MSN Backend Sneak Peek [Archive] - Search Engine Watch Forums

Dodger

07-01-2004, 02:58 AM

I was looking at the "look before it is gone" page and saw that they claim to have 1 billion pages in the index already on this page (http://techpreview.search.msn.com/about.aspx). They also claim that it will grow very quickly. I will buy that, the MSNbot is everywhere.

A few samplings have been very favorable (for me anyway). So far I am liking what I am seeing. Although there are a lot of results from one domain flooding the results. The As Seen on TV seems to be pretty prevalent over here too.

These pages are showing the handiwork of the MSNbot's crawls. There are tons of pages in there. More than what Slurp is picking up on three new sites that were started around March.

Nacho

07-01-2004, 03:13 AM

Finally after I was able to get in (very difficult), tried to submit my post for 3 times because IE was crashing, this is what I got.

I searched for one of our keywords, and I saw that we were ranked for it on the #2 and #3 listings.

The strange listing was #3, since it was refering to a link from our (old) banner management system that (which was replaced by a new one 3 weeks ago). Somehow it is giving it some type of high popularity score (not "link popularity") for being shown all over the site. Then adding the title and product description snippet of the landing page.

Why would it do something like this :confused:

Dodger

07-01-2004, 03:31 AM

I found a result that mentioned a product just once in a page of about 450 words total. There is a link on that page that goes to another site where that product is sold, the link does not contain the term in its anchor text.

That page has nothing to do with the term, it is just mentioned in passing as a "we also carry" type of item on the page. It is in close proximity to the outbound link though.

Probably just a bad algo, but it ranked number one for it. :D

Depending on your viewpoint I guess, some people are going to really love MSN.

polarmate

07-01-2004, 04:22 AM

I liked what I saw!! But it looks like they have pulled the tech preview for now. I see this:
We are sorry, the web site for MSN Search Technology Preview is currently unavailable. Please try again later.

dannysullivan

07-01-2004, 05:15 AM

Keep an eye on the MSN Sandbox (http://sandbox.msn.com/). They have links to the US, UK and some other versions of their new search engine. The UK link is working for me, at the moment.

For those coming into this thread, be aware that the new search engine Microsoft released is not being used to power the results at the non-beta MSN Search that ordinary people use. That site has had some cosmetic changes, and this thread (http://forums.searchenginewatch.com/forum/showthread.php?t=437) describes them more.

I also posted a SearchDay story here: MSN Search Gets New Look; Microsoft Gets New Search Engine (http://searchenginewatch.com/searchday/article.php/3376041).

Chris_D

07-01-2004, 08:07 AM

Great catch Lex!

:)

The biggest thing I noticed is that there appears to be no result clustering - there were 11 consecutive results from the same domain on one page.....

doppelganger

07-01-2004, 10:38 AM

It also doesn't seem to be handling things like dashes very well...

Compare the results on a search for "e-commerce" on this new MSN, versus Google and Yahoo...

All I have to say is... YIKES!!!!!!!!!!!!!!!!

pleeker

07-01-2004, 01:47 PM

(as requested, reposting my comments from last night when there was just one thread about the MSN changes....)

The techpreview SERPS are awful.

Found a page ranked No. 1 on a 2-word query that has nothing but a large graphic on the home page and a text copyright notice at the bottom. Oh, it also has an overbloated META Keywords tag and that large image has an even more bloated "alt" attribute. But at least the page was created in FrontPage, so it must score points for that. :)

Authority sites aren't being recognized. (No eBay in the Top 15 for "auction" or "auctions"?)

Too many listings from a single domain.

etc.

But at least it exists.

Chndru

07-01-2004, 04:27 PM

atleast now i know, that MS sure is into search rather than just press releases. yahoo dropped by 5%.. hmm..

robwatts

07-01-2004, 05:04 PM

Nice and simple..but I dont like it.

Way too many Sites Positioned Above Mine ;)

Ive grabbed 3 serps for 3 popular travel kw's. Gonna look at them weekly and see what happens.

