Looks like an Update at MSN Search

[Marcia]

Some down a bit, some up a bit and some have appeared out of nowhere. I'm definitely seeing noticeable changes across every search I've tried so far, and still checking.

Anyone else seeing anything? How is MSN treating you lately?

[ohcho]

Obviously, MS is challeging Google. I think Google looks like having real challenger. As I read from other web-site, MS is using neural network (NN) based search engine. If MS optimizes NN properly, it will be hard to beat.

Most of my Google top page doing well on MS. Previously was very messy. Now most of them are improved or the same. Traffic is also growing. But the question is that MS can woo back those from Google?

[Darkon]

Something's going on, that's for sure.
We were in the top listings, but now our indexpage is gone and the links we can find to our site are old. BUT in the meantime their spider crawls our site for hours every day!

[AussieWebmaster]

I am not seeing too many different results....

[mtweed]

I've noticed activity, but more from Yahoo. I'd say MSN is just trying to keep up.

[AussieWebmaster]

My newest site is doing very well in MSN compared to Yahoo and Google.

[martinuboo]

I have noticed some of my rankings change pages for a particular keyword phrase.

Example: Page-1 has historically ranked for term X with mid first page listings. With this latest update, I now see page-2 rank for the same term X, but at a much lower position -10 to -30 places. The original page-1 is no longer ranking for that term (at least not in top 50). Page-1 ranks well with G & Y! though. Strange!!

Overall though, most of my pages have improved significantly with this update.

martin

[shazbot]

I was doing AWESOME about a week and a half ago. Very high rankings for just about every keyword I care about. Just check today and I got bounced (at least from the top 70) for about 80% of my words and the 20% that were still decent got moved back significantly. My % of overall MSN traffic was cut in half according to my Urchin stats.

Anybody have any explanation?

I don't even know WHY I was so hign in MSN before, and now why I was bounced. Any suggestions would be appreciated.

[Robert_Charlton]

Originally Posted by Marcia

...Anyone else seeing anything? How is MSN treating you lately?

Marcia - On searches I monitor, I'm seeing basically the same results as what I saw on July 21st, which is the last time I'd looked. I will say that I was very pleased (and in one case astonished) by what I saw then and continue to see now.

[Marcia]

shazbot

I don't even know WHY I was so high in MSN before

Aside from other sites, I've had one of mine slip from #1 and #2 for the top two phrases down to second page and less and another one of mine move up out of nowhere to the top couple of spots for several keyword phrases. And I don't know WHY either.

It's obvious that there was an algo change, because the sites stayed the same and IBLs aren't an issue at all in these instances. But I continue to be astounded at MSN's almost uncanny capability of extracting individual words from pages and putting them together with others that aren't even contiguous on-page, and returning exactly what the queries asked for.

Also, the way they're clustering - or indenting - pages isn't at all like the other engines do it in a couple of cases I've seen with my own sites. I really can't describe it enough to make it clear, but let's say PageA links to PageB and has words 1 and 2 on it. Then, PageC links to PageD and has another couple of words on it. I've seen where PageA and PageD are the indented pair, where with other engines you'd expect an A/B or C/D combination to be paired, rather than A/D.

There's no point in struggling to figure it out, by the time they upate next it could easily be totally different again. But what I'm seeing with what they do with lexical analysis - that's why I bumped that older thread on Term Vectors in Orion's forum. I've got a page that's ranking for a four word phrase and two of the words are NOT on the page (or in meta tags) and they are NOT in any anchor text linking to the page. They are on other pages on the site probably - I'd have to take a day to find them. But I'm starting to suspect that they know what words are in *sites* and can return accurate queries even if the words in the entire phrase aren't on the page itself.

It's almost like a latency factor operating. I know they've got some LSI gurus on their team - but I'm not going there.

[DaveAtIFG]

I was doing AWESOME about a week and a half ago.

