“Google could do more [to stop typo-squatting] if they wanted to. But they profit from each and every click that’s associated with these practices, so they have every incentive to look the other way.”

http://www.theregister.co.uk/2008/10/23/google_and_typosquatting

See: http://gaggle.info/post/98/google-earnings-call-nobody-wants-to-turn-away-a-customer

nmw

NLCS is probably a fairly uncommon way to place the query. Who calls it that, except for people in the media?

Perhaps it’s a bit of a conspiracy theory to suggest Google is purposely making bad search results in order to drive ad clicks. Really this is an example of why search is so hard a problem – everyone has their own way of placing the question, and it’s hard to do a good job on all of them.

thanks for correction. it looks live is one year late…

I apologize that I am not being more clear about this. I am talking about offline.

All throughout this, I’ve asked you multiple times about the method by which query fixes and feature tweaks are prioritized, as a result of whole-page clickthrough behavior. Query fixes and feature tweaks are not something that you do, online, any more than a hand-drawn cartoon could be filmed before a live studio audience. They’re what you do for research. Offline.

So again, let me just requote what I already wrote above:

I wrote: Or, if such a team does exist, can I truthfully say that this team does not communicate with you in any way about their full-page knowledge; they do not give you information about what is happening on the other (advertising) half of the page that could in any way sway the 450 tweaks/year that you make to your algorithms — neither the way the tweak actually works, nor even which 450 of the possible 5,000 tweaks that you could decide to work on?

So yes, you do have a group that looks at the whole page. That’s fine. But when they look at the whole page, what do they do with the knowledge of how the activity is split between organic clicks and ad clicks? Why does anyone even bother measuring it, if the firewall in place ensures that no one ever uses this joint knowledge?

If all they do it just tell you when a change is better or not (presumably by doing some sort of NDCG calculations or whatever), why do they need to look at the whole page? Are you telling me that there is a group that looks at the page as a whole, but then makes recommendations to you based solely on the organic results? Why don’t you just have a team that looks at the organic results? (And then a separate team that just looks at ad quality results?)

Their job is to tell search quality whether a given change is an improvement or not, and even though they can look at the entire page, their job is to help us improve our organic search results, not to encourage us to earn any more money.

I am not saying that they directly encourage you to earn more money. I’m asking about more indirect influences here.

Look: You have to decide at some point what changes you are going to make to your algorithms, and why. You have to collect a bunch of sample queries that are not performing well, or look for new queries that no one has yet examined, and then look for patterns in those queries, to try to create new tweaks.

And what you are telling me, let me make sure I absolutely understand, is that the team that looks at the page as a whole, and feeds you those poorly performing queries.. uses absolutely no knowledge of anything else that either was displayed or was interacted with on that page, to tell you whether or not to work on a particular query? The queries that they feed you as examples for you to work on.. those are chosen by the “whole page team” in a manner so as to have nothing to do with anything else that happened on that page, in that query?

I don’t know how I can be more clear than that.

Just a simple “yes” would suffice, if true. And you’re also welcome to say “bugger off, JG” if my persistence has got you annoyed. But I feel like we keep dancing around the core of my concern, and I’d just like a satisfactory answer.

Cheers.

But you pretty much answered my question anyway, so thank you. You say:

The systems are currently orthogonal and independent, so search quality wouldn’t have the ability to say “Don’t show any ads for this query” in the same way that the ads group wouldn’t have the ability to say “Don’t show any organic results for this query.”

I ask this question because I have this notion of relevance from the end-user perspective.. how ultimately all the user cares about is meeting his or her information need.

And in service of that information need, there must exist a theoretical best ordering of all possible information available online — web pages, videos, images, and yes, even ads.

And in that best ordering, it is also very possible that the best advertisement has a relevance score, or a relevance value, equal to (let’s say) the 73rd-ranked web page. That is to say, the top 72 web pages are *all* more relevant than the rank#1 advertisement.

So it seems to me the “right” or “non-evil” thing to do would be to wait to show an advertisement, until the 72nd organic result had been reached by the user. And filling the remaining space with the more relevant organic results.

By instead promoting an advertisement to the very top of the results page, and putting it on (separate but) equal par with the top organic result, you are really not serving the user’s best interests. You’re cluttering up their interface and wasting all that valuable space.

Note that I am not talking about showing fewer ads. I’m still talking about showing the appropriate, relevant number of ads. But you should wait to show them until there no longer exists more relevant organic information. You shouldn’t automatically promote every single ad to the very top of the right column. It’s an anti-user, anti-relevance stance, imho.

And note that I am also not talking about mixing the ad in with the results. No, still keep them separate and clearly delineated. I’m just saying that the ad, even in its separate position, shouldn’t slide its way up to the top, when more relevant organic information is available.

Stated in another fashion, I am essentially asking you why Google does not dynamically resize the amount of information presented on the first page. Why is it always the exact same format.. 10 objects. It’s almost like there is no awareness on Google’s part that some topics and queries are smaller, and some are much better. It always only 10, every single time.

(And I know, I can set the default to 50, or 100, but that misses the point. If I reset the number of links, then it would be always 50 or always 100. And it would still not dynamically resize based on my query.)

Xin, look deeper. That #1 MLB result on live.com is from 2007 (last year), not this year.

Book, we think of ads as useful-but-independent information in their own right. The ads team tries to show more relevant (which can also mean fewer) ads, and the search quality group tries to show more relevant search results. The systems are currently orthogonal and independent, so search quality wouldn’t have the ability to say “Don’t show any ads for this query” in the same way that the ads group wouldn’t have the ability to say “Don’t show any organic results for this query.”

We have done tests to see whether showing 20 results is more or less helpful (it tends to be less helpful). We’ve also run tests to fit more information into the same amount of space, e.g. doubling up our video boxes for a search such as [matt cutts].

JG, it’s unclear if you’re talking about online (as the search results are being computed) or offline (e.g. doing new research on ways to improve our algorithms). If you mean online, I don’t think my webspam group’s code has any knowledge of whether or how many ads might show for a query. I believe the query is sent to the ads system and the indexing system independently and then merged at the last millisecond before showing the results to the user. See the Google Cluster Architecture paper that was published not long ago at http://www.scribd.com/doc/312186/THE-GOOGLE-CLUSTER-ARCHITECTURE for example.

We certainly do have a group that evaluates the whole page of search results for quality, but not in the way you are worried about. Their job is to tell search quality whether a given change is an improvement or not, and even though they can look at the entire page, their job is to help us improve our organic search results, not to encourage us to earn any more money. You can read about the search evaluation group at http://googleblog.blogspot.com/2008/09/search-evaluation-at-google.html to see that it’s not doing the job you’re worrying about.