Early days, but interesting nonetheless :)

garyp

07-01-2004, 05:08 PM

Since this is an early demo, I wouldn't be suprised if some of the following comes and goes.

+ Unlike Google and like Yahoo, no stopwords.

+ 15 results (instead of 10 at Google, 20 at Yahoo). I wish results were numbered (as they are at Google)

+ Boolean NOT appears to work (as does the - sign)

+ Boolean OR works

+ Site: appears to work. Domain: does the same thing?

+ Backlinks (link, title, and url limits do not work)

+ Nested searching (using parens) appears to work

+ I was unable to view more than the first 150 (approx) results

St0n3y

07-01-2004, 05:12 PM

not terribly imopressed. Lots of room for improvement, which I'm sure will be forthcoming.

Dodger

07-01-2004, 05:24 PM

They have a little "rainbow" graphic next to each listing that you can give feedback on it. I would imagine that this thing may be getting hammered by the likes of some TechNet people and possibly some of us SEWatchers. There is a drop down list of the types of feedback you can send them which include not fresh (out-of-date content), spam or junk, dead link (title link does not lead to a page or site), is missing a description or the description is too short, has a description that does not match page content, and some illegal and adult related feedback.

I was talking with a buddy and he noticed that he had entered in a two-word term and checkedt he results (usual comments everyone is saying here), but when he returned to check again for the same term it popped up a "no results found for (two-word term)". He checked later and they were back.

It seems that we might be working in a LIVE test environment here, and they are working on it while we are using it. Thus sporadic results as above, plus sometimes the site is inaccessable.

Dodger

07-01-2004, 05:34 PM

+ Site: appears to work

I am glad you brought that up, I forgot too mention that. It is one that I tried earlier on too.

It seems to be working, but hard to verify for they are not showing the regular line "Results 1-15 of xxxxxxx" as is being done in other SE's (even MSN Search proper). I tried it on some smaller sites, because you have to page thru the results and hand count them.

I would have to agree that it is working properly because if they have one billion pages (reportedly on their FAQ page) then what I am seeing could be an accurate accounting given the veracity of the MSNbot of late. Now compared to Yahoo's 4billion plus (or whatever it is now) and they only having at most one-third of the pages that Google is showing (if that) then I would say Yahoo is fudging the figures a little bit -- MSN is definitely showing a hell of a lot more pages than Yahoo is right now. I would place Yahoo pages (guesstimatedly) at no more than half-a-bill if they're lucky. The MSN one billion I think is a pretty good and solid figure.

pleeker

07-01-2004, 05:43 PM

There is a drop down list of the types of feedback you can send them which include not fresh (out-of-date content), spam or junk, dead link (title link does not lead to a page or site), is missing a description or the description is too short, has a description that does not match page content, and some illegal and adult related feedback.

That is a nice touch, and at this point, probably the best thing they have going for them. The feedback system beats the pants off the SERPs. :)

Terry Plank

07-01-2004, 06:36 PM

Couple of things I noticed from spending a little time in an industry I work with a lot.

1. Plural vs singular seemed to have more of a variety of results than some of the other engines.

2. Search for plural "officiants", were results that were more directories in the field than the singular "officiant". Also more top ranking sites.of the URLs that had the keyword phrase in the URL.

3. Seemed to be a pretty deep indexing of pages within the domains I evaluated.

Will be interesting to do more extensive research when it warrants not wasting time before it settles into a final form. :-)

Also found it interesting in the About section they invite SEO's to be involved:
"Webmasters, SEOs & search enthusiasts
Be part of the conversation - MSN Search newsgroup.
Get information - MSNBot and your site."

Nacho

07-01-2004, 06:49 PM

It seems to be working, but hard to verify for they are not showing the regular line "Results 1-15 of xxxxxxx" as is being done in other SE's (even MSN Search proper). I tried it on some smaller sites, because you have to page thru the results and hand count them.
I did a site:www.mysite.com and tried to hand cound my urls indexed. Guess what?

I couldn't get past page 10. Seems like they are limiting results to 100 or so serps per query.

