Google Sign in options
Remember me
This is not recommended for shared computers

Privacy Policy
Sign In
Create Account

Javascript Disabled Detected

You currently have javascript disabled. Several functions may not work. Please re-enable javascript to access full functionality.

Subscribe to HRA Now!

Are you a Google Analytics enthusiast?

Share and download Custom Google Analytics Reports, dashboards and advanced segments--for FREE!

www.CustomReportSharing.com

From the folks who brought you High Rankings!

Redirect Scripts And A/b Split Testing

Started by rfwilson , Jun 14 2004 04:47 PM

Please log in to reply

8 replies to this topic

#1 rfwilson

HR 1

Members
2 posts

Location:Rocklin, CA 95677

Posted 14 June 2004 - 04:47 PM

I've been studying A/B testing software. One variety of this kind of software uses a a redirect script (Meta Refresh tag) to a rotator that sends visitors to separate test pages. What if you use one of these tests on your home page so that visitors would be redirected to index1.htm and index2.htm, testpages which will disappear after the test is over?

My understanding is that if Google finds a redirect script they will index the target URL, not the URL with the redirect script on it.

But what about...

www.ibm.com
www.jcpenney.com
www.sears.com

each of which redirects you to another page. Do these big companies risk getting their home page URL altered by Google. Why or why not?

Do you have any wisdom on this?

God bless you,
Ralph F. Wilson

Edited by rfwilson, 14 June 2004 - 05:20 PM.

Back to top

#2 Jill

High Rankings Advisor

Admin
32,324 posts

Posted 14 June 2004 - 05:01 PM

Welcome Ralph! :raspberry:

I asked Ralph to post this here because he emailed me the question, and I really didn't have the answer. I'm sure some of you more technical people will be able to answer it though!

Jill

Back to top

#3 Randy

Convert Me!

Moderator
17,540 posts

Posted 14 June 2004 - 06:52 PM

Welcome Ralph !

Personally, I'm not a big fan of META Refresh redirects. Leaves too much to chance. I'd much prefer using either server-side redirects, or if you need to assess some information first, a scripting solution that uses a 301 Permanent Redirect.

Can some sites get away with less than perfect solutions? Yes. Especially if they have links pointing to internal pages so that the spiders can find their way in.

Two of the three you mentioned above are using Server-side redirects. Incorrectly IMHO, though IBM's isn't a bad solution. Frankly, with what they're doing on the server-side redirects it would be as easy to do it correctly --with a 301-- as it is to do it incorrectly --with a 302. But they're so big and have enough backlinks pointing internally that they can kind of get away with it.

Let's take a look at what each is doing on their main domain name just for fun, compliments of WebBug.

First www.ibm.com:

HTTP/1.1 302 Found
Date: Mon, 14 Jun 2004 23:38:11 GMT
Server: IBM_HTTP_SERVER/1.3.26.3  Apache/1.3.26 (Unix)
Location: http://www.ibm.com/us/
Connection: close
Transfer-Encoding: chunked
Content-Type: text/html

ca 
<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 2.0//EN">
<HTML><HEAD>
<TITLE>302 Found</TITLE>
</HEAD><BODY>
<H1>Found</H1>
The document has moved <A HREF="http://www.ibm.com/us/">here</A>.<P>
</BODY></HTML>

Personally, I would use a 301 instead, but at least their server writes a good old plain A HREF link for spiders to follow.

Next up, www.jcpenney.com:

HTTP/1.1 302 Object moved
Location: http://www3.jcpenney.com/

All I can say to that one is ACK! FWIW, it looks like they're doing some load balancing because one time it will be www1, the next time www3, etc. Still, use a 301, not a 302!

