Google Algorithm - What would you do differently?

[irldonalb]

Hi

We all have a good idea of what's in Google’s algorithm. Matt C alluded to the fact there are over 200 variables in it. We attempted to create a list [webmasterworld.com].

However I was wondering if you had control of the algorithm, what you'd do different? Where are the flaws?

Mine surrounds the lack of punishment for "Paid Links". If I can tell which of my competitors are buying links, surely Google can tell.

In addition, feedback is important. I know it's improving in Google'’s WMT but it’s too little and too slow.

Regards
Donal

[tedster]

I don't understand the inner workings of the Google algo well enough to suggest to much about what I would do differently. I can't imagine the kind of pressure that web spam puts on algo development.

One thing I have always thought they should do if they don't want paid links to influence ranking (and I agree that they shouldn't) is just identify them and don't let the link pass any juice, rather than handing out penalties.

[aristotle]

They should giving a special artificial rankings boost to Youtube videos.

[maximillianos]

Penalties for gaming the system should be permanent. If a site achieves a certain amount if success and inbound links from traffic it received via black hat tactics, the penalty against that site should never be lifted, or at least some sort of penalty should remain to be fair to all other sites in that niche who are honest businesses. Or perhaps all past inbound links should be nullified after the penalty is lifted. Or at least a percentage of them since they were acquired during the time the site was gaming SERPS.

[enigma1]

Penalties for gaming the system should be permanent.

Yes, but the problem is sites who follow these tactics are new for the most part. So they'll just get another domain and do the same again. If the algorithm can't filter them in the first place (dup content, spam, fraud etc) the problem will persist.

Then with old domains you can't easily tell when the administration changes in which case a penalty won't be effective in the long run.

[futureX]

Mine surrounds the lack of punishment for "Paid Links". If I can tell which of my competitors are buying links, surely Google can tell.

But can you write an algorithm to sort paid links from organic links automatically?

[jdMorgan]

I would modify the PageRank "link juice" calculation so that it was based not on how much "link juice" flowed to a page, but rather on how much flowed through a page.

This would put a stop to the silly "nofollow" linking craze and the PR-hoarding that is Balkenizing the Web.

I'm sure that some kind of balance between "to-juice" and "through-juice" will be needed, since there are some legitimate kinds of pages that won't have many outbound links. But if the Web is to remain a Web instead of becoming a huge collection of end-stubs, then that's a fairly easy fix.

Jim

[mrguy]

I would get rid of visible pagerank.

That would put a damper on a lot of the link selling since nobody would know what the percieved page rank of a page is.

No normal surfer ever looks at page rank anyway and the only people who do are those obsessed with page rank.

[dvduval]

What's dangerous to me is people are losing interest in google, especially webmasters. Instead they are moving to places like Facebook. It will take a couple of years for the full effect to be known, but I feel that in their attempt to squash spam through a "set it and forget it mode", they then also forget that some of their most frequent users are webmasters. This is dangerous. There is really little to talk about with the algorithm right now.

[Sgt_Kickaxe]

Google has a two-tiered search engine already, by that I mean they have short term rankings and long term rankings. They already do an excellent job with the short term (finding/indexing new content) but the long term is far too often dominated by sites/pages that aren't the best on an individual basis.

Too much weight is given to major websites. Too much weight is given to companies with physical addresses. Too much weight is given to a dozen other factors that lead to "not the best page" being returned at #1. Until Google finds a way to rate actual on page content on its own WITHOUT requiring outside influence factors I doubt that will change.

I wouldn't change anything until that point is reached, change for the sake of change is often a bad idea as illustrated by Mayday.

[brysonmeunier]

Great question. Agree with a lot of what's been said about paid links. I've always thought that Google's core search ranking algorithm gives too much weight to exact query in domain. It's too easy to buy a two or three keyword domain with a lot of query volume, pay attention to title tag and body content and rank above the fold with little to no link popularity. I've heard this from a number of webmasters and am surprised that it's not here yet.

[irldonalb]

But can you write an algorithm to sort paid links from organic links automatically?

FutureX - I couldn't write it but I'd be very surprised if Google couldn't. Websites that sell paid links have certain characteristics. They should be easy to find. Then if a website's have an unproportionate amount of links from these websites, I think it's safe to assume they've paid.

There is a risk that you could buy links to your competitors. But not if you adopt Tedster's approach of ignoring these links instead of penalising for them.

I've always thought that Google's core search ranking algorithm gives too much weight to exact query in domain.

brysonmeunier - I agree with that completely. However I've often wondered if it's because of the keywords in the domain name or because that domain gets links naturally with the keywords in the anchor text.

[supafresh]

I want the google algo to devalue links from other countries.

If a site has 3000 links from China it should show up well in Chinese SERPS and not USA Serps.

It should however to keep things fair have some weight on USA Search.

[alika]

However I've often wondered if it's because of the keywords in the domain name or because that domain gets links naturally with the keywords in the anchor text.

I have a 1 week old website where the only links come from my 2 other sites and my Twitter posts (no retweets yet or tweets about my new site), but it is already #1 for its search term which is its domain name. This is a fairly competitive keyword with about 306 million search results.

[brysonmeunier]

I have a 1 week old website where the only links come from my 2 other sites and my Twitter posts (no retweets yet or tweets about my new site), but it is already #1 for its search term which is its domain name. This is a fairly competitive keyword with about 306 million search results.

