Forum Moderators: mack
Search for <snip> - 9 of the top 10 is <snip>.com....almost all serps seem to look pretty much this way...url only titles and just poorer than ever.
[edited by: Brett_Tabke at 7:34 pm (utc) on Feb. 16, 2006]
[edit reason] I bet even <snip> itself has bad results ;) [/edit]
Create keywords subdomains, spam blogs, rule the pitiful MSN spam engine.
As god awful as those other results were, they were probably still superior to these roled back ones. At least those other ones seemed almost random. The rolled back ones remain universally miserable (if the terms are really targeted by anyone) in how one simpleton tactic works.
The homepage is # 1, # 2-9 are all subdomains from the same site.
what the heck is going on?
C
Correction rolling out?
If I may quote David Lee Roth: "Big Bad Bill (Gates) is sweet William now." These results spark zero fear in the industry.
Cheers,
CaboWabo
Cheers,
CaboWabo
We did roll out an improvement to RankNet just a few hours before the thread started (you guys are quick!), but the tests we used to qualify that net did show a small improvement. We saw the negative responses on WebmasterWorld almost immediately, but our own personal “sanity-test” queries (mostly technical and scientific) worked fine, so we assumed it was just a few people grousing about their over-optimized sites getting hit. However, as time passed and the thread continued to be a) very busy and b) nearly 100% negative, it became clear that something was wrong. Deeper evaluation revealed that the problem was in the qualification test itself.
We rolled back to the old net after only about 48 hours. The new one might have been up considerably longer if you guys hadn’t been so vocal so quickly.
So we’d like to thank all of you for your feedback – no matter how negative it was.
On another note we did roll out some other changes as part of the release and thankfully we have not had the need to roll those back :) Here is a summary of some of the more visible changes that we made or changes that you all would likely detect.
User Experience: We lightened up and streamlined the UX a bit. Thanks for the positive feedback on this: [#*$!.com...] Isn’t the super-sized search bar at search.msn.com also refreshing?! That is a big search bar and it makes a big difference :)
Snippets / Contextual Descriptions: We made some changes to how we create contextual descriptions / snippets for pages. One clearly visible change is that we are now doing hit-highlighting in the title. We also have more subtle improvements around using page structure to get a true summary of the page. A decent example of this at: [search.msn.com...] or [search.msn.com...] You will notice that the descriptions tend to read like real sentences. This was not always the case.
Depth of crawl: Over the past month or so we rolled out an improvement that will allow us to crawl high quality domains more deeply. We are generally pretty content with the improvements we have seen. We still have a lot of work we want to do to improve our selection, however, this is a step forward. If you have any feedback on this in terms of what you are seeing on your end we would love to hear it.
Keep the feedback coming.
- msndude(msd) with input from the techy RankNet folks
Glad that depth of crawling is being addressed - hopefully improvements in this area will continue.
In another thread I pointed out that at the moment I cant get past page 25 of results - are you aware of this - or is it just me? Something to look at perhaps.
Another thing I cant understand is that pages that appear in a site:domain.com search dont appear in a normal serp search for the keyword - even if I go to the end of all results - as if the page disappears - would you like examples of this?
One thing which is still bothering me most is that the first page in some finance searches contains only one site.
This can be done with a single entry and "More results from www.anysite.com"
Just a gentle reminder - we don't encourage links to blogs - except maybe yours and Google Guy's ;) - Your post didn't even need manual intervention to get expleted out!
Yep - the guys here are quick. You could give us an hour's warning and you'd have a running commentary as the rollout goes live. People who make their livings out of providing the most relevent content for the most relevent search query (and those making a living providing the top result regardless of relevency!) aren't going to hang around until the dust settles... even if they should at times imho! so when a factor changes in our environments, people all over the world are on your case quite quick.
Thanks again for pitching in and giving the feedback MSNDude. Revising that search quality testing is something I am sure the guys here would have a view on as well if you wanted to put such a thread into the field...
Dixon.
Those searches I sent you last week (per your request) are still showing ALL subdomains with the same spammy title.
Are these going to be removed?
It's nice to know there's at least one search engine that's listening and reacting to feedback. Thanks!
The results themselves are of course shameful.
