How To Split Test SERP Results with CTR Results
What if you could compare your page CTR for different positions in SERP results. What’s the effect of being in 6th position as opposed to 3rd with the same page title and description? How can you quickly test which of 5 page title variants may work the best when shown in SERPS? How many times are you the *second* visit choice in SERPs (usually after Wikipedia..!)? Despite the various tools out there for collecting data and interpreting results – it’s still really hard to split test these changes and get results fast.
There is a potential answer out there, a new tool called SERP Turkey (also free!).
I originally planned to do a long post with a how-to guide about SERP Turkey, but its creator Tom Anthony has already written a great SERP Turkey guide on SEOMoz – and it’s a great read and explains all about SERP Turkey. He’s also got a post on his blog about how to integrate SERP Turkey with Amazons Mechanical Turk for crowd-sources results – genius. If you want a really quick overview, keep reading! SERP Turkey allows you to create your project one or more copies of search result listings – each of which you can change slightly to create a variant. Your project has a unique URL, which when visited, displays any one of your variant search result pages. Visitors then click on the link that appeals most to them, and you can download a CSV of which links they clicked. For example, you might want to see if you’d get more click through simply be being in position 4 instead of 6, or if changing tweaking your page title would result in higher CTR. You can change the following elements of any of the search result entries:
You can distribute your project/test URL between friends, colleagues or use a crowd-sourcing platform such as Mechanical Turk to get results.
A Quick Test Drive
I’ve create a test case for the search term ‘dresses’. Shown below is the original SERP shown in SERP Turkey – I’m going to tweak the ASOS entry slightly to see if it performs better. I’ve create a test case for the search term ‘dresses’. Shown below is the original SERP shown in SERP Turkey – I’m going to tweak the ASOS entry slightly to see if it performs better.
With some minor tweaks: So, if you visit my test case URL, you’ll be presented with one of the two variants: When you click a result, those stats for each variant are available to me in a great CSV format from my SERP Turkey dashboard – allowing me to easily see how each variant performs. Here’s an example of the results screen, with some sample data in it (not a real test, just a demo).
It highlights whats been clicked very nicely, and you can download a CSV of the data for this variant. As Anthony mentions at the end of his SEOMoz post, you want to pass your test URL out to an appropriate target group of potential customers to get good results, not just your colleagues or friends! Also if you use a crowd-sourcing platform such as Mechanical Turk – make sure that your crowd is as closely aligned to your target group as possible (if you can).