My experience is similar. I have a year old site that ranked on the first page for its target keyword phrase in a week and has held the #2 position out of 800k results since the second week ahead of Boing Boing, AOL, the official site, and a number of other sites with a great deal of authority, simply because I controlled the order and placement of keywords in the domain and title, and put relevant content in the body. This is what Google Webmaster Tools says about the link graph:

We do not know of any links to your site. If your site is new, we may not have found any links yet, or other sites may not link to yours. Check back soon for more data.

I would imagine that some sites do get more links because they have the exact phrase in the domain, but in this case controlling the domain and title is enough. Even though I'm benefiting from it in this test case, it's not really fair to larger sites that can't change the domain every time they want to indicate relevance, and in my view should probably be deemphasized a bit.

[vordmeister]

I would return results relevant to the search. It is becoming increasingly difficult to search for something obscure in Google without having to scroll through pages of results for something they think is similar but isn't. I would prefer they just say "sorry we don't have anything on that".

I did a very specific search the other day and they changed one of the words and gave me pages of results about motorcycles. I wasn't searching for anything with wheels.

They used to be better in this respect. I'd change to Bing but they seem to be doing the same!

[blaze]

Google needs a 'credible' or 'relevant' button that people can put on their web pages ala like.

From there they could create a social network around websites you believe are relevant.

Everyone would of course put this button on their website and crowd out the like buttons.

[wheel]

Great question. Agree with a lot of what's been said about paid links. I've always thought that Google's core search ranking algorithm gives too much weight to exact query in domain. It's too easy to buy a two or three keyword domain with a lot of query volume, pay attention to title tag and body content and rank above the fold with little to no link popularity. I've heard this from a number of webmasters and am surprised that it's not here yet.

It has to be that way, they've already learned this. It's because your domain name can and in many cases should be, your company name.

If they changed this feature (and I believe they tried this at one point years ago) then when people do searches for 'nike', they may not get nike's website - when it's clearly what people want.

Plus, there is every expectation that a keyword rich domain name is going to be very on target for an exact search - so despite people's complaints about how those domains rank well, in most cases Google's actually providing relevant results. Too bad it's not your website is all :).

There's an easy solution. Go get your own keyword rich domain. I've got a couple. they're on target, and provide great traffic.

[incrediBILL]

All paid SEO-only directories would be penalized along with every site listed in them.

These directories are easy to spot as they don't rank for anything that would pass traffic to their members, they never attempt to rank for anything other than being a paid directory.

[Edge]

1) Less search results emphasis on domain-keywords and rank web sites on content.
2) Less search results emphasis on universities (.edu) results. Again let the website stand on the content. .edu websites are typically about the educational institution, rarely about useful content search results.
3) Less search results emphasis on Wikipedia – enough please…
4) Get over the “Paid-link” issue… The longer I look at this issue the more it looks like trying to control advertisers and funneling them into the adwords system. Let the free market rein!

[jimbeetle]

One of the things I might have done back in the early days would be to consider some sort of get together with Inktomi, though I don't know if either would have been in a position to purchase the other.

Ink was the first SE to use links as one of its major ranking factors and had much more experience than G in fighting link spam -- and wasn't afraid to pull the plug on any sites that crossed the line.

It took a few years for Google to develop any expertise in this and whatever it deployed still seems at times to be applied in a hit or miss manner.

[jimorandy]

I think the search engine algo should rely on links for NO more than 30% of "ranking". Most of it should come from user behavior.
I think Google is already at about 50/50 - I mean, 50% of ranking comes from links and 50% from user info (click through rates, pages viewed per visit, length of visit). Intimately, your website is for users. Whatever "metric" to measure your visitors satisfaction is what Google should use to rank your website.

[MWpro]

Not artificially boost Google services in the rankings such as blogspot... I cannot stand this and is what I hate most about google.

[buckworks]

All paid SEO-only directories would be penalized along with every site listed in them.

A new opportunity for competitive sabotage? Oh, such fun!

[indyank]

and has anything changed recently in google algorithms...I used to get into "Google SEO News and Discussion" through the sitelinks that appear under webmasterworld.But I don't see this particular sitelink for webmasterworld now..

or is it that the sitelink has been blocked through the GWT here...

[indyank]

If a site has 3000 links from China it should show up well in Chinese SERPS and not USA Serps.

I don't think regional treatment is good for internet.What makes internet so great is the ability to get to so many pages from different countries and with different culture and get to know the best...if such regional treatment is given, then the end users may not always get to see the best things in the world...

2) It would also mean that big sites with servers all around the world will always dominate the web and this isn't good either...

[indyank]

Not artificially boost Google services in the rankings such as blogspot... I cannot stand this and is what I hate most about google.

I think there is a kind of reservation system being applied to blogspot blogs. the first page in google has one blogspot position reserved for any keyword and if any blogspot blog can have the right kind of content and link juice, it is allocated to that blog post...

i have seen many blogspot blog with exact keyword match usually fitting into this place...but it does have some decent and relevant content for the keyword...

[tedster]

sitelinks that appear under webmasterworld

Google algo change [webmasterworld.com] - currently this forum is at position #4 for the same search.

[roycerus]

Ban adsense accounts which are MFA's [one line text with ads] - take the help of the community to identify. That will reduce web spam by many factors. Google is funding their own demise for a few bucks.

[SEOPTI]

Real time results, they are late, tennis or basketball results, all delayed by 30-45 minutes ... this is slow as hell.