Slashdot Mirror
MSN Search Engine Favors IIS
Scud writes "It appears that if you want to rise up in the rankings over at the MSN search engine you would do well to host your page on IIS. Ivor Hewitt has done a study and it appears that by using IIS, you are likely to increase your odds of a higher listing by several percent."
Mirror of Ivor Hewitt's site
To be conclusive, it needs to be a controlled experiment with the same text and same outgoing/incoming links.
Just the webserver alone changing. This can happen by taking a popular site and then changing what it reports to the MSN search robots.
But until such an experiment is done, the data is open to too many interpretations.
MS software installs MSN as the default search engine in many places, and there are an awful lot of users who don't bother to change anything, so when they hit 'Search', it's going to be MSN.
- Google 7873
- Yahoo 3163
- MSN 199
- AOL 65
- Dogpile 44
- Unknown 41
- Earth Link 28
- AltaVista 16
- Excite 14
- A9.com 9
- Others 77
...which comes out to about 2% MSN.
"msnbot/1.0 (+http://search.msn.com/msnbot.htm)"
"Time is an abstract concept devised by carbon-based lifeforms to monitor their ongoing decay." - Thundercleese
You can have a "conspiracy of one" if that person acts in multiple roles.
As an example, let's say that one person is a company's bookkeeper and CFO. (This isn't uncommon in small companies.)
As a bookkeeper she cooks the books to cover her embezzlement.
As CFO she prepares false financial documents for her company and its investors.
One person, criminal acts in two roles, so in many states she can be charged with conspiracy in addition to embezzlement.
BTW, this isn't a "conspiracy" in the legal sense since it's not a crime to give preferential service on the basis of web server. It's sleazy unless it's fully disclosed, but it's not a crime unless they actually sell the search engine as an unbiased tool.
For every complex problem there is an answer that is clear, simple, and wrong. -- H L Mencken
Believe it...
First off, I looked at the difference in means for Apache rankings in MSN and Google. 61.5% (MSN) vs. 64.3% (Google) for 970 observations Right there, you ought to be able to eyeball it and see significance. But, to make sure, here are the results of a t-test which checks the likelihood that two matched sets have different means (forgive the crappy formatting):
M G
Mean 0.615061856 0.642948454
Variance 0.01100624 0.008740111
Observations 970 970
Hypothesized Mean Difference 0
df 969
t Stat -10.51551356
P(one-tail) 7.26569E-25
t Critical one-tail 1.646427658
P(two-tail) 1.45314E-24
t Critical two-tail 1.962415113
As you can see, the P is 1.45 x 10^-24, which at least makes us think the results are not pure coincidence. I don't intend on speculating on the causality, though...
Add something like this pseudocode to your server:
if $Browser = "MSNSearchBot" then $Server = "Microsoft-IIS/6.0"
One line blog. I hear that they're called Twitters now.
Try typing "online music".
On Google the top two references are iTunes and iTMS. On MSN you'll have to go through a few pages before you'll see anything about iTunes.
Yeah, I trust Microsoft to provide unbiased search results. Sure I do.
m.m.
Yeah the trouble is that the person who submitted the story linked to the old "original" result set i.e. the "/orig/" in the url rather than the more complete more recent results at: http://www.ivor.it/goog
I guess the "MSN against Google" report is more attention grabbing.