Step by Step: Why and How to Test Your Paid Search Ads

Paid search ads are still a large and important part of most companies’ and brands’ online marketing strategy. For one thing, search is one of the few places online where users can’t block ads, and Google has gone to great lengths to optimize their positioning and make them more similar to organic results.

These are just a couple of reasons why you need to make sure your ads are successful, and that you’re funneling money in the right direction and seeing your expected ROI.

Why test your paid ads?

You should test because you want to improve the performance of your paid traffic and to get customer insights. Testing helps you understand what customers like and do not like.

Do they want to save time and money, or get premium services and products? Through testing, you learn what appeals to them.

Another reason to test is to remove banner blindness. When people constantly see the same ads from a brand, they become accustomed to them and no longer really “see” them. When that happens you’re just wasting money.

When it comes to testing, many people think of A/B testing, or single group testing, which I wrote about in an earlier post. It talked about the higher click through rates and improved quality scores that can result from testing the various components of your ad. (Please note that post was written for the classic ads rather than the expanded text ads, but the principles of testing one thing at a time still apply).

Single Group Testing

However you test, you need to be scientific about it. Brad Geddes, a Guru in this domain, talks a lot about single-group and multi-group testing.

At the single group level, rather than testing what consumers like across all your products or services, you are instead testing whether you have the right message for that ad group targeting.

Single group testing is good for your branded groups and those groups with a lot of activity. Be cautious about testing ads that have seen low traffic, since there is not enough data to achieve results that are statistically significant.

Source: AdAlysis.com

Multi Ad Group Testing

The multi-ad group or pattern testing is useful if there is an idea you want to test over multiple groups or if you want to scale your testing across hundreds or thousands of ad groups.

This method will show you what is working at a high level. After you test your idea, you can aggregate the resulting data across those ad groups.

With this method, you can also get some insights into consumer behavior, and it’s a useful technique for testing ad groups that have a low volume of data.

Source: AdAlysis.com

What’s your primary question?

What are your main questions about your ad performance? Are you in a place where you need to test but aren’t sure what’s the most important feature to start testing?

Work through some of the questions below with your team to identify what is relevant for your business.

  • Is price important? Are your customers looking for deals, are they sensitive to pricing?
  • Do they want a selection? How important is variety to them?
  • Is location relevant? Do you need to test the inclusion of geographic terms in your ad copy?
  • Is shipping an issue? Do customers want options such as free or expedited shipping?
  • How do they respond to what you ask? Do they want to find, discover, or learn?

Obviously, not all of those items will be relevant to your specific brand, but they can get you started with relevant points on your testing. Moreover, they can help with marketing on other channels.

What about content?

What copy will you use to run your test? At some point, you may hit a creative block when trying to decide what it is you want to test, or might find yourself repeating the same ideas over and over again in your ad copy.

There are a few places to explore for new creatives:

  • Social media content: the 140 characters limit on Twitter is a great source for new ideas since you are also limited in your ads. You may be able to discover some commonly used acronyms that you can incorporate in your paid traffic content.
  • Customer support questions: What do customers ask you? What do they want and how do they describe their problems? Use their wording in ads. If you have an online customer support forum, you can see exactly what they asked through their inquiries or posts.
  • Reviews: Similar to questions, you can go through reviews to see their wording. How do they say you helped them in reviews – or didn’t help them! Negative reviews still provide content you can use.
  • Search query data: As you review your search query data in your paid account, you may discover new phrases to test in your ads

Decide on your metric

Once you have decided on the above, you have to decide what metrics to test based on your goal. Is it impressions, clicks, revenue? Of course, this comes back to the why of your test and you need to know in advance how you will determine a winner.

Impressions tell you the possibility of conversion at the time of the impression. By comparing conversions to impressions, you can determine the likelihood of a conversion resulting from an impression of your ad.

For eCommerce businesses looking at revenue, you can subtract the cost from the revenue and divide by impressions to get a metric. Although some of these exact metrics are not in AdWords, they are easy enough to calculate manually.

Source: Unbounce

Drafts & Experiments

Now that you are ready to test, you can do so with drafts and experiments in AdWords.

With this feature, you can easily do A/B testing directly in AdWords to test for remarketing, keywords, bid strategies or ad copy, and see how they would affect the results of your campaign before applying the changes.

Before starting your experiment, you can first create a draft by selecting this option at the top right of your screen. Once you create a draft, you can use it to run an experiment (or apply those changes to a campaign).

When you choose an experiment, the experiment is created immediately from the draft and will share the budget with existing campaigns. Give your experiment a name that indicates what you changed, such as increased bid, so you know what the campaign was about when you revisit it later.

Hold off from changing your original campaign while the experiment is running.

After the experiment is created, there will be an “all experiments” tab on the left hand side. Note that with experiments, you can only run one at a time – the tool doesn’t allow for multivariate testing.

When your test – or experiment – is done running, you will see a scorecard that allows you to compare performances between the experiment and the campaign with arrows that indicate whether the change has statistical significance.

After viewing the results, you can update your original campaign or create a new one. From there, you may choose to continue experimenting by selecting another feature to test.

AdWords experts know that not testing paid traffic is not an option. You cannot run a paid campaign and not make changes, yet expect a positive ROI. But testing paid traffic has to be done strategically to make it worthwhile.

Before creating a campaign, decide what you want to test and what success looks like. What is the metric you want to focus on? How will you modify your messaging? Once you know what you want to test, allow enough time for meaningful results in your data before applying changes to your live data.

These are the basics of paid search ads testing. I’d love to know how it turned out for you, and what improvements you managed to achieve. Let me know in the comments!

Note: The opinions expressed in this article are the views of the author, and not necessarily the views of Caphyon, its staff, or its partners.

Author: Emily Reiffer

Emily Reiffer is general manager at Digital Monopoly, parent company of Paid Traffic, an Australian based PPC advertising agency. She is a marketing fanatic and entrepreneur with a passion for everything search engine related. You can reach out to her on linkedin.

Leave a Reply

Your email address will not be published. Required fields are marked *