Split Testing, also known as A/B testing, once explained is very simple. One of the great things about split testing PPC ads is that it is easy to do – as long as you understand the protocol in conducting these tests. Once you have the foundation, you will be able to start impressing your clients with some killer split tests. Of course, once you understand the fundamentals, what will really distinguish you between being a PPC Analyst and a PPC Ninja, will be how thoughtfully you can interpret data and how well you can implement.

What The Heck Is It

The best way to explain what A/B or Split testing is, would be through this example; imagine you have two possibilities for your Body Text and another two for your Headline, you will want to run four different tests to see which combination is the most effective (AB AA BB BA). The main way most professionals test their PPC ads is through A/B testing. This type of testing is an efficient method for deciding on the best promotional and marketing strategies for your business online. This method can be used to test nearly everything from, ad copy to email ads. Although it may take some time, the results will go far in yielding more conversions. The main objective of using A/B testing is to hone in on the most effective aspects of your ads, and building off of them. Your efforts will result in a much more effective and fruitful ad campaign as a whole.

One of the first things you want to do before you begin your Split Testing is figuring out what exactly you would like to test. It all depends on what you want to improve and test exactly, be it a call to action button or ad copy, you will need to figure this out before you proceed. After you have that sorted, you will want to decide which variables you will want to test. Just be sure you have a clear image of what kind of results you are hoping to get. Some testable aspects that could have a potentially big impact could be, for example, your headline, product descriptions or your call to action. It is really all up to the user’s discretion.

When it comes time to determine if the experiment was a success, you can also decide to run all of the changes fully (for example, if I’m experimenting on bids, any experimental bid changes will be applied full-time) or to remove the changes (we’ll keep the failures between us!)

So now that you know how to test PPC ads, the question is what do you test?

One of the first things you want to do before you begin your Split Testing is figuring out what exactly you would like to test. It all depends on what you want to improve and test exactly, be it a call to action button or ad copy, you will need to figure this out before you proceed. After you have that sorted, you will want to decide which variables you will want to test. Just be sure you have a clear image of what kind of results you are hoping to get. Some testable aspects that could have a potentially big impact could be, for example, your headline, product descriptions or your call to action. It is really all up to the user’s discretion.

When I’m crafting a new ad to test against my current ad, I create a list of variables that are lacking in my current ad. Here is an example of a table that I create when I’m sketching out the variables I want to use.

PPC Ad Test Variable Examples
Variable Control Ad Experimental  Ad Value Change Aesthetic Change
Call To Action Request a free account audit Instant Free Account Audit Time-Specific Capitalization
Headline Top WordPress Influencers 99 WordPress Influencers. Number Period Included
Call To Action Get 5% Off All Jewelry Offer Ends Nov. 1st. Call Now! Deadline Specific Date
Display URL example.com/Flowers example.com/Tulips Product Specific Keyword
Structured Snippet Free Wi-Fi 100 Megabit Free WiFi Specificity Number
Call-out Full-Time Support 24/7 Support Number Number
Headline Cheap Flowers Flowers Start $29.99 Specificity Number
Description Movers In All 5 Boroughs. NY Metro Area Movers. Location Specificity

Now that we know the variables, what are the metrics that we want to test?

Goals, goals, goals – what are you looking to improve upon based on your ads’ current performance. Let’s look at a sample ad to see the metrics that we want to measure first:

Your most important starting metrics are going to be clicks, impressions, click-through-rate (CTR), average cost per click (CPC), average position (where your ad appears on search results, a lower number is better), conversions, cost per conversion, conversion rate, and bounce rate (this is especially an important metric, as it tells you if people are having an awful landing page experience).

Once you can get your metrics to align with your goals, you’ll also want to see how the ads are changing your keyword’s quality score. Especially if you are building a lot of single keyword ad groups, the subtle differences in your ad can produce big results for quality score.

Since quality score measures CTR, relevancy, and landing page experience, these are the initial to-dos you want to have lined up. CTR is especially going to be your go-to number in the beginning.

How to determine the winning ad

When it comes time to determine if the experiment was a success, you can also decide to run all of the changes fully (for example, if I’m experimenting on bids, any experimental bid changes will be applied full-time) or to remove the changes (we’ll keep the failures between us!). With any of your measurements, you’ll want to see that the results are statistically significant but also for a consistent period of time. Just because one ad has a higher CTR one day than the other doesn’t guarantee anything, but if I see consistent results for a few weeks, then you’ll definitely be able to convince me which ad is performing better. Before even checking on statistical significance, you’ll want to see that the gains are positive across all metrics – if the ad has a higher CTR but a much lower conversion rate, then that doesn’t make for a winning ad.

One of my favorite blogs for the data analysis side of PPC is Okam’s Razor by Avinash Kaushik which has provided me with a simple test that I use for determining the statistical significance of any test results. It’s a one-tailed test, so the biggest question that the test answers is if the skewed impressions / clicks of either ad influenced the results too much. This statistical significance template is meant for you to play around with so be sure to have some fun with it!

Now go out and test your ads – there’s never a bad time to do it…

…as long as you have a good reason to. Know your variables you want to measure and know your measurements. And as you learned in middle school, follow the scientific method when it comes to experimenting with your ads!

Leave a Reply

Your email address will not be published.

Pin It on Pinterest

Shares
Share This