AdWords experiments are a fantastic and easy-to-use tool within the AdWords interface. They are an absolute must for any PPC marketer who loves testing and validating their ideas - so that would be you! In this post, I will walk you through the entire process of setting up, running and finishing experiments.
What are experiments & how do they work?
Experiments let you test changes to your campaigns in a controlled setting, and without having to make changes to your original campaign. With experiments you can try out new optimisation strategies, or minor and major changes to your campaigns, such as bid adjustments, ad scheduling or new bid strategies..
You can set the duration and the budget & traffic allocation for your experiment, and compare its performance against your original campaign. Once your experiment has finished, you can easily apply your tested changes to the original campaign or convert the experiment into a new campaign.
A few things to bear in mind:
How to Set Up & Run an AdWords Experiment
Setting up the Draft
Click on +New Draft and name your draft. Choose something descriptive and easily identifiable.
For example, if your original campaign is called "Kitten Lollipop", you want to test the bid strategy "Maximise conversions", and you're starting the experiment in January 2018, then name your draft "Kitten Lollipop - Maximise Conversions - Jan18".
Hit "Save", this will create an exact copy of your current campaign.
Note that this is only a draft, not an active campaign. So don't worry if you have to stop and come back later, the draft has no impact on your original campaign and is not serving any ads at this stage.
You will find all drafts for your account at the bottom of the left-hand navigation, and within any specific campaigns.
Setting up the testing parameters
Next, you want to change your testing variable within the draft.
In our example "Kitten Lollipop - Maximise Conversions - Jan18", we are testing the bid strategy "Maximise conversions".
In order to change the bid strategy for the draft / experiment, go to "Settings" > "Bidding". Click on "Change bid strategy" and select "Maximise conversions", then hit "Save".
This is the only feature we are changing.
Do not make any other changes. Bear in mind that we want to test the impact of using a different bid strategy in our example. So in order to attribute performance differences to the bid strategy, you have to make sure that the draft / experiment and the original campaign are - and remain for the duration of the experiment - identical in all other things.
Note: of course there can be instances in which you want to run an experiment with more than one change: when you're testing out an entirely revised strategy for example. Just make sure you've noted down all changes and that your experiment doesn't follow a scattergun-approach.
From Draft to Experiment
At the top of your AdWords interface, you will find your draft status (Drafted), the campaign type and the reference to your original campaign. Click on "Apply", next to this information.
Select "Run an experiment" and hit "Apply" again. Name your experiment; typically you'll give it the same name as your draft to make things consistent and easy for you identify.
For most A/B tests, you will want no defined end date and a 50/50 split.
Having no end date means your experiment will run indefinitely, and you decide at what point you have collect sufficient data.
The experiment split is the percentage of your campaign's budget which is going to be allocated to the experiment campaign, meaning your original campaign and experiment will share the budget.
In addition to the budget share, the experiment split also determines the percentage of auctions in which your experiment will participate instead of the original campaign.
You can choose a lower experiment split if you are highly uncertain about the experiment's positive outcome and can't afford to lose any performance drops. However, in order to gain the most reliable data, it makes sense to use a 50% split.
You can choose a start date for your experiment in the future or start right away. Note that it will usually take AdWords a few minutes to start the experiment. If your campaign is quite large, it may take up to a couple hours. The experiment's status will say "Creating" whilst it is being set up, and switch to "Active" once it is running.
Monitoring your experiment's performancE
Once your experiment is up and running, it will show up in your normal campaign tab and in your left-hand navigation, indicated by the experiment icon. You will also find any active and ended experiments in your left-hand side sub-navigation, under "Drafts & experiments" in the "Campaign experiments" tab.
To understand exactly how your experiment is performing against your original campaign, click into your experimental campaign through one of the paths above.
AdWords shows you comparison stats at the top, indicating how your campaigns are doing. It will start showing a little blue asterisk once enough data has been collected to be saturated and the data is statistically significant.
Select the metrics for which you want to compare performance, and hover over the results to see more details.
Keep monitoring your experiment until you can see conclusive results. Be patient - depending on your campaign's scope, an experiment can last from days to weeks and even months!
Applying your experiment's outcome
Once your experiment has run its course and you are certain about the results, it is time to end your experiment and apply your changes if the outcome was positive.
Go into your experimental campaign and click on "Apply" in the top right. AdWords will ask you whether to apply the changes you made in the experiment to the original campaign, or if you want to create an entirely new campaign from your experiment. In most cases, you will want the former, in order to update your existing campaign.
After applying your changes, AdWords will end the experiment and you're done. You will always be able to come back into your ended experiment, or you can remove it in the "Drafts & experiments" > "Campaign experiments" tab.
If your experiment showed no significant performance change
Sometimes you will run an experiment with high hopes, however even after a long run time, the results are underwhelming or insignificant.
Bear in mind that your experiment may not have an influence on all performance metrics, so perhaps only your CTR or avg. CPC changed, whereas your conversion rate stayed consistent.
It is up to you to decide which metrics are indicative of your optimisation goals.
If none of your results are statistically significant or if they are undesirable, good! Negative outcomes or no performance changes are still a result. End your campaign and go back to the drawing board. There is no shortage of experiments you can run on AdWords, simply cross this one off the list and start testing a new variable.
What types of experiments do you usually run and how do you execute them? Let me know in the comments!
Hi, I'm Eva
Thanks for stopping by, I'm so glad you're here!