Tracking SEO tests

[ErinTM]

Trying to get some best practices on testing SEO changes.

We are going to make a bunch of changes on subsets of pages. Say testing about 5 different on-page changes.

Originally we were going to submit separate Sitemaps to GWT and see if our test sets get indexed, how quickly, etc. But we noticed that GWT says some pages in our Sitemaps aren't indexed even though we know they are (what gives?).

So we thought, for each test, let's put a unique code on the page so we can see how many get indexed by Google.

But that doesn't solve the issue: how many people clicked on our test pages. So we are thinking of putting a tracking pixel on the test pages, specific for each test. But then I am thinking, why not just create a separate Google Analytics profile and place that code on the test pages (set up goals to track visits per test since we aren't going to change the actual URLs).

and on and on

This is where you come in. What kind of tracking do you implement when you set up tests?

Advice appreciated!

E

[JakubMovic]

Google's Website Optimizer is ok and it's free. I also highly recommend https://www.optimizely.com/ it's paid tool but it's really great. For SEOmoz pro members there is 20% off.

[Ascedia]

setting up a 1x1 pixel is simple. Just create it, drop it on the page. In Google Analytics in the content section you can search for that pixel to get your numbers. I really think , without knowing your specifics, that you are over complicating this.

You can run multiple GA scripts, but they of course don't recommend it. Remember, if you are playing around with the main stats account, if you screw anything up, you cannot repair it.

I would at least set up a new profile before doing anything so that your main stats are untouched.

[Ascedia]

That's not really true, especially in the case of Google tools. There are many free tools out there that offer features and functionality far beyond the skill level of most mid-level IM's. Optimizly works well, but it costs. The experiements (new GWO) is more than enough for what most people need.

I have used Optimizly and it was decent, definitely makes testing easy, but I go back to GWO every time. Having the tests integrated into the stats is a godsend. Always be testing.

Your link is broken. I assume you meant  http://monetate.com/products/testlab/#axzz1zl1r4yWU

[Ascedia]

You have a better chance of being indexed if you create decent content and have a couple links pointing at the page. Don't speculate about a way to improve your love from Google. Smarter people than us have spent a lot of time removing the predictability of Google.

Create good content, run a clean site, make people happy = get paid.

[Ascedia]

I have personally never seen GWO ever effect rankings or behave in any other way than it should. You can set up multivariate and a/b tests directly in Google Analytics now, much easier than GWO (which was easy) to set up, less code, and it uses your defined goals from Google Analytics.

If you want to see the testing in GA, its in experiments in the content section. Works like a charm. Are you testing  if the changes you are making have any effect on the amount of time it takes to be indexed? Or is it about ranking.

Keep in mind, if you are testing very small elements on multiple pages, you are going to have to deal with duplicate content warnings, also if these pages are the same pages, you are going to be competing against yourself in the search engines. Typically, when testing pages, the test pages are blocked from search engines for this very reason.

Don't get too hung up in trying to figure out what you can do to a page to get indexed, getting indexed is the easiest part of the show.

[DaveGerecht]

The funny thing is that I got 6 thumbs down for that answer - but I think it's still true!

[TheGrid]

David,

I have heard this theory before. If you have adsense on a page, google must visit it in order to determine the content to set the ad properly. But as you said ther eis no evidence for this directly.

It sounds to me like you can just have GA on ALL pages and simply drill down to your subset for testing on what ever group you are interested in in the content view and compare accordingly. No need to get too tricky.

[BKSearch]

It seems to me that your test is a little different than how most people would consider SEO.

Most considering higher rankings and more traffic as a success. You are considering a higher percentage of pages as indexed as a success.

I think you should step back and rethink this test. With millions of pages the indexing of your site might get better and your rankings get worse with the same changes. That would probably turn out to be a net loss in benefits and actionable traffic.

I would use SEO and page content for higher rankings.

I would think of linking patterns (internally and externally), site structure, website coding and less redundant content as the way to increase percentage of pages indexed.

[Gab-Goldenberg]

Interesting question:

I would set up with goals as you mentioned.