Last up is www.sears.com:

HTTP/1.1 200 OK
Server: Netscape-Enterprise/4.1
Date: Mon, 14 Jun 2004 23:42:12 GMT
Content-type: text/html
Etag: "d88971ec-1-214-4098cd11"
Last-modified: Wed, 05 May 2004 11:16:33 GMT
Content-length: 532
Accept-ranges: bytes
Connection: close

<html>
 <head>
 <title>sears.com</title>
<META HTTP-EQUIV="Pragma" Content="no-cache">
<META HTTP-EQUIV="Expires" Content="-1">
<META HTTP-EQUIV="Cache-Control" Content="no-cache">
 <script LANGUAGE="JavaScript">
 
   var queryString = window.location.search;
   
   if (queryString.indexOf("eid") > 0){
      var indexPage = "/sr/entry.jsp";
      newPage = indexPage + queryString;
   }
   else{
      var indexPage = "/sr/index.jsp";
      newPage = indexPage;
   }
   
   window.location = newPage;

 </script>
 </head>
</html>

Now that's just silly! The main page returns a 200 OK code, then is relying on Javascript to do the forwarding to a new URL (/sr/index.jsp) which then sets a cookie. Try going to their main page without JS enabled and you'll get exactly nowhere.

Of the three choices, IBM's is the best, but it could be improved. If you want to see proof of how something like this can play havoc with a site, check out these "site:" commands on each domain.

IBM
JC Penney
Sears

Look at all of those pages that can't be spidered on the latter two. Also look at the number of pages Google has indexed for each. You'll never convince me that JC Penney's site is only 400 pages. Or even that Sears only has 5,800 pages in their site.

IBM, while they're not doing it the way I would, are far, far better. Almost 70,000 pages indexed. I'd be willing to bet that the other two would be up in that range if they weren't making life so difficult for the spiders.

Back to top

#4 Jill

High Rankings Advisor

Admin
32,324 posts

Posted 14 June 2004 - 08:04 PM

Hmm...I just remembered once that a guy from JC Penny came to one of the SES conferences...guess he wasn't paying good enough attention! :dance:

Back to top

#5 Randy

Convert Me!

Moderator
17,540 posts

Posted 15 June 2004 - 07:23 AM

Just goes to show that even the big boys don't always get it right.

The difference is that when you have a Recognizable Brand that lots of people link to and tens-of-thousands of pages you can sometimes still get a few hundred or few thousand pages listed.

Wouldn't it be fun to remove those spider obstructions and see what happened? :huh:

Back to top

#6 rfwilson

HR 1

Members
2 posts

Location:Rocklin, CA 95677

Posted 15 June 2004 - 10:33 AM

Thanks for all the insight on redirect procedures. That's helpful.

But my application is how to do A/B split testing with the type of software that sends the visitor to webpage A or B or C in an even-handed manner. How do I keep from getting Google to now index A or B or C rather than the main URL I am testing?

Thanks for your input.

Ralph F. Wilson

Back to top

#7 Jill

High Rankings Advisor

Admin
32,324 posts

Posted 15 June 2004 - 10:36 AM

Do you have an example of it where we could see the code it produces?

Jill

Back to top

#8 Alan Perkins

Token male admin

Admin
1,559 posts

Location:UK

Posted 15 June 2004 - 10:58 AM

Hi Ralph

While the trial is running, use the meta refresh as at present. (There are other options, such as a 302 or a JavaScript redirect or a frameset).

When the trial is over, remove index1.htm and index2.htm. Set up a 301 from index1.htm and index2.htm back to your home page. Also, use robots.txt to exclude index1.htm and index2.htm from the search engines' indexes.

Later, when you see few or zero hits on index1.htm and index2.htm, you can remove the 301 and the robots.txt exclusion. Ideally, make sure you have a 404 handler in place to catch the occasional hit.

Alternative: if you're only concerned about Google, then use a 302. Google indexes the content at the target URL under the URL of the 302, which is exactly what you want.

Back to top

#9 Vertster

Google wristbar installed

Active Members
327 posts

Location:Salt Lake City, Utah

Posted 15 June 2004 - 03:29 PM

Hi Ralph,

Rather than using a Meta Refresh redirect, what about using a Javasript redirect? The
problem is that both of these strategies rely on client side code to push the browser to the rotation script. Presumably, the spider couldn't follow a javascript link (until the next googlebot rolls out.) You could provide a hard link on the page for the spiders and people with JS disabled, and a rediret to the splitter for everyone else.

Certainly, using server side code is much prefereable, although out of reach of non technical website managers. I think the issue with the meta refresh you have described might be sending the spider to another domain for the actual A/B splitting.

Of course, I recommend against ttesting on the home page, but this is more due to the variability of homepage traffic. Its true that splitting traffic "equalizes the playing field," but with the dynamic nature of website traffic, the traffic mix a site gets one month may vary a lot the next. As will the response to a given promotion.