Same here for a VERY competitive phrase. In my sector, it appears sites MSN liked better than mine were ADDED to their index above mine. These are sites that are prominent on Y and/or G but were missing from MSN's previous index.

[AussieWebmaster]

In my niche a lot of the dreck that was prevalent has been dropped but not replaced by mirrors of G or Y. They have chosen various sites that have high relevance but not nessarily the large IBLs that Google is influenced by.

[Marcia]

It can't possibly be another update this soon, or could it? It sure looks like some major movements to me this morning, every one of my sites has moved around - some pages quite a bit.

[Rob]

I noticed big changes in indexed pages and BL yesterday over the previous week on MSN. I thought it might just be a glitch, but if others are seeing it, perhaps it is another update already.

I did have a client who occupied a #1 spot for weeks for a very competitive phrase but he dropped a couple weeks ago to 25 or so and stayed there. Yesterday he was 19 and today he's 17 - perhaps they are moving things around again.

The changes I've been seeing seem to be affecting those sites that have had significant changes in the past 3 months. Any sites that haven't changed, or changed before that 3ish month time period don't seem affected by any fluctuations.

[floridian]

Originally Posted by ohcho

Obviously, MS is challeging Google. I think Google looks like having real challenger. As I read from other web-site, MS is using neural network (NN) based search engine. If MS optimizes NN properly, it will be hard to beat.

Most of my Google top page doing well on MS. Previously was very messy. Now most of them are improved or the same. Traffic is also growing. But the question is that MS can woo back those from Google?

I have done research using neural networks, and they can be handy tools. But most neural nets aren't that different from regression models or other algos in terms of the problems they can solve. Setting things up right is critical, as you noted.

And I like the end results - I ranked well in MSN first, am starting to climb in Yahoo, still ignored by G. I think MSN is throwing some major computing power behind what they are doing - more frequent indexing, dynamic updates.

[AussieWebmaster]

They are still really stuck on subdomains.... we have one site that has 5 or 6 of the SERPs due to this issue.... if it remains will have to do the same thing ourselves.

[AussieWebmaster]

Or try this one at number two for exchange rates:
http://www.currencyguide.com/exchange.html

It is using a frameset of Yahoo! converter

[ohcho]

Originally Posted by floridian

I have done research using neural networks, and they can be handy tools. But most neural nets aren't that different from regression models or other algos in terms of the problems they can solve. Setting things up right is critical, as you noted.

The difference is that neuralnets can learn complex things with complex situations with far much detail than you can do with regression models. For simple problems, they are more likely produce the similar results. But ranking serps, regression won't be as effective as neuralnets. What Google and Yahoo are using is "MANUALLY DEVELOPED" regression models. If regression works for serps, MS would have chosen regression models. Note that it's far cheaper to implement!

[Marcia]

What Google and Yahoo are using is "MANUALLY DEVELOPED" regression models.

ohcho, what do you mean by "manually developed?"

[ohcho]

Originally Posted by Marcia

ohcho, what do you mean by "manually developed?"

Neural networks and regression are methods of predictive modeling. They develop models using past or sample data collected, using statistical or data mining systems. Regression models are basically mathematical formulas. What people often refer is "algorithm change". IMHO, this is wrong term to use. More accurate one should be "formula change".

Google and Yahoo ranking formulas are developed by human engineers with their hunch and stats, NOT by computer software using past or sample data. This is what I mean by MANUAL.

What MSN does is very different. It uses serps that may have prepared carefully by human specialists, as input data to neuralnets. Or they can Google serp pages if they wish to make Google like results. Or they can use Yahoo to produce Yahoo like serps. Then they train (or teach) neural network until it can correctly rank result entries as they wished.

Neural network is what our nerve systems work. If you train them with good examples, it exibit good behaviour. Obviously, MSN uses carefully prepared quality serp data to train networks. I personally like current MSN serps rankings, than Google. Note that it's not because it ranks my site better. However, MSN has messier result texts!