Ad Testing Made Easy: Running An Ad Test

spies with shoe phone
"We accept your mission!"

In the third installment of our series, Ad Testing Made Easy, I’ll be talking about implementing your ad test. After reading Felicia’s post yesterday, I’m sure it’s safe to assume you accepted her mission and are ready to dig in and implement some ad testing. So, where do you start? Where do you end? This is your field guide to putting effective PPC ad testing into practice.

Step 1: Identify One Element To Test

After you’ve identified your control ad with Felicia’s help, you’ll need to identify one element (that’s right, ONLY one) to test. Why only one element? Because you’ll want to know what worked (or what didn’t). If you change more than one element from your control to your test ad, you won’t be able to pinpoint what searchers were reacting to. Here are some suggestions for elements to test:

  • Landing pages (would only recommend testing this for conversion-focused ad testing, as this obviously won’t effect CTR)
  • Call-to-action
  • Promotions
    • Pricing information vs. generic product benefits
    • Test two benefits against each other, like free shipping and bulk discounts
    • Tone of ad (emotional vs. statistical appeal)

I have a couple of comments on writing your test ads. Ideally you’ll want to keep the number of ads per ad group to two or three, and I’d strongly recommend keeping this to two. As I mentioned previously, the fewer variables, including the number of ads,you introduce into your ad testing, the better.

Step 2: Uploading New (and Old) Ads

In both Google and Bing, you’ll want to start your testing with a clean, sparkly slate. In order to do this, you’ll need to pause all existing ads in your ad groups (even your control), and upload both your control ad and your testing ad as new. In order to get a copy of your control ad to successfully upload, you might need to change one minor facet of the ad, like adding a slash at the end of your display URL, just something minor that will make it seem unique. Don’t change any text of course, otherwise that defeats the purpose of using your proven control.

Step 3: Hide in the Bushes and See What Happens

No good spy can remain on the scene for the duration of their mission! After you implement your ad test, you’ll need to simply let the ads run untouched until you have statistically valid data. What does this mean? You need to let your ads collect enough data so that you can determine your winning ad with a high level of confidence.

There is no set benchmark for statistically valid data. Maybe for your ads you determine this to be 200 clicks for each ad, or 1,000 impressions. Or maybe, it’s simply a length of time, like a full month. If you want some more in depth help on creating statistical validity in your ad testing, Jessica wrote a very thorough post on this, and I highly recommend taking the time to read it.

There are also a number of good (and free) online tools to help you determine whether or not you have collected enough data to determine your winning ad. Basically, you can use these tools to enter your clicks, impressions, etc. and your statistical success will be calculated for you. Check out this PPC Ad Testing Tool from WebShare.

Step 4: Spying Your Winner

After you’ve gathered enough data to declare one ad your winner, now what? Well, umm…more ad testing! The idea is to constantly test ads to see what works best in your account. If your control ad wins the first test, choose another element to test to see if another variation of your control outperforms the original. Ad testing is more or less a never-ending process, especially when you’re only changing one element at a time. But ultimately in continuing to do ad testing, you’ll increase not only your engagement with your account, but the account’s long-term success.

I know you’re ready to jump in and implement your ad testing, but don’t put away your spy gear yet. Check back tomorrow for Kayla’s post on the importance of ad testing, and some common challenges we’ve found when testing!