Dodger

07-01-2004, 07:02 PM

I did a site:www.mysite.com and tried to hand cound my urls indexed. Guess what?

I couldn't get past page 10. Seems like they are limiting results to 100 or so serps per query.

oooh...really? That is interesting. Gotta be a reason for that.

It seems to be really slow too. I am not sure what they are running it on, but knowing Microsoft it has to be one of their Server Platforms. That is going to raise the question of how fast these results are going to pop up.

They are not showing the total results in the subset, nor as we can see they are not showing more that 100 results. I am wondering if they are using their full index for testing right now -- hmmm.

Nacho

07-01-2004, 07:58 PM

"It'll be later this year that we actually roll out what's entirely our own back end driving the search", said Bill Gates in this article on June 28th:

http://www.zdnet.com.au/news/software/0,2000061733,39151731,00.htm

What do you guys think? Can they reach this goal as it looks now?

Dodger

07-01-2004, 08:08 PM

What do you guys think? Can they reach this goal as it looks now?

Yes. They will roll something out by the end of the year.

But expect MSN Search SP1-RC1 to come out sometime in mid-February or March of next year. ;)

Nacho

07-01-2004, 09:58 PM

LOL :D and SP2 for Local Search after that.

When I do a search for "pizza 92037" I get #6 result with:

Indian Restaurants in USA - USA Indian Cuisine Restaurants
St, Lakewood, CA 90713 Ph 562-867-3340 Desi Pizza 5806, Bellflower Blvd, Lakewood, CA 90713 Ph 562 ... Gallery 5662, La Jolla Blvd, La Jolla, CA 92037 Ph 858-551-8610 India Palace 7514, Girard Ave ...
www. gourmetindia .us/

And #8 as . . .

Oregon Consulting Management Inc.
LLC 3352 Holiday Court, Suite 224 La Jolla, CA 92037-1808 Summary: Providing Construction Management ... In Design Papa Murphy’s Take ‘N’ Bake Pizza – Camas Relocation Client: James Lovelace 5512 NE ...
www. ocmi .net/projects.html
Keep working on it guys! You've got a lot of room for improvement. :rolleyes:

detlev

07-01-2004, 11:55 PM

Hello everyone,

I am seeing less emphasis on relevancy with keywords in TITLEs that are at least 50 chars in length. Unless the TITLE is just about a perfect match, (and short), I am not seeing the same results I typically get from other engines. I like longer TITLEs because they still work for relevancy and you can include a call to action for getting the click.

I bet MSN will keep tweaking the TITLE algo because it appears to be raw and seemingly a pure density thing at the moment. If you have a high density of the keyword in the TITLE, it looks like you have a good chance for a high position when the terms are backed up with content and links into the site. Seems ordinary, but the other engines recognize the TITLE length can provide the searcher a more verbose listing qualifying for the click.

Try a search for: flowers. Almost all TITLEs are short with a high density.

Just something I've noticed early on - and this can change in 24 hours.

*cheers*
-detlev

Dodger

07-02-2004, 12:02 AM

LOL :D and SP2 for Local Search after that.

When I do a search for "pizza 92037" I get #6 result with:

Indian Restaurants in USA - USA Indian Cuisine Restaurants
And #8 as . . .

Oregon Consulting Management Inc.
www. ocmi .net/projects.html

Those are relevant. If they had a cache version of the page it would have said something like:

pizza : does not appear on this page for they were consumed after delivery

Keep working on it guys! You've got a lot of room for improvement. :rolleyes:

Gotta love them guys. They are like an old Timex watch - "they can take a licking, but they keep on ticking"

British Chap

07-02-2004, 12:11 AM

I had 7 out of 10 listings and all have disappeared

David

Dodger

07-02-2004, 12:29 AM

I had 7 out of 10 listings and all have disappeared

Don't worry about that. Someone else noticed results that disappeared, but they came back a couple of hours later. Check back later.

I think they are tweaking on this thing while we are using it. I am sure you will notice things flip-flopping around on a regular basis.

garyp

07-02-2004, 12:33 AM

Nacho:
You are correct about serp limits. I made this point in an earlier post.