There is more to life than exact match of duplicate subdomains.
(and I'm still wondering how MSN can list a non-www URL that hasn't been accessible since before the search engine existed)
I noticed that only the top 250 results are shown for searches as well, Dayo_UK.
Also I have noticed deep crawls and some serp's I monitor are producing deep level result sets which in most cases helps the end user really find what they are after. It definetely is not conducive when you see a site's default index page ranking for something that is 5 levels deep in the site.
Thumbs up MSN, your headed in the right direction!
I realize this is cosmetic, but I like it - much improved.
I'm also glad to see a SE take interest in the comments here. Clearly there were some results that just were not reasonable.
Thanks for listening.
One thing which has bothered me for a very long time when using MSN to search is how far off the results can be when the order of the keywords being searched for means the difference between good and very bad results. I wish I could provide specific search terms here, but its against the TOS for WebmasterWorld.
For example:
The following is not a real search terms folks ... its just an example because I can't use a real one!
A search for "miss muffet" serves up results for "muffet miss".
The searcher was looking for pages relating to the nursery rhyme but instead gets results for a bakery in Dudangania. Its frustrating!
MSN needs to have a good long look at how it evaluates and interprets the order of keywords in a specific search.
There are all sorts of marinas where I live, there is also an island actually named "marina something". If you search for the island on MSN ... you have to dig deep to find it even though the word "marina" comes first in the island name and all marinas here have the word marina second in their names. So why can't a simple thing like the order of keywords used by a surfer be addressed?
I have sent a feedback form giving a specific search phrase for what its worth!
-msndude
Thanks for coming in on the thread MSNDude.Glad that depth of crawling is being addressed - hopefully improvements in this area will continue.
In another thread I pointed out that at the moment I cant get past page 25 of results - are you aware of this - or is it just me? Something to look at perhaps.
...but our own personal “sanity-test” queries (mostly technical and scientific) worked fine...
msndude - Thanks for your feedback. It strikes me that this may provide a clue to some of MSN's quality problems. I'm glad you listened and made the adjustment you did. Current results are also superior, I think, to what you had a week or two ago.
But results in scientific areas are not likely to be spammed to the extent that you'll see in the more competitive commercial areas. I think that MSN really needs to beef up things like link quality evaluation. To do this, you're not going to able to apply one-algo or one-sanity-test fits all to the entire web.
See a brief exchange on MSN quality evaluation (with regard to your very bad results of last week) in this thread [webmasterworld.com] between Receptional and me. I mention in msg #24...
The thought hit me the other night that they might be varying their results to test out various parameters and then trying to use some automated measure of user satisfaction to feed back into the system and readjust the algo... in other words, sort of a combination of #1 and #3, but perhaps with some randomization thrown in.As I was spinning my wheels about this, one of the things I wondered about was whether, if they did this, they were topic-area specific, or whether they might be applying an "algo" derived from an overall web sampling to every topic area (or, vice-versa).
It appears that you'd done the "vice-versa"... ie, taken a sampling of satisfaction from a small topic area and applied it to the whole web.
Your algo of a couple of weeks ago, it seems, had cranked up the anti-optimization measures so high that it was discriminating against classically optimized pages on reputable sites and showing a lot of random junk instead. I'm pleased to see what you have now, but long-term, I think, you folks need to do better.
but our own personal “sanity-test” queries (mostly technical and scientific) worked fine
That is how you tested? Seriously? I was a software tester for 9 years before getting into SEO and if we would have done our testing like that, we would have been fired.
This is not a step forward. Searches for VERY common items bring up a huge amount of domain spam. This is not complaining against any of our sites getting hit with this update. Honestly, we get such little referral traffic from your engine that isn't the case. We want to see you do well. We need the competition for Google and Yahoo! not have the power they do ... a ban from Google these days nearly does in a web-based business.
It has been a year since you have released your engine and this is the state it is in?
Disappointed and in the need of a Waborita.
CaboWabo
Lol - to see my pages so intelligence data then :) - unlikely many users would want to go beyond page 25 :)
Fair enough if you stop at 25 pages. I would have thought from a user point of view it would be better to take the links to page 26,27,28,29 out and take out the next link when you get to the 25 page - but that is just cosmetic.
