Testing attribution before you spend on ads

What a pre-launch test proves

A test proves that the chain from click to event is connected in your app, on your build. It answers four questions: does a tap on a link get recorded, is the install matched to that tap, do events arrive with the right values, and does the link open the right screen. It does not prove that numbers will be right at scale, which is why you check again with live data later.

What you need

• A build of the app that contains the SDK, installable outside the store or through a test track.
• One iPhone and one Android phone if you ship on both.
• Each phone registered as a test device in the attribution tool, so repeat installs count as new.
• One tracking link made for the test, with a campaign name such as "internal-test" that you can filter out later. How tracking links work explains what the link carries.

The test, step by step

1. Delete the app from the phone.
2. Tap the test tracking link on that phone. Let it reach the store page or the download page, then stop.
3. Install the test build and open it. The first open is what the tool counts as the install.
4. Look for the install in the dashboard's real-time or test view. Confirm that it carries your test campaign name.
5. Trigger each key event once: sign up, start a trial, make a sandbox purchase.
6. Confirm that each event appears once, under the same campaign, with the right revenue value and currency.
7. With the app now installed, tap a link that points at a specific screen and check that the screen opens.
8. Delete the app, tap that link again, install and open. The same screen should open after install. This is the deferred deep link path.

Write down the time of each step. When something fails, the timestamps let you or the vendor's support find the click and the install in the logs.

Reading a failed test

A note on iOS test phones

Apple requires an app to get the user's permission through App Tracking Transparency before it tracks them or reads the device's advertising identifier.^[1] Without that permission the identifier is all zeros.^[1] Run the iOS test twice, once allowing tracking and once refusing it, so you know what your reports look like in both cases.

When to repeat the test

Repeat steps 1 to 6 before any release that updates the SDK, changes an event or touches the purchase flow. It is a short check once the phones are registered. Decide who owns it using who does what, and keep the test link for reuse.

After the test passes

Ship the build and wait for it to reach users before you start a campaign. Then move to the live-data checks in what to check during a free trial, beginning with a comparison of reported revenue against your store data.