GWO can screw up your rankings and is not advised - speaking from personal experience.

Separately, with tens of millions of pages, maybe the basic problem is

1. getting more links

2. trimming the fat and seeing if you can consolidate certain pages together.

[glennfriesen]

Though I've never personally tested this, technically, multiple tracking codes for different GA accounts should be able to run simultaneously.

http://www.webmasterworld.com/analytics/3781851.htm

http://www.google.com/support/forum/p/Google%20Analytics/thread?tid=608cc854afe9629d&hl=en

I'm not sure if this is the most efficient or effective methodology for a multivariate test, or, rather, for measuring a multivariate test. I generally recommend Google Website Optimizer for most multivariate testing, and Omniture for folks with large, profitable sites.

[nwinternetmarketing]

Make life easier on yourself, spend some coin in a solid multivariate testing framework.

Free tools rarely offer the firepower of a commercial product.

This one seems decent: http://monetate.com/features/power-features-list/

[DaveGerecht]

Well - 6 thumbs down? I guess I was wrong!

---

I have no proof about what I am about to say, but using logic and a hunch, I would suggest  that if one were to have direct links to Google, lets say via AdWords and/or AdSense, then your site has a better chance of getting more of your pages indexed.

In theory.

With no proof at all whatsoever behind this idea.

[TeamArmando]

Have you looked into Google's Website Optimizer? It will serve two different versions of a page, roughly half and half, so you can compare which one converts better, which one ranks better, all that good stuff.

[ErinTM]

Hmm, ok.

I understand that linking to deep content will help that content, but we are trying to do discrete tests. My company was founded by a computer scientist, so everything must be scientific, even if my inclination is to simply do all the best practices and expect (hope) for the best. The test sets are too large and not in different directories, so going through logs would be . . . horrifying.

Here's the issue I am curious what peeps think:

What if I set up an additional GA profile and have two sets of GA code on each test page? Then I am getting credit for views in my main GA profile and can easily see if test sets are being viewed in the second profile.

Or, can someone give me info about how to set up a 1x1 pixel and see those stats in GA?

[StalkerB]

CRO = Conversion Rate Optimization.

I only asked as you were looking for link click metrics, which suggested you were looking to get users moving round your site more.

Users visiting specific pages doesn't mean that Google is going to index more of your pages. It's kind of 2 separate tests you want to run.

Your absolute best bet to see what pages Google is visiting is to check your server logs and see what Google is looking at. I appreciate it's confusing looking through them though and it still doesn't show what's being indexed.

I think I understand what you mean by the sitemaps now, treating a subfolder as a different site in WMT and seeing how many pages they index from the sitemap?

Could work...

Eh, normally when I want to see what pages are functionally indexed I would check in analytics to see what pages have actually brought a visitor through search. Anything that's not is effectively not indexed, though I appreciate that's not strictly true.

If a particular section has an increase in the number of landing pages bringing traffic then more pages are being indexed.

I appreciate it's onpage changes your testing but building links into pages further down the site structure should also prod search engines to crawl more pages.

It's not an easy task all in really. I guess sitemaps in WMT, functional indexing and long nights sifting through server logs is my best offer. Sorry it's not quite a solution but something to think about.

[ErinTM]

Basically we are looking to get more of our pages indexed. So rather than do everything to every page (we have tens of millions of pages) we are looking to test ideas on subsets. Whichever test gets indexed faster and more click throughs we will promote to all of the pages.

What is CRO?

[StalkerB]

Are you testing CRO or SEO?

The only reason I ask is that you can only present the search engines with one version at a time for them to index and rank, so the way to tell would be if the pages you've made changes on are getting more or less traffic which is easy enough through analytics.

CRO is an easy thing to test as well but if you could just clarify I'll have a think

[ErinTM]

We are making changes to a subset of our pages that are crawlable by google, but we aren't linking to them in a special way, so I am unsure how the URL builder will help us. But I may be missing something!

[CahabaWeb]

You could just use the Google Analytics URL Builder that will create some virtual tracking opportunities for you. Are you creating the test, running for a period and altering the test?