Track Split Tests

On Adwords & On Server

[tobrien]

Newbie post - lurker for a while and would appreciate any input you guys/gals have to offer. Apologies in advance for not responding quickly to any replies but I'm off on holiday in an hour ;)

Problem:
Adwords allows you to split test ads - the only way you know which ad results in conversion is if you identify the creative (explicitly with tracking parameters or implicitly from specific landing page). Keywords would have to be identified using referral URLs.

Now let's say you want to split test landing pages in addition to the adwords split test- concurrently, how can you do so with existing tools/analytics out there?

Optimally solutions for hosted and local analytic/test systems are viable.

If anyone is successfully doing this I'd appreciate your advice.

All the best,
Tom

[Receptional]

Good question.

Keywords would have to be identified using referral URLs

Actually - that is not how I would do it. I would track entry page URLS and then, on the adwords, change the URLS as follows:

example.com?source=adwords&kw=keyword-ad1
example.com?source=adwords&kw=keyword-ad2

The entry page is then the same to the user, but different for tracing. The entry page URL is much more effective for tracking than the referring URL.

We use javascript tracking to then store the kw variable and retrieve it at the confirmation page, using a pretty well know tracking system, but from there you have a number of ways to track through to conversion.

I am surprised that Adwords' own ROI tool doesn't compare conversion rates of two ads - but going back to Google university, I agree - it doesn't look possible, but presumably you could get away with separate adgroups and comparer that way as well.

[Receptional]

Ah - it dawns on me that you are not using exact match, so that is why you need to use referring URL to catch the keyword.

In this case, I would suspect that you will not get enough clicks to get statistically relevent data even if your tracking caught every keyphrase accurately in referral URLs. If this is an important test for you, then it is probably important on two very specific phrases - to I would try running this test on two exact match ads for comparison purposes, rather than trying to learn or deduce too much from a wider sample.

Dixon.

[tobrien]

Hi Dixon,

Thanks for the response. Yes to your point re using broad and phrase matching - it's a pain that you cannot identify the exact ad creative and keywords sent from Adwords and from what I can see there's no easy solution.

Take your point regarding the broad match and possibly not having enough statistically valid information - that would of course be determined by the volume over a given period of time.

I'll do a bit of digging into the best 3rd party analytic tools out there.

All the best,
Tom