Test which Custom Product Pages perform best for your Apple Ads campaigns. This guide walks you through how to configure and launch a CPP experiment in SplitMetrics Acquire.

Make sure the following are ready before opening the setup form:

✅ Active ad group with historical data. The ad group must be actively delivering and have at least 30 days of impressions, taps, and downloads. This historical data is used to estimate experiment duration and expected traffic.

✅ Custom Product Pages created in App Store Connect. Your CPP variations must already exist, be approved by Apple, and be available for use. You need at least two product pages, including the Default Product Page, to run an experiment.

✅ CPP assets are finalized. Avoid making changes to screenshots or promotional text during the experiment. Any updates to creative assets can affect the validity of the comparison.

✅ Primary KPI defined. Decide which metric matters most for your test, such as TTR, Download Rate, CPT, CPA, or another available metric. This metric is used to determine the winner.

✅ Understand what is frozen during the experiment. When an experiment runs, the original ad group is paused. All automations, including Rules, Bid Strategies, Budget Allocation, Keyword Expansion Strategies, and CPP Scheduler, are disabled for experiment objects. Bulk updates will skip items included in the experiment. Plan your workflow accordingly.

Apple Ads was not built specifically for A/B testing. There is no native way to split traffic evenly across product pages. Instead, the auction determines how impressions are distributed and optimizes for its own objectives rather than a controlled split.
​

Because of this, every CPP experiment must work within the constraints of the auction. Each method approaches this limitation differently, and the right choice depends on your traffic volume, the number of variations you want to test, and how quickly you need results.

🔜 Switch (recommended for most cases)
Switch is the most reliable way to test CPPs. Only one variation is active at a time, which avoids auction competition entirely. Each variation receives full traffic during its period.
​
With weekly rotation, each variation collects data across a full week, capturing both weekday and weekend behavior. This makes Switch the safest option for drawing reliable conclusions.
​

The main consideration is timing, as a two-variation test typically requires at least 14 days. Supports up to 4 product pages at either the Ads or Ad Group level.

Dayparting
Dayparting is a strong alternative that also avoids direct competition in the auction. Variations rotate within each day on a fixed schedule. Both variations receive daytime and evening traffic each day, which helps reduce time-of-day bias.

The system controls when variations are activated via ad group scheduling, thereby minimizing the impact of Apple learning phases. Limited to 2 product pages.

Parallel
All variations run simultaneously, making this the fastest way to collect data. The system creates identical ad groups with the same keywords, bids, and audiences, so the only difference is the product page.

However, since all variations compete in the same auction, Apple Ads determines how traffic is distributed. This can lead to uneven splits, as the auction may favor one variation over another.

This method works best for high-traffic ad groups where speed matters and uneven traffic distribution is acceptable. It supports up to 4 product pages.

Go to CPP Experiments in the left sidebar and click Create CPP experiment.

Choose your app from the App dropdown. The list includes all apps connected to your SplitMetrics Acquire account.

Click Select ad group to choose an eligible ad group. The system will display its storefront next to the name.

Click + Add product page and select 2 to 4 product pages, depending on your chosen method.

Key details:

Once you’ve made your choice, click Select. The chosen pages will appear as Variation A, B, and so on, with previews.
​

Click Set as control variation under the page you want to use as the baseline, typically the currently active CPP. This variation will be marked as Control in the test.

Enter a clear, unique experiment name (e.g., "Holiday_CPP_test_US" or "Christmas test"). Optionally, add a description to document the goal or context.

Choose one of the three methods: Dayparting, Parallel, or 🔜 Switch. The selected method will be highlighted.

Click More details under each method to see a visual timeline showing how each option works.

After selecting the method, you’ll see two additional options: Level and Period.
​

For Dayparting and Parallel, these settings use default values. For the 🔜 Switch method, you can customize both options depending on how you want to run your test.
​

- Ad group: Creates a separate ad group for each variation. Each variation runs independently with its own auction history, which provides the most accurate and reliable comparison. This option is recommended if you want cleaner results, but it will increase the number of ad groups in your campaign.

- Ads: Rotates variations within the same ad group. This setup is simpler and keeps your campaign structure unchanged. However, all variations share the same auction history and performance signals, which can make results slightly less precise.

- Weekly (recommended): Each variation runs for a full 7-day period. This allows you to capture the full weekly traffic cycle and reduces the impact of Apple Ads learning phases after each switch. This option is recommended for most campaigns.

- Daily: Each variation runs for one day and rotates at midnight. This helps you collect data faster, but frequent changes may trigger short learning phases, which can affect performance stability. This option works best for high-traffic ad groups with enough daily volume.

From the Experiment primary metric dropdown, select the KPI that will be used to determine the winner of your experiment.
​

Available options include TTR, Download Rate (Tap-Through), CPT, Avg CPA (Tap-Through), and other Apple Ads funnel metrics.

How to choose the right metric

- TTR: Shows how well your ad attracts attention and drives taps. This is an early signal based on your screenshots and messaging. It needs less data, so results come faster, but a higher TTR doesn’t always mean more installs.

- Download Rate (Tap-Through): Shows how well your product page turns taps into installs. This is usually a more reliable indicator of overall performance than TTR.
​

- CPT: Helps you focus on the cost of each tap. Useful when your goal is to improve cost efficiency at the tap level.
​

- Avg CPA (Tap-Through): This is the most complete performance metric, but it usually needs more data, so experiments may take longer to finish.

In the Experiment estimations section, set the timeline for your experiment.

- Start date: By default, the experiment starts at 00:00 (midnight) the next day in your organization’s time zone. You can also choose a different future start date if needed.

- End date: Estimated completion date showing when the experiment is expected to stop and produce a result, such as a winner, no clear difference, or not enough data.

- Estimated duration: The system automatically estimates how long the experiment will take based on your ad group’s last 30 days of performance and your current setup.

- Run experiment until winner is found checkbox: Select this option if you want the system to automatically stop the experiment once a clear winner is identified, instead of running it for a fixed time.
​

Click Continue at the bottom of the page. A Summary window will appear with all your experiment settings.

Carefully review the setup, including the objects that will be created for the test. If anything doesn’t look right, click Cancel to go back and make changes.
​

When everything looks good, click Start CPP testing to launch your experiment.

After you click Start CPP testing, the system automatically takes over and runs the experiment for you.
​
​1. Test ad groups are created

For Parallel and Switch (Ad Groups level), the system creates a cloned ad group for each variation. Each clone keeps the same setup as the original, including targeting, bids, CPA caps, keywords, negative keywords, and audiences.
​

For Switch (Ads level), new ads are created within the existing ad group instead.

2. Original ad group is paused

At the scheduled start time, the test ad groups begin serving. After a short overlap period (around 15 minutes) to ensure a smooth transition, the original ad group is paused and stays paused for the full duration of the experiment.
​
​3. Automations are disabled

All automation rules, Bid Optimization Strategies, Budget Allocation settings, Keyword Expansion Strategies, and CPP Scheduler actions are automatically turned off for all experiment objects.
​

Any bulk updates in Ads Manager will skip experiment-related items.
​
​4. Data collection starts
The system begins collecting data based on your selected method:

Your experiment is now live and collecting data. Next, you can track performance, monitor health signals, and see when results are ready.
​

Next article: Understanding Your Results and Making a Decision