Target ROAS experiments for Shopping let you test Target ROAS against your standard Shopping Campaigns using manual CPC or eCPC bidding and measure the impact. The experiment splits your budget between the experiment and the original campaign, which is the most reliable way to remove any bias and seasonality from the experiment.
When you are finished with your experiment, you can apply the experiment directly to your original campaign. Target ROAS experiments are only available for standard Shopping campaigns using manual CPC or eCPC bidding.
Before you begin
Before you start your Target ROAS experiment for Shopping, you need at least 15 conversions in the past 30 days. This will allow you to directly compare the performance of that Target ROAS bid strategy against your original campaign. If you want to specify a ROAS target, you should keep it close to historical average and then change it to the new ROAS target after a week or two. This is to ensure that the campaign gets enough traffic to learn from in the first few weeks.
You should pick a test window between 30 and 60 days based on the length of your average conversion delay. If your average conversion delay is less than 7 days, you can test for 30 days. If your average conversion delay is more than 21 days you should test for 60 days. Target ROAS experiment reporting will show 95% confidence intervals to help you determine when the experiment results are statistically significant.
Create a Target ROAS experiment
- Sign in to Google Ads.
- Click Settings in the page menu on the left.
- Select the campaign you want to test.
- Under “Bidding and budget”, in the “One-click target ROAS” section, click Try target ROAS to optimize for conversion value. You’ll see information about your experiment based on the original campaign and parameters based on experiment best practices.
- Click Change options if you want to change any of the following settings:
Note: If you make any changes to the original campaign, they will not be reflected in the experiment. You must manually make changes to the experiment settings or create a new experiment.
- Name: Give your Target ROAS experiment a name.
- Start date: Date the experiment is created.
- End date: Recommended number of days to test.
Note: If your campaign tends to have a long conversion delay, you may want to test for longer than 30 days.
- Experiment split (50%): Budget split between experiment and original campaign. We recommend you don’t change this.
- Target ROAS (%): Recommended Target ROAS is calculated from the last 28 days average ROAS, excluding most recent days due to conversion delay. If you want to set a different Target ROAS, be sure to set an initial target that is in range with historical ROAS.
- Click Create. The new experiment will be listed in the Experiments page under the “All experiments” section. You can manage your experiment: edit, run, pause, or delete. Read Find and edit your experiment.
Make changes to your experiment
You can make changes to your original campaign and your experiment, but be aware that significant changes can make it difficult to compare. If you make any structure campaign changes on the original campaign (for example, adding a new ad group, remove or add products) these changes will not be automatically applied to the experiment. Remember to also make these same changes to the experiment arm for a clean apples to apples comparison.
Apply your ROAS experiment
After a few weeks have passed you should review your experiment. You can adjust your Target ROAS based on desired results:
|Want more conversion value?||Want higher ROAS?|
If volume of sales is your goal, and you want to drive more conversion value, lower your ROAS target.
Lowering your ROAS target will allow Target ROAS to enter more auctions and drive more conversion value. This might be at the expense of efficiency.
If efficiency and higher ROAS is your goal, raise your ROAS target.
Raising your ROAS target will limit Target ROAS to entering auctions that Google believes are more likely to lead to a conversion. This means higher ROAS, but less volume.
If you are happy with the current level of efficiency and/or volume, leave the ROAS target as is. When you are ready to apply your experiment, read Apply your experiment.