App Attribution After iOS 14: What Actually Works

When Apple rolled out App Tracking Transparency in iOS 14.5, it broke the attribution industry. The IDFA — the identifier that let tools like AppsFlyer and Adjust track users across apps — went from available-by-default to opt-in. Most users opted out. Depending on whose numbers you believe, somewhere between 75-90% of users decline tracking when asked.

That was 2021. We're five years in now and the dust has settled, but the confusion hasn't. Indie developers especially are stuck in a weird middle ground: they know they should track attribution, they've heard the old tools don't work as well, and they're not sure what to do instead.

Here's where things actually stand.

What ATT broke

Before ATT, the flow was straightforward. User sees an ad. Clicks it. Downloads app. Attribution tool matches the IDFA from the ad click to the IDFA in the app. Done. You know which ad drove which install.

After ATT, most users don't share their IDFA. The attribution tool can't match the click to the install through that identifier anymore. The whole system that the mobile attribution industry was built on — deterministic matching via device IDs — stopped working for the majority of users.

What the big players did

Probabilistic matching (fingerprinting). AppsFlyer, Adjust, and others pivoted to probabilistic attribution. They collect device signals — IP address, screen size, OS version, carrier, language, timezone — and use statistical models to guess which ad click matches which install. It's less accurate than IDFA matching, but it works reasonably well for large-scale campaigns.

Apple has been cracking down on this too. Their privacy policies technically prohibit fingerprinting, and they've gradually tightened enforcement. Whether they'll fully block it is an ongoing question, but the direction is clear.

SKAdNetwork (now AdAttributionKit). Apple's own attribution framework. Gives advertisers aggregated, delayed, privacy-preserving conversion data. The data comes back in batches, up to 72 hours late, with limited granularity. You get told "some users who saw this ad converted" but not which specific users.

SKAdNetwork is useful for large advertisers spending thousands on ad campaigns. For an indie dev spending £50 on a test campaign, the aggregated delayed data is nearly useless.

Web-to-app attribution. Some tools track the web session before the user goes to the App Store. If someone clicks a link on your website, you can identify them via cookies or session data. When they install and open your app, you check if there's a matching recent web session. This works, but only for traffic that goes through your website first.

What actually works for indie developers

Here's the honest picture. If you're an indie dev or small team, you probably don't have the budget for AppsFlyer ($500+/month), you don't run enough ad spend for SKAdNetwork data to be meaningful, and you don't have a team to interpret probabilistic attribution models.

What you need is simpler: which of your marketing efforts drove a purchase?

Link-based attribution. Create a unique tracked link for each marketing channel. TikTok gets one link. Instagram gets another. Your email newsletter gets a third. When someone clicks the link, the click is recorded server-side. When they purchase in your app, the purchase is matched to the original click.

This bypasses the ATT problem entirely. You're not tracking users across apps. You're not using the IDFA. You're tracking a click on a link you own, then matching it to a purchase event in your own app. That's first-party data, which Apple is fine with.

How it works technically. The tracked link redirects to the App Store (or your website, or wherever). On the redirect, the server records the click with a timestamp and whatever identifying info is available — typically just an anonymous click ID stored in a cookie or passed through as a URL parameter.

When the user installs your app and opens it, the SDK checks for a matching click. If it finds one (usually by checking if the device recently visited a tracked link), it associates that user with the original campaign. Future purchases from that user are attributed to that campaign.

Limitations. This only works when users actually click your link. If someone sees your TikTok, searches your app name in the App Store manually, and downloads it without clicking anything, that install isn't attributed. This is the same gap that affects all non-fingerprinting attribution.

For most indie apps, this is an acceptable tradeoff. You won't capture every conversion, but you'll capture enough to see clear patterns about which channels work and which don't.

Comparing methods

The practical setup

If you're using RevenueCat or Stripe for purchases, link-based attribution is straightforward to set up. With LinkOwl:

1. Create links for each channel (takes 2 minutes)
2. Add a lightweight SDK to your app (two lines of code)
3. Connect the RevenueCat webhook (point it at your LinkOwl endpoint)
4. Share the tracked links in your marketing

No IDFA required. No ATT prompt. No privacy concerns. The tracking is on your own links, not on the user's device.

You pay 5p per attributed purchase. No monthly fee. If ATT killed your old attribution setup and you've been flying blind since, this gets you back to making informed decisions about your marketing spend.

What's coming next

Apple's direction is clear: less tracking, more privacy. Each iOS update tightens the screws further. Any attribution method that relies on collecting device signals or cross-app tracking is on borrowed time.

First-party, consent-based attribution — where you track interactions with your own links and your own purchase events — is the approach that aligns with where Apple is heading. It's not as comprehensive as the old IDFA world, but it's sustainable. It won't break with the next iOS update. And for indie developers who just need to know "is my TikTok marketing working?", it's more than enough.