Directional experiments provide a quick and scalable way to A/B test multiple creative assets and receive directional performance insights. This feature helps you make data-driven decisions that impact how you can better optimize your campaigns’ performance.
This article explains how Directional experiments works for App campaigns and the benefits of using it. Learn how to Set up a Directional experiment for App campaigns.
On this page
- Benefits of Directional experiments
- How Directional experiments work
- Keep in mind
- When to use Directional experiments
- Considerations around Directional experiments
Benefits of Directional experiments
- Identify top-performing assets: You can test multiple assets at scale with up to 20 videos tested head-to-head. This helps you to get clear, directional insights into which assets perform and resonate best with your audience.
- Make data-informed decisions: You can use the test results to help guide your asset production and overall campaign strategy.
- Iterate quickly: The rapid turnaround of results allows for agile creative development and optimization, which can potentially reduce the time and cost to test new assets.
How Directional experiments work
Directional experiments let you test assets or groups of assets head-to-head in a controlled environment. You can select different assets, create distinct test groups, and choose a success metric relevant to your campaign.
The test results provide you with quick, directional insights into which test group performs best so you can prioritize top-performing assets and refine your creative strategy.
Keep in mind
Directional experiments aim to have an even impression distribution across your test groups to help provide a fair comparison. Here’s a simplified overview of how it works:
- Randomization: When an ad auction occurs, one of your designated test groups is randomly selected to participate. This gives each test group an equal opportunity to be shown.
- Impression serving: All test groups are given similar preference for serving. Test groups with similar assets are more likely to receive a similar number of impressions. However, significant differences in assets, such as video orientation, length, or quality, can still lead to variations in impressions. For instance, different video formats may be eligible for different ad placements, which receive varying amounts of organic traffic.
- Pacing: Impression delivery is paced consistently across each test group to help ensure that they’re shown to people at a comparable rate throughout your test.
By managing impression distribution this way, you get more reliable directional results. This allows you to make informed decisions about which assets to use in your campaigns.
When to use Directional experiments
Directional experiments can be used when you want to:
- Identify top performers: Test new video assets against each other to find the most effective ones.
- Compare existing video assets: Understand which of your current video assets are performing best.
- Evaluate different durations or orientations: Test the impact of different video lengths or formats on campaign performance.
- Assess creative concepts: Evaluate which initial ideas are worth developing further.
- Perform cross-platform validation: Test if top-performing video assets from other platforms perform well on Google.
- Conduct seasonal testing: Assess the impact of specific video assets during promotions or events.
Considerations around Directional experiments
- Directional experiments are currently available for video-only App campaigns for installs (ACi).
- To get significant test results, you should create a new, separate campaign specifically for your test. Your test campaign should:
- Target a geographic area that is excluded from your main campaigns to avoid campaign overlap and ensure cleaner results.
- Have only one ad group that contains any mandatory video assets for your App campaign and video assets that you plan to include in your test.
- Focus on a single variable. For example, test for color or orientation, but not both at the same time.
