Best way to test a new ad? - Ad rotation Optimize

I let Google run the test - when is it over?

[pavlovapete]

Hi,

I wanted to challenge an ad that we've had running for a couple of months - 20k impressions, 7% CTR.

I chose the "Ad rotation Optimize: Show better performing ads more often".

One week later Google is showing me that the new ad has 300 impressions and a 4% CTR.

Interestingly Google has given less than 2% of the traffic for the keyword to the challenger (98.40% vs 1.60%).

My questions are:
- when will Google decide the experiment is 'over' or will a 2% of the traffic 'always' be sent to the ad? Is 2% an industry standard (I think it was 4% earlier in the week - I didn't keep a record)

- have I done this experiment the right way? Is simply adding another ad the way to go here? I guess I can only assume that the quality of the traffic to the challenger is the same as the quality of the traffic to the control ad.

Thanks in advance for any comments or thoughts

[briggidere]

Hi Pavlovapete,

I never use the optimize feature when testing ads. They just never get an even enough share of impressions. I always have at least 2 ads running simultaneously and up to 8 at the most.

Google bases the decisions on which ad is better by using the CTR when the conversions are the much more important figures to me.

One thing I sometimes do is re-create the original ad so it has no performance history attached to it. I believe this provides a more level playing field for the new competing ad. I only do this when I am doing a quick test on high volume terms and only slightly different ad copy.

[pavlovapete]

Thanks for your detailed response briggidere

I always have at least 2 ads running simultaneously

I don't understand how you do this. How are you influencing which ad is shown within the same adgroup?

the conversions are the much more important figures to me

I'm not set up properly yet to track conversions from keywords/ads but I do agree with you.

Thanks for your time.

Cheers

[netmeg]

There's a setting in the campaign settings to rotate ads more evenly. Note that they aren't necessarily completely even - I asked once and Google told me if one ad was very obviously outperforming the others, even if it was set to rotate evenly, they'd still show the better performing one more often.

[pavlovapete]

Hey netmeg,

thanks for your response. I considered the (un)even rotation but thought I might be putting too much risk into the mix.

Have you any ideas how briggidere might be controlling the display of different ads in the one adgroup? Just that idea has me scratching my head.

Options:
rotate evenly
pick the best one for me Google
briggidere's mysterious "at least 2 ads running simultaneously and up to 8 at the most" method

[briggidere]

Hi Pavlovapete,

To change the ad rotation you need to go to your campaign settings. Under advanced settings you'll see one for ad rotation. Click edit and you'll see 2 options, optimise and rotate.

These are campaign settings, not adgroup settings. If you wanted to test just one adgroup, it has to be in a separate campaign.

Without proper tracking in place you are running blind tests so you won't really know which ad is doing better for you. As mentioned, Google will use the CTR to rate your ads, but these may not be the best ads for bringing in the quality visitors.

In my experience the following happens for the generic search terms:
Very specific ad (Lower CTR but higher conversion rate)
Generic ad (Higher CTR but lower conversion rate)

Personally, the first thing I'd do would be to get the tracking in place. What sort of site is it? Do you have contact forms, e-commerce etc?

[pavlovapete]

Thanks for clarifying briggidere

I forgot the settings were campaign specific. I guess this means that the setting is applied to all adgroups in the campaign? For example, say I have three adgroups, one with 2 ads, one with 3 ads and one with 5 ads and I select optimize - the optimize setting will be applied to each adgroup.

OK - I was definitely misunderstanding you - there is no way to have an adgroup with 3 ads (1 control and 2 challengers) and split traffic (for example) 80%, 10%, 10%?

I think I am getting there with your helps.

Cheers

[briggidere]

There are ways to do what you want but it takes a lot more mucking around. If you really want to test it here's what you could do, but I wouldn't recommend it unless it's very important to you.

Create 3 campaigns with the same adgroup in each.
Each campaign adgroup has one ad which you want to test.
Give one campaign 80% of the allocated test budget and the others 10% each.
Use ad scheduling to try to let the ads have a similar time slot, but not compete against each other. (important as certain times of the day perform better than others)
The results should be close to your 80/10/10 split.

This will only work if you have tracking in place though. You need to know what happens to the user after coming to your site.

It's not something I'd do as there are too many campaigns, adgroups and ads to test.

[pavlovapete]

OK briggidere, thanks so much for spending the time to explain how things work.

Use budget to split test - genius idea.

I agree though that that sounds like a lot of work - and I especially agree that doing it without tracking in place is approaching pointless in value.

OK - thanks - I am off to implement robust Adwords tracking on our website.

Thanks again

Cheers