Deep links are only successful when they open the intended app screen, preserve campaign context, and provide a useful fallback when the app is unavailable. This guide gives marketers and developers a repeatable way to test Universal Links and Android App Links across devices, browsers, install states, QR codes, email, and messaging apps before a release or campaign launch.
Overview
A mobile deep link contains a destination beyond the app's home screen, such as a product, article, account, or checkout page. On iOS, the main web-to-app mechanism is the Universal Link. On Android, the comparable mechanism is the App Link. Both use an HTTPS URL, but each platform verifies the relationship between the website domain and the installed app differently.
The test objective is not simply “does the app open?” A release-ready deep link should answer four questions:
- Does the correct app screen open when the app is installed?
- Does the browser or store experience work when the app is not installed?
- Are path, query string, and campaign parameters preserved?
- Does analytics record the click, install, open, and destination consistently enough for reporting?
Separate routing failures from attribution failures. A link can open the right screen while losing utm_source, a campaign identifier, or an internal referral value. Conversely, analytics can record a click even though the user lands on a generic home screen. Test both outcomes independently.
| Area | iOS Universal Links | Android App Links |
|---|---|---|
| Web association | Uses an Apple App Site Association file served by the website. | Uses an asset links JSON file served by the website. |
| App configuration | Requires the app's associated-domains configuration and matching website paths. | Requires an HTTPS intent filter with matching host and path rules. |
| Primary test question | Does the verified domain open the intended route in the installed app? | Is the domain verified for the installed app and selected as the handler? |
| No app installed | Should remain a useful web experience or follow the planned install path. | Should remain a useful web experience or follow the planned install path. |
| Common investigation area | Association-file paths, entitlements, app state, and link-opening context. | Asset statement, package and certificate values, verification status, and intent filters. |
Platform behavior can change with operating-system releases, browser updates, app settings, and the way a link is opened. Treat this table as a planning reference, then confirm behavior on the OS versions and devices that matter to your audience.
How to estimate
Use a small test matrix rather than relying on one successful tap. Estimate coverage by multiplying the number of meaningful variables, then remove combinations that do not apply. The basic planning formula is:
Test cases = platforms × app states × link surfaces × destination types
For example, two platforms, three app states, four link surfaces, and two destination types produce 48 theoretical combinations. You may not need to execute every combination if your risk assessment shows that some are equivalent, but recording the matrix makes those exclusions deliberate.
Use these categories:
- Platforms: supported iOS and Android versions, including at least one older supported version and the newest release available to your team.
- App states: installed and closed, installed and already open, and not installed. If your app has login or onboarding states, include those as separate states.
- Link surfaces: mobile browser, email, QR code, and messaging app. Add social or ad placements when they are important campaign sources.
- Destination types: a public content page and a route requiring authentication or special handling.
Assign each case a simple result: pass, fail, blocked, or not applicable. A useful release estimate is not a guessed conversion rate; it is the percentage of planned cases that pass without a known attribution or fallback defect:
Readiness rate = passing cases ÷ executable cases × 100
Do not use this rate as a business performance metric. It is a QA indicator. A single failure on a high-value checkout or campaign route may require action even when the overall percentage is high.
Inputs and assumptions
Domain-association checklist
Before testing devices, verify the pieces that allow the operating system to associate the domain with the app:
- Use the intended HTTPS domain, including the exact host used in campaign links.
- Confirm the iOS association file is available at the expected well-known location and contains the correct app identifiers and permitted paths.
- Confirm the Android asset links file is available at its expected well-known location and contains the correct package and signing certificate information.
- Check that files are returned as valid JSON, without an HTML error page, unexpected redirect, authentication requirement, or accidental environment-specific values.
- Compare app configuration with the production domain, path rules, package identifier, team or application identifier, and signing configuration used by the build under test.
- Test the exact path patterns used by short links, QR codes, email links, and ads. A root path working does not prove that a nested route works.
Test-device matrix
Record the device model, OS version, app build, browser or source application, app state, URL, expected route, actual route, and attribution result. Include both a clean device or simulator state and a device with prior app history. For QR testing, scan the printed or rendered code from the same approximate distance and context users will encounter, rather than testing only the raw URL.
UTM-preservation assumptions
Keep campaign parameters on the web URL unless your routing design explicitly maps them into an internal deep-link payload. A representative campaign URL might be:
https://go.example.com/summer/shoes?utm_source=newsletter&utm_medium=email&utm_campaign=summer_launch&product_id=123
The expected result is that the app receives the product destination and the measurement layer can access the campaign values. If a redirect is involved, test whether it preserves the complete query string. If parameters are transferred into an app-specific format, document the mapping and verify that sensitive values are not exposed unnecessarily in URLs or logs.
Worked examples
Example 1: installed app from email
Send a controlled email containing a branded or shortened HTTPS link. On an iPhone with the app installed, tap the link from the mail application and confirm the expected content screen opens. Repeat with the app closed. On Android, verify that the intended app handles the link rather than a browser or an unintended app. Record whether the source, medium, campaign, and content identifiers arrive in analytics.
Example 2: app not installed
Open the same URL on a device without the app. The fallback should be intentional: a mobile web page, an app-store route, or another documented destination. After installing the app, repeat the original campaign path and check whether the user reaches the intended content or merely the app home screen. If deferred deep linking is supported, define exactly which event carries the original destination and campaign context, then test it as a separate capability.
Example 3: QR code and messaging app
Generate the QR code from the final campaign URL, not a temporary staging URL. Test scans on both platforms with the app installed and absent. Then paste the same URL into a messaging app and tap it from the conversation. Compare the destination and attribution with the browser result. Some surfaces may preview, rewrite, or delay navigation, so a successful browser test does not eliminate the need for surface-specific testing.
Fallback decision tree
- Is the URL valid and reachable? If no, repair the redirect or destination before testing app behavior.
- Is the app installed? If yes, test the verified app route. If no, continue to the planned web or install fallback.
- Does the association verification match this domain and path? If no, use the web fallback while correcting configuration.
- Does the destination require authentication? If yes, preserve the intended route safely through sign-in; do not discard it silently.
- Were campaign values preserved? If no, inspect redirects, parameter parsing, and analytics handoff separately from UI routing.
When to recalculate
Re-run the matrix before every major app release, campaign launch, domain change, redirect change, QR print run, or analytics SDK migration. Revisit it after an operating-system or browser release when link-opening behavior may differ, and whenever support reports that users land on the wrong screen or lose campaign attribution.
For a practical release workflow:
- Freeze the final URL patterns, destinations, UTM naming, and fallback policy.
- Validate association files and app configuration in the release candidate.
- Generate test links through the same short-link, redirect, QR, or campaign system used in production.
- Execute the device and surface matrix, recording evidence for each result.
- Review routing, fallback, and analytics as separate pass criteria.
- Retest every failed or blocked case after the fix, including a clean-install case where relevant.
- Archive the matrix with the build number, domain, test date, and known limitations.
Keep a small set of permanent smoke-test links for a public page, an authenticated route, a campaign URL, and a QR destination. Running those links after each release provides an early warning, while a fuller matrix remains necessary for launches and domain migrations. Pair this process with a UTM QA workflow, automated link health checks, and documented short-link lifecycle rules so routing and measurement remain maintainable after launch.