June 2005 MSN Search Update & Neural Net Tech

[ephricon]

For one of my clients I just noticed alot more traffic and correspondingly (word?) improved rankings for some reasonably competitive terms. This is well deserved IMO as its a great site and extremely relevant, but I have to question why all the good news now where as two days ago the top 10's were top 30's and traffic was much lower. Small sample size but today and yesterday MSN traffic is up on this one site by a two to three times what I would normally expect.

Anyone else notice anything?

[CaseyC]

I wish we were as fortunate as you... we are seeing a steady decline in our MSN numbers. Our indexed pages have been cut by nearly 60% this week and they are dropping daily. We have experienced very little movement in Google or Yahoo, but MSN has gone straight down the tubes for some reason.

[JohnW]

Something is going on with many of my sites too, not good or bad but noticeable. I am seeing pages disappear from the top 30 (didn’t check any deeper) but in most case are being replaced in similar position by other pages from the same site. And some pages have improved, some new pages have made an appearance.

[mickisdaddy]

My MSN experience has been up and down in the number of pages in their index. At one time my site was up to 2,000 indexed pages, then it dropped down to about 700. I am now back up to about 900. My site log shows MSNbot hitting my site every day. It seems like their index has a leak in it or something.

Edit
A lot of the pages that have been dropped were ranking pretty well for the term I was shooting for.

[CaseyC]

mickisdaddy, sounds like you are having the same problem we are. Did you make anychanges to your site to help get some of your pages back? When can not come up with anything else to do to help with this.

[mickisdaddy]

I did not really make any changes. The number of pages in the MSN index has just been like a yo-yo. MSNbot started indexing my site right about the time of my last major re-do of my site (about 4-5 months ago). The number of pages in their index started going up a little each week then all of the sudden bam it dropped by like 1,000 pages. It started to go up again and then bam it would drop back down.

I see MSNbot daily in my logs hitting like 100 or so pages a day I have not really looked at what pages are crawled versus what pages are in the index. I may have to analyze that and see what I can figure out. (ie what pages are actually making it into the index and if I can what pages are dropped).

[CaseyC]

Sounds like the exact same thing we are experiencing. Please let us know if you find anything out.

Thanks
Casey

[Marcia]

Just today I noticed that the MSN traffic for one of my sites went through the roof. Best I can tell it started yesterday.

[fredmcain]

Originally Posted by CaseyC

I wish we were as fortunate as you... we are seeing a steady decline in our MSN numbers. Our indexed pages have been cut by nearly 60% this week and they are dropping daily. We have experienced very little movement in Google or Yahoo, but MSN has gone straight down the tubes for some reason.

Casey,

I just joined this group. I was hoping someone on the group could give me some advice. We have a website with a Route 66 theme. It took us over a year and a half, but we finally worked up to page 2 on MSN based on the search words "Route 66". We were ecstatic when we made page one over this last weekend!

Then, yesterday morning much to my dismay, we got bumped all the way back to page 6. How and why can that happen? We changed nothing on our site that I'm aware of. My question is, is there anyone at MSN that we can complain to? Or, is there any recourse at all?

[dannysullivan]

MSN's just posted that they're using some new technology to rank results, so that may explain what people have been seeing. Details from MSN here: http://blogs.msdn.com/msnsearch/arch...21/431288.aspx, and I've blogged some comments here: http://blog.searchenginewatch.com/blog/050622-082709. In particular, they make mention of a neural network technology that may be similar to what this research paper (PDF file) describes: http://research.microsoft.com/~cburg...ML_ranking.pdf

[fredmcain]

Would it do any good if we all bombarded them with complaints? Of course, I guess those who saw their rating rise wouldn't have much to complain about.

But if someone really wanted to complain, what e-mail address should they use?

[Marcia]

Would it do any good if we all bombarded them with complaints?

No, not if it's based on sites ranking because there would be a corresponding number of people satisfied with their new rankings.

Of course, I guess those who saw their rating rise wouldn't have much to complain about.

Exactly, they'd say "Great job, MSN!"

[fredmcain]

Originally Posted by Marcia

No, not if it's based on sites ranking because there would be a corresponding number of people satisfied with their new rankings.

Exactly, they'd say "Great job, MSN!"

Marcia,

It's just frustrating to us. We went through this very same thing on Google. After about a year, we finally worked our way up from page 50 or something absurd to page 5. Then Google made somekind of a change and botta-boom! We got slammed back to page 18. Now, 6 months later, we've worked our way back up to page 5 again where we hope to remain until Google makes their next change.

Now, MSN has done the same thing. I don't think this is fair. For example, should a reference to a mass transit website that just happens to have a bus or trolley line tagged "Route 66" be ahead on a search engine of a website that is really about Route 66, The Highway? That just doesn't make sense to me.

Is there something that we should be doing different that we're not doing? I'd like to get on page one (or at least page two) and stay there!

