-ups on the PowerPoint and presentations, but not as much coverage of the questions.

Just to give you a flavor of the sort of questions that people asked afterwards, here are a few as I remember/interpreted them:

Q: (This was for David Bailey, the Google rep.) You showed a snapshot of entering Google’s search results via universal search. How can a video site get included in Google’s universal search?
A: Quite a few sites are already in there, and we would love to open that up more. Factors include things like reliable playability, lack of copyright concerns, no porn, etc. They’ll keep working to expand the sites that can participate.

Q: Do you expect to be crawling the web for videos?
A: Not right now. For the time being, you could submit your video to YouTube or other search engines if you wanted to.

Q: (for Yahoo’s Tim Mayer, I think) Do you expect to use 3rd party rating such as BizRate to help with ratings and abuse?
A: For now, we’re just using ratings on Yahoo.

Q: We have 20-30 videos on our corporate site. We wouldn’t show up in universal/blended/3D search?
A: Not right now, but you could always submit your videos to the different engines.

Q: (for Google) What other types of data do you expect to surface?
A: A good guess is the types of data that we already help search over, e.g. things like patents or code could be interesting, or Google Base has different feeds for real estate and jobs. Data that we already have is the most likely near-term, but most people probably care about well-known types of data like news, video, local, etc.

Q: Different IPs will see different search results? How can I see what someone in San Diego would see?
A: There’s not a great way right now. Some people use proxies.

Note from Matt: Google does provide quite a bit of this functionality. For example, searches as if you’re in the U.S., and returns Bank of America at #1. But does the search as if you’re in the UK, and returns Lloyds TSB at #1.

You can even look at ads based on lat/long, regions, cities, U.S. ZIP codes or U.S. designated market area (DMA). See this way informative post for more details.

Q: What future plans do you have to extract text from audio or video?
A: Everyone was silent for a while. David Bailey of Google gave the only reasonable answer that most search engine employees can give when you ask about future plans: we have researchers that work on such projects, but we have nothing to announce at this time.

Q: What are your three best optimization tips for video?
A: (various panelists answered.)
1. Choose a good title that describes your video
2. Tim Mayer from Yahoo mentioned exploring MediaRSS.
3. Erik Collier from Ask said “Make a kick-ass video.” Well-said, and a good reminder that compelling content makes optimization much easier. 🙂

Apologies if I’ve paraphrased any questions or answers incorrectly. Feel free to comment if you think I got something wrong.

Back from Kentucky

I just flew back from Kentucky, and boy are my arms tired. 🙂 My brother is getting married in early September — woohoo!! — so on Friday my wife and I flew to Kentucky for a wedding shower. We flew back today. It was really nice to see lots of family and friends and wear my Ale-8 T-shirt with pride.

The only real downside (other than all that flying time) is that I’m several days behind on email now. Not the situation that I’d like when going into a multi-day search conference tomorrow, but such is life. If you see me at the conference, please say hello! Also be on the lookout for several other Googlers who will be presenting or hanging out at the webmaster tools booth.

Last but not least, here’s a really useful piece of advice. If you’re attending the Google Dance on Tuesday night at Google’s headquarters, don’t just spend your time chilling with Googlers, dancing, and drinking. There will also be a “Meet the Google Engineers” room in building 43 where you can ask lots of different questions. There will be folks from search quality, search infrastructure, Google Reader, AdSense, Feedburner, invalid clicks, mobile search, and many more teams. Save some time to drop by that room and meet some Google engineers. I plan to be there too.

Talk like a Googler: parts of a url

Let’s dissect the parts of a URL (uniform resource locator). I’ll tell you how we typically refer to different parts of a URL at Google. Here’s a valid URL which has lots of components:

Here are some of the components of the url:

  • The protocol is http. Other protocols include https, ftp, etc.
  • The host or hostname is
  • The subdomain is video.
  • The domain name is
  • The top-level domain or TLD is uk. The uk domain is also referred to as a country-code top-level domain or ccTLD. For, the TLD would be com.
  • The second-level domain (SLD) is
  • The port is 80, which is the default port for web servers. Other ports are possible; a web server can listen on port 8000, for example. When the port is 80, most people leave out the port.
  • The path is /videoplay. Path typically refers to a file or location on the web server, e.g. /directory/file.html
  • This URL has parameters. The name of one parameter is docid and the value of that parameter is -7246927612831078230. URLs can have lots parameters. Parameters start with a question mark (?) and are separated with an ampersand (&).
  • See the “#00h02m30s”? That’s called a fragment or a named anchor. The Googlers I’ve talked to are split right down the middle on which way to refer it. Disputes on what to call it can be settled with arm wrestling, dance-offs, or drinking contests. 🙂 Typically the fragment is used to refer to an internal section within a web document. In this case, the named anchor means “skip to 2 minutes and 30 seconds into the video.” I think right now Google standardizes urls by removing any fragments from the url.

What is a static url vs. a dynamic url? Technically, we consider a static url to be a document that can be returned by a webserver without the webserver doing any computation. A dynamic url is a document that requires the webserver to do some computation before returning the web document.

Some people simplify static vs. dynamic urls to an easier question: “Does the url have a question mark?” If the url has a question mark, it’s usually considered dynamic; no question mark in the url often implies a static url. That’s not a hard and fast rule though. For example, urls that look static like may require some computation by the web server. Most people just refer to urls as static or dynamic based on whether it has a question mark though.