Ny Times Article On Google's Search Team, Best insight I've seen on how the search team works Options

[donp]

The NY Times has quite an article on the inner workings of Google's search team - doesn't give away the store of course, but very interesting none the less.

NY Times Article -Sunday June 3rd

This post has been edited by donp: Jun 3 2007, 09:16 AM

[qwerty]

I bet we're going to see a lot more use of the terms "signals" and "classifiers" on the blogs after this.

[Bradley]

Great article, one I'll certainly point clients to..

[Michael Martinez]

Freshness and diversity would be the more critical hints the article shares, in my opinion. I doubt I'll pay attention to anyone who blogs about "signals" and "classifiers". They won't be saying anything useful. But if someone wants to speculate about freshness and diversity, point me to their blog.

[qwerty]

That's more or less my point. A lot of people seem to think they can pass as experts because they know an inside term or two.

[jodo]

Lots to think about it in the article which I thought was very good for a general interest piece.

If I understand the article correctly and if the info is good; Google is using personal info for searches if one simply uses e.g. gmail:

Increasingly, Google is using signals that come from its history of what individual users have searched for in the past, in order to offer results that reflect each person’s interests. For example, a search for “dolphins� will return different results for a user who is a Miami football fan than for a user who is a marine biologist. This works only for users who sign into one of Google’s services, like Gmail.

I use several google services and I'm wondering if the results I'm seeing are anything like others see who have different business and personal interests. The article implies their search results are different. So, importantly, you don't need to sign up for personal results to get them.

Is this why I see so many of my sites pages ranked so well? Because they know I'm interested in them? I always thought it was because I followed the good advice I found here on HighRankings. (IMG:style_emoticons/default/whitehat.gif)

Also very interesting was this:

“If you have a lot of different perspectives on one page, often that is more helpful than if the page is dominated by one perspective,� Mr. Cutts says. “If someone types a product, for example, maybe you want a blog review of it, a manufacturer’s page, a place to buy it or a comparison shopping site.�

This is as I suspected and it means in general there are really only about three positions available on the front page for a given type of web site/page. This explains why the shopping web sites can rank so high even when the page is sometimes actually empty of any product for sale.

And most amazing is they tweak the algo virtually everyday. What fun.

[qwerty]

If I understand the article correctly and if the info is good; Google is using personal info for searches if one simply uses e.g. gmail

Yes. As long as you're signed in to any of their services, personalized search is on by default. I'm pretty sure that was made public a few months ago.

[redsonia!]

Aha! So this means that if a person is using any of the Google services the results (thereby site rankings) they see will be completely different from what another person sees. So if you use a computer at home where you primarily search for shopping sites and comparison sites, any search query you use on that computer will be "biased" toward shopping sites. And if you use a computer at work primarily to find information, the same query you used at home will return a different set of informational websites at your work computer. So basically there will not be any way for a business to consistently rank on page 1 of the results for a given key phrase - because the SERPS are always going to be fine tuned to the individual searcher's perceived preferences.

Therefore, you can only make your site the best it can be and hope (IMG:style_emoticons/default/crossfingers.gif) that your website will come up on the first page of the organic SERPS for as many people as possible. I think this means that PPC advertising is going to be MUCH more important if you want to show up in the SERPS. Ah, better fine tune those PPC campaigns!

It also means that it will be harder to prove to clients that they rank on the first page. Because the SERPS that the SEO sees will most likely be significantly different from what the client may see on his/her computer. Ooh, what a headache. (IMG:style_emoticons/default/ohno.gif)

[OldWelshGuy]

Redsonia, you have hit the nail firmly on the head there and THAT is good news.

Clients measure success by Position, and it is completely the wrong metric. RESULTS are the the only metric. They want to be o #1 for a reason, what is that reason? find it and use THAT as the metric.

IE sales, conversions, signups etc.

I have been a long term fan and if you go back 2-3 years on this very forum, there is a post where I set out what I concidered to be the future of search. I stated then that it would be personal search where people can opt out of sites, opt in to groups, remove shopping sites, remove directories etc. And that personal search would be personalised through algorithmic refinement.

That future is now here, and just around the corner to becoming mainstream.

[Nueromancer]

The NY Times has quite an article on the inner workings of Google's search team - doesn't give away the store of course, but very interesting none the less.

NY Times Article -Sunday June 3rd

The quote:

"it has to double-check the engineers’ independent work with objective, quantitative rigor to ensure that new formulas don’t do more harm than good."

Made me smile - not quite sure a lot of people icluding meself would 100% agree with that comment more wack it out and see what breaks is how I think they work.

[projectphp]

I disagree Necromancer. They are always testing, have several datacenters (all of which can run slightly different algorithms) and they always test UI changes extensively before they become standard features. Seems to me like the ideal format for, and us of, extensive testing of changes before they go live, and seems like they do practice what, in this case, they preached.

Besides, the fact that many people are caught by changes kinda proves that the changes worked, as the goal would always be to filter something and someone.

This post has been edited by projectphp: Jun 4 2007, 11:40 PM

[Lakshmi Narsimha...]

OldWelshGuy, you must be a Search astrologer. (IMG:style_emoticons/default/appl.gif)

It is one of the many different factors that Google calls as “signals", from the NY article. G also considers many other signals as on Web pages — like words, links, images, data patterns uncovered in the trillions of searches that Google has handled over the years, all the links going to a page, how the content is changing on the page over time and also it states that Google is using signals that come from its history of what individual users have searched for in the past, in order to offer results that reflect each person’s interests.

This should be the areas of a website where one should focus on to gain good ranks on Google.

[Nueromancer]

I disagree Necromancer. They are always testing, have several datacenters (all of which can run slightly different algorithms) and they always test UI changes extensively before they become standard features. Seems to me like the ideal format for, and us of, extensive testing of changes before they go live, and seems like they do practice what, in this case, they preached.

Besides, the fact that many people are caught by changes kinda proves that the changes worked, as the goal would always be to filter something and someone.

Ime sure they do test it is just that i don't belive its "objective, quantitative rigor" and i dont call punting sofware out to j random datacentre and letteing the users do the testing objective, quantitative rigor either.

I would expect better in house testing (this has areal efect on peoples jobs after all) I once ran a system in parallel for 9 months when I was working on billing systems for BT - and that was just the baby one the big systems somthig like CSS is planed to the nth degree and missing a drop is not somthing that happens.

[projectphp]

Then what is mate?

I once ran a system in parallel for 9 months when I was working on billing systems for BT

False comparison, as the % for success is very different, and search is not a closed, known system. What constitutes a bad transaction is known, but a "better" algo? What is that?

And the SE that spends 9 months goes broke. Really, that is just an old software model that is outdated for the web. But I think we just have to agree to disagree on what constitutes good testiong in this circumstance.

[Nueromancer]

Then what is mate?
False comparison, as the % for success is very different, and search is not a closed, known system. What constitutes a bad transaction is known, but a "better" algo? What is that?

And the SE that spends 9 months goes broke. Really, that is just an old software model that is outdated for the web. But I think we just have to agree to disagree on what constitutes good testiong in this circumstance.

no test in parallel I ment and not for 9 months but if they expicitly use terms ""objective, quantitative rigor" that implies they do such
testing.

Google should be able to have a big enough set up to the on the same set of data with old and new algo/heuristics and check that the desired result of the change and look for any un expedted changes - which they can the analyze and do a cost benefit anlysus

You could automate this I suspect - A bit like big well organised teams have nightly build and smoke tests for teh developers to come in and work on in there morning.

One of my client't had to lay off staff becasue of algo changes so unexpected consequnces do have a cost.