Regards,
Fred M. Cain

[lorenbaker]

Definately seeing some keen results in the new MSN RankNet rankings. Looked a bit more into the paper that Danny linked to and trying to pinpoint some patents on the technology.

Two Microsoft patents dealing with RankNet and Neural Net technology which may be relevant to RankNet:

The first patent identified is Method for Scanning Digital Info Content which mentions neural net in the Abstract :

Computer-implemented methods are described for, first, characterizing a specific category of information content--pornography, for example--and then accurately identifying instances of that category of content within a real-time media stream, such as a web page, e-mail or other digital dataset. This content-recognition technology enables a new class of highly scalable applications to manage such content, including filtering, classifying, prioritizing, tracking, etc. An illustrative application of the invention is a software product for use in conjunction with web-browser client software for screening access to web pages that contain pornography or other potentially harmful or offensive content. A target attribute set of regular expression, such as natural language words and/or phrases, is formed by statistical analysis of a number of samples of datasets characterized as "containing," and another set of samples characterized as "not containing," the selected category of information content. This list of expressions is refined by applying correlation analysis to the samples or "training data." Neural-network feed-forward techniques are then applied, again using a substantial training dataset, for adaptively assigning relative weights to each of the expressions in the target attribute set, thereby forming an awaited list that is highly predictive of the information content category of interest.

And Chris Burges, mentioned in the MSN Search Blog post and head author of the Learning to Rank with Gradient Descent paper, was one of the co-authors of this patent application which describes neural network; "System and method for identifying content and managing information corresponding to objects in a signal." The abstract:

An "interactive signal analyzer" provides a framework for sampling one or more signals, such as, for example, one or more channels across the entire FM radio spectrum in one or more geographic regions, to identify objects of interest within the signal content and associate attributes with that content. The interactive signal analyzer uses a signal fingerprint extraction algorithm, i.e., a "fingerprint engine," for deriving traces from segments of one or more signals. These traces are referred to as "fingerprints" since they are used to uniquely identify the signal segments from which they are derived. These fingerprints are then used for comparison to a database of fingerprints of known objects of interest. Information describing the identified content and associated object attributes is then provided in an interactive user database for viewing and interacting with information resulting from the comparison of the fingerprints to the database.

[hiero]

I'm finding that MSN is finding new pages and changed pages much quicker than Google or Yahoo. Is anyone else seeing the same?

[bethabernathy]

I just went through my reporting for May / June and all my sites are still ranking on the 1st page of MSN for designated search phrases (most in position 1-5). I did notice on 6/15 that one client had their positioning drastically drop, but as of today their position is back to normal.

In looking at what search engine brought in the most traffic, overall MSN provided 1/3 of what of Google and Yahoo did (this is for May, so I guess that doesn't count). I'll look at June at the beginning of July.

Whatever their doing - no problem for me - except for on and around 6/15/05.

[niceland]

Hi all

I have been checking out the MSN after the new update and it is much better and results are more relevant. Some of my sites are doing better and some worse but overall much better.

On the blog MSN says they are developing and improving. With this speed of improvement MSN will soon go past Google. MSn is already indexing faster, and deliver very relevant results and it is more fun to use with all the options, and last but not least I can not see any evidence MSN penalize sites in the same way as Google and Yahoo.

[sem4u]

I am not seeing a huge amount of changes yet, but it is good to see inanchor and linkdomain searches available. They just need to get the anchor text search to work correctly now...

[PhilC]

What has to be realised is that search engines never get it right, and they rarely improve their results via algo changes. For instance, Google came to the forefront because of the relevancy of its results.Since then, they've made major changes that have had the effect of shuffling the serps around and generally pushing some less relevant results nearer the top at the expense of some relevant ones - but they've never improved the serps, imo - they've only caused them to deteriorate bit by bit.

The MSN engine is less that a year old, but already they are incorporating a new ranking technology. It's something that search engines feel they have to do, and, in the long term, they do have to do it. So they shuffle the serps around, cause some very relevant pages to sink, and probably cause some other relevant pages to rise along with some non-relevant pages. It's par for the course, and we have to live with it. It gets annoying when relevancy takes a back seat to new ways of ranking pages (e.g. I posted about some post-Bourbon Google results where 10 of the top 20 have nothing to do with the searchterm), but we have to accept that it happens, that the engines really do get things wrong, and we have to work from whatever position we find ourselves in at any particular time.

Having said that, they do provide us with links to tell them about poor results, so there is something that we can do. It isn't going to make any difference in the short term, though, so we still have to accept what happens, and get on with whatever situation we find ourselves in.

Current thinking has said that meta tags are no longer paid attention by search engine spiders. Not true. Definitely not true!

If you go to MSN and search "stereoviews" my sites come in #1, #2, and #10. #1 uses meta description below the listing. #2 uses top text and #10 uses meta keywords. It seems to choose different things for different listings! Weird huh?

So if you have been ignoring meta tags or not using literal descriptions and copious keywords, this could explain your changes in ranking. I have always used them and my rankings have always been great for the searches that matter to me.

David