Split Run A/B Test Without Angering Google?

Split test gone wrong...looking for advice

[SniperRyan]

We have been working on a new general landing page for our site for a few months now, and we finally were ready to begin testing it 1.5 weeks ago. I set up a split test through a well-known service to serve our control page (the current portal page) and the test page (our new design) in an A/B fashion.

This implementation requires that some Jscript on the homepage redirect a visitor through the testing-service's servers and then to PageA or PageB. We placed this code at the top of our current default homepage, with nofollow/noindex instructions to keep the spiders from following the redirects.

This past week I saw massive fluctuations in the SERPs in our sector, and once we all started to recover I noticed that our homepage is no longer indexed (site:mysite.com returns 45,000 results, no default.aspx). It appears that this is the only page which has been dropped, as there appear to be the correct number of pages in the index, and we are still appearing in some SERPs, but for subpages on our site.

Does Google not like this implementation? This is a test of visitor behavior on our site to see if one style would be better for humans than the other. It has nothing to do with gaming the engine or SEO. Would a reinclusion request be appropriate for this situation? Are there any suggestions on a change of implementation that could help?

[Philosopher]

Yep...that javascript redirect is likely causing you problems.

While what you are doing is perfectly valid and legit, javascript redirects have been used for a very long time as a "poor mans" version of cloaking.

Google has gotten better at parsing javascript for this very reason (looking for redirects).

The sudden drop may have nothing to do with it, but I think it's more likely the redirect got picked up by one of the spiders and action taken.

[jimbeetle]

We placed this code at the top of our current default homepage, with nofollow/noindex instructions to keep the spiders from following the redirects.

If I read this correctly (and I'm not sure I am) then it appears that G-bot is doing exactly what you're telling it to do and it has dropped your current index page.

Can you refiddle things to just include the noindex on the test page and still allow the bot to continue to index the current home page?

[SniperRyan]

Here is the code we are using:

<meta name="Robots" content="noindex,nofollow">
<script language="JavaScript" src="http://rotating.service.com/rotate.js" type="text/JavaScript"></script>
<script language="JavaScript">
<!--
// Rotation Script
// ------------------------------------------------
// Test: Split Path Test
// Test ID: 
// DO NOT MODIFY THIS CODE UNLESS INSTRUCTED BY SUPPORT

var fs = "'http://www.oursite.com/default_a.aspx'";
var qry = Request.QueryString;
[code]
if (document.images)
 

location.replace('http://rotating.service.com/?group=2886&failsafe=' + fs + '&' + qry);
 else
 location.href = 'http://rotating.service.com/?group=2886&failsafe=' + fs + '&' + qry;
 //-->
</script>

<noscript>
<meta http-equiv="refresh" content="0;URL=http://www.oursite.com/default_a.aspx">
</noscript>

(URLs and titles have been changed to generic words)

Below this code is our normal homepage. I had attempted to get the robots to ignore this code and just move down to the regular page, but it appears this might not have worked as intended.

The nature of our site is such that we don't have lots of individual pages to promote through SEO, but rather we promote the homepage. Losing that page from the index has/will cost us dearly in terms of traffic.

We run on ASP + SQL, is there another way to do this that won't cause Google to drop our homepage? Should I begin the reinclusion process now?

[piatkow]

Unless I have been totally confused by everything that I have read about the ROBOTS tag you have explicitly told the bot NOT to index your home page by using NOINDEX

[WW_Watcher]

Noindex & Nofollow,
If the page is no longer in the index, then google and other search engines have done exactly what you told em to do. This will not do much good for the pages linked from your home page also.

It does not care about the javascript, it never got past the noindex.

Back to watching
WW_Watcher

[jimbeetle]

Sniper, if the code you posted above is for your current home page, the control page in the test, then I'd take the robots meta off that page and simply include one on the test default page.

That will allow bots to keep indexing your current home page and will still block them from the test page.

[SniperRyan]

Thanks much for the help everyone. I'll post back with the results after I remove the meta.

[jimbeetle]

But be sure to include one on the test page.

[SniperRyan]

Update: We removed the code three days ago. Usually I can see changes to titles/descriptions etc. within three days, but so far our homepage is not yet back in the index.

I have submitted the page to be crawled. We still have all of our backlinks to the site. Is there anything else I can do to speed up the recovery process?

[WW_Watcher]

Hey SniperRyan,
IMHO, there is nothing else you can do. G-Bot will spider it in time, then it will be added to the index, in time, how long that takes is totally up to G.

You can Fret, scream and hollar, send e-mails, & spend all day checking datacenters, none of which will help. I would not have even bothered to re-submit the page to G, If you have corrected your mistake, it is just a waiting game now.

Back to Watching,
WW_Watcher

[SniperRyan]

Thanks for your input WW_Watcher. That was my assessment as well.

If this were to happen on one of my personal or client sites I would have just brushed it off, but since this is a corporate site owned by someone sending me paychecks I wanted to make sure some "due dilligence" had been applied.

[SniperRyan]

Near full recovery as of this evening. Recovered 45 #1 listings (of 47 originally). I expect more good news by Monday.

Thanks again WebmasterWorld members.

[airpal]

Ryan, try a Content Split A/B/ test, and see if that works well. It's another option in the same software package.