AdWords Experiments: How to Avoid the Under-Reporting Risk
Published: September 21, 2017
Author: Brendan Davitt
Most AdWords professionals know the value of campaign drafts and experiments: low-risk testing of things like bid changes, bid strategy, ad copy, creative, targeting, or other settings. You can assign a percentage of spend to the experiments, and once they’re set up and live, you’ll be able to see metrics within AdWords as if it were a normal campaign.
If you haven’t used them before, the process is simple:
- Propose and test changes to your search or display campaigns with less risk
- Make a draft of the campaign you’d like to test
- Change elements and direct a percentage of traffic from your original campaign
- Set live as an experiment
- Track results and apply or end
However, I recently discovered that even though the UI shows the metrics in the UI, the data does not show up in the spreadsheet when it is downloaded.
This could potentially lead to under-reported spend, accidental spend above budget, and an unhappy client. This can best be seen from the example below:
There are two different workarounds to include experiment data in reporting.
The first way to incorporate the data is to pull each experiment individually. Although it is tedious to go experiment by experiment, it gets the job done. This data is pulled at the ad group level and allows for custom columns to be pulled in. From there, you must do some manipulating. Follow this process:
- Pivot your data
- Add Day & Device (or an additional segmentation) to the pivot rows
- Include any other desired metrics
- Formatting process:
- Design
- Report Layout
- Show in Tabular Form
- Repeat All Item Labels
- Subtotals
- Do not Show Subtotals
- Paste values only into a new sheet
- Input your campaign name
The second method is a process is hidden within the UI. In the upper tabs of the UI, next to Opportunities, there is a Reports tab. From there, follow this process:
- + Report
- Table
- Pull in columns needed
- Download
- Filter for the experiment campaigns needed
Unfortunately, this process does not allow for customs columns to be pulled in. If you are using custom columns, I would resort to using option number one. Also, depending on the report, you should update your campaign lists. This way you can add data into an existing campaign or just simply add a new campaign.
Everyone will eventually have a situation where using experiments will become necessary. They are easy to set up and do not pose much of a risk to your account performance. If you do chose to set one up, remember to pull your experiment data as well!