Also, the tech preview does not have a query term limit. Google truncates after the 10th term.

Btw, speaking of Bill Gates, note the fact that when you run a search for "bill gates" you're not given a chance to comment on the result (yes, you will not see a "rainbow button" for his home page. :)
http://techpreview.search.msn.com/results.aspx?q=%22bill+gates%22&FORM=SMCRT

Nacho

07-02-2004, 12:41 AM

Less emphasis on longer TITLEs...
Right!

www.and-a-huge-emphasis-on-keywords-on-url-that-perhpas-if-I-add-pizza-92037-here-MSN-will-rank-it-high.com

Dodger

07-02-2004, 12:46 AM

polarmate

07-02-2004, 02:30 AM

If Google can give themselves a PR of 10, I guess ol' Bill Gates can make sure that there is no feedback button for his sites. LOL! Maybe they're just being v realistic - think of the kind of crap that would have to wade through as feedback!!

pleeker

07-02-2004, 02:27 PM

What do you guys think? Can they reach this goal as it looks now?

Of course they can! He just said it would roll out this year; he didn't say it would be any good. :) Besides, we all know MSFT believes in the release first, fix later approach to new products.

BTW, great find Gary. Funny!

seomike

07-02-2004, 04:50 PM

I've seen 5 cloaked domains already.

Thought this was going to be spamproof. LOL

rustybrick

07-02-2004, 05:01 PM

I've seen 5 cloaked domains already.

Thought this was going to be spamproof. LOL

I guess MSN is on the side of cloakers who say cloaking is not spam.

mocara

07-02-2004, 06:55 PM

I've noticed that sites which use refferal systems such as myaffiliate.com are missrepresented.

All the links such as www.myaffiliateprogram.com/u/user/b.asp?id=112 go through to a site such as www.somesite.com they are not pages merely redirects, in fact 301 moveds, yet the engine shows them as pages. Very wierd.

Surely these should not show up in search results?

Anyone got any good links on search engines and 301s? I'm intrested in how it effects an attempt to change your sites domain name.

Mocara

Dodger

07-02-2004, 07:34 PM

Btw, speaking of Bill Gates, note the fact that when you run a search for "bill gates" you're not given a chance to comment on the result (yes, you will not see a "rainbow button" for his home page. :)
http://techpreview.search.msn.com/results.aspx?q=%22bill+gates%22&FORM=SMCRT

It appears that they are "pinning" certain results to the top of the page.

George Bush (http://techpreview.search.msn.com/results.aspx?q=George+Bush&FORM=SMCRT) : John Kerry (http://techpreview.search.msn.com/results.aspx?q=john+kerry&FORM=SMCRT)

Google (http://techpreview.search.msn.com/results.aspx?q=google&FORM=SMCRT) : Yahoo (http://techpreview.search.msn.com/results.aspx?q=yahoo&FORM=SMCRT)

Pepsi (http://techpreview.search.msn.com/results.aspx?q=pepsi&FORM=SMCRT) : Coca Cola (http://techpreview.search.msn.com/results.aspx?q=Coca+Cola&FORM=SMCRT) : Budweiser (http://techpreview.search.msn.com/results.aspx?q=Budweiser&FORM=SMCRT) : Coors (http://techpreview.search.msn.com/results.aspx?q=Coors&FORM=SMCRT)

Michael Jordan (http://techpreview.search.msn.com/results.aspx?q=michael+jordan&FORM=SMCRT) : Michael Jackson (http://techpreview.search.msn.com/results.aspx?q=Michael+Jackson&FORM=SMCRT)

Still Paul Allen (http://techpreview.search.msn.com/results.aspx?q=Paul+Allen&FORM=SMCRT) has not been pinned ... so much for freindship, eh? :rolleyes:

Phoenix

07-02-2004, 10:05 PM

Local Search

Hmmm... yucky I get the same exact results for "italian restaurant 78666" and "italian restaurant 78209", and so on. They got close with Lubbock, unfortuately that about 400 miles from the zip codes above. That would be one long drive for italian. lol