How to Set up Google Adwords Campaign Experiments
Want to improve your PPC performance without risking the performance of your top campaigns? There is no such thing as the perfect PPC marketing campaign, in fact, you should be constantly testing new ideas on all your campaigns regardless of their performance; discovering that a given change can improve your top campaigns will bring the highest impact on the results of your account. That’s where Adwords Campaign Experiments fit in.
Nobody wants to take a false step adjusting a setting or changing a landing page. Often it’s easier to tinker with underperforming campaigns that won’t seriously affect the overall results of your account and not touch what seems to be working. However, this fear of testing new ideas could be exactly what’s holding you back from taking your campaign performance to the next level. In this article you will learn you how to set up an experiment in Google Ads in the Drafts & Experiments section. After reading it, you will be ready to trial the changes that have the potential to skyrocket the performance of your campaigns.
How to create a Google experiment
Step 1: Create a Draft
A Draft is essentially a copy of the specific campaign that you want to test. When creating a draft, you will have to choose an original campaign that Google will replicate for you to make the changes you wish to test this new version, not to your precious top campaign. So, you will be making all your modifications in a completely controlled environment where there is no risk of negatively affecting your original campaign. The following are some examples of changes you might want to try on the “mirror campaign” that Adwords will create for you:
- Comparing the performance of an ad with Headline 3 and Description 2 against ads with only two headlines and one description.
- Experimenting with different copies to determine which variation works better
- Analyzing the effect of a change in the bidding strategy of a given campaign. For example, testing Enhanced CPC agaisnt Target CPA.
- Evaluating the performance of a geographic targeting by region, city, or radio and coordinates.
*Do NOT try to test landing page performance in this section. A/B Testing Adwords landing pages should be done with a 50% split between each URL, which Adwords cannot guarantee. For this purpose you can use Google Optimize or other third-party providers.
It’s a risk-free testing ground which allows you to open up, experiment, and take advantage of all the data that is available to everyone using Online Paid Media to grow their business today.
Ad Variation Section
This section is specifically created for you to create ad variations and test ad copies within your campaigns. You can add any variations in your ads’ headlines, descriptions, paths, final URLs, or final mobile URLs. The process is very similar to that of “find and replace” in Microsoft Word or Google Docs – just search for the item that you want to modify (ad text, URL, etc.) and then replace it with your new alternative. To avoid changing parts of words, Google provides the option to check the box: “Match whole words only”, which should solve this problem.
Step 2: Create an Experiment
First, go to the Experiments section and create a new one. You will have to choose a Draft, and by doing so you will also select the campaign that you will compare it to (since this draft was created over an original campaign that will serve as a base for the test). Next, give a name to your experiment and a description if you want. Then, you can decide between two split options:
- Search-based: this split enables you to assign users randomly to either your experiment or original campaign every time a search occurs. With this split, a single user might see both the experiment and your original campaign multiple times.
- Cookie-based: with this split, users will only see either the ads from your original campaign or your experiment campaign, thanks to the user’s cookies, he or she will only see a single campaign version.
Optimize your PPC spend
You must be asking yourself, which is best? Well, as in many cases, it depends. The following are some of the most common problems you might encounter when with each split option:
- If your original campaign is not able to gather clicks and conversions quickly, or if you value how fast the experiment will have statistically significant results more than the experiment’s accuracy, you may prefer the search-based split option, as you will get statistically significant results faster.
- If your original campaign has no problem with gathering clicks and conversions quickly, or if you value the accuracy of the results of the experiment over how fast can it be statistically significant, your best option will be the cookie-based split, as you will know that a user won’t see the two versions of your campaign and so will avoid your test from being impacted by other factors.
You’re almost set! Just two fast, but important steps on your way. Now, you need to decide how much of your original campaign’s budget you will give to your “mirror campaign”. We recommend to split the budget of the original campaign to 50%, as having the same budget will make the experiment more reliable. If you set a different split, say 30% and 70%, after the experiment is done you might worry about the effect that this budget allocation had in the performance of your experiment.
Finally, set the start and end date of the experiment, which will depend on how fast your campaigns are able to gather traffic. From personal experience, I would recommend running the experiment for at least one month, as it should be enough for the majority of your experiments to garner statistically significant results.