How App Tracking Transparency Changed iOS Ad Income

In spring 2021, with iOS 14.5, Apple started requiring apps to ask permission before tracking people across other companies' apps and sites. The prompt is called App Tracking Transparency, or ATT. It changed what advertisers can know about the person who sees their ad, and that shows up in what they are willing to pay.

Five years on, the change is still the main reason iOS ad income is harder to predict than Android income. This guide explains what the prompt does, what is and is not known about opt-in, how it feeds into eCPM, and which parts of the problem you control as a publisher.

What Apple actually changed

Before ATT, every iPhone shared an advertising identifier, the IDFA, with any app that asked. Ad networks used it to recognize the same device across apps, build a picture of its interests and match an ad click to a later install. After ATT, an app must show Apple's system prompt before it can read the IDFA. If the user declines, or never sees the prompt, the identifier comes back as a string of zeros and is useless.

Two details matter for publishers. First, the rule is about tracking across other companies' apps and websites. Using data inside your own app is not what the prompt is for. Second, the prompt is per app, so your own opt-in rate depends on your audience and your timing. Apple's pages on App Tracking Transparency and on user privacy and data use are the primary source, and the wording is worth reading yourself because it gets revised.

Why a missing identifier lowers bids

Advertisers pay for outcomes: installs, purchases, subscriptions. Bidding algorithms learn which impressions lead to those outcomes. With an IDFA the loop is tight, because the advertiser can see that this device saw the ad, and later bought something. Without it, the same advertiser is guessing from coarse signals such as country, device model, app category and time of day.

When a buyer is less sure what an impression is worth, it tends to bid more cautiously. That cautious pricing, spread across many advertisers, is the mechanism behind the eCPM effects people describe. Fill can shift too, because fewer advertisers are confident enough to enter an auction for an anonymous user.

Opt-in: the number everyone asks about

Opt-in rates are commonly reported to be a minority of users, and they vary a lot by app category, audience and how the prompt is introduced. We will not quote a single figure here, because any number depends on whose data it comes from, and your own dashboard is the only one that counts. Your ad network or mediation platform may show the share of iOS requests that arrive with a usable identifier.

To see why the exact rate matters less than it seems, run the arithmetic with invented values. Suppose a tracked impression pays $10 per thousand and an untracked one pays $7. At 25% opt-in, your blended eCPM is 0.25 times $10 plus 0.75 times $7, which is $2.50 plus $5.25, or $7.75. At 15% opt-in it is $1.50 plus $5.95, which is $7.45. Ten points of opt-in moved the blend by $0.30, about 4%. The gap between the tracked and untracked price matters more than the opt-in rate. If that gap were $10 against $4, the same ten points would move the blend by $0.60.

What happened to iOS eCPM

iOS has long paid more per impression than Android, partly because many advertisers see iPhone owners as a valuable audience. After ATT the picture got less tidy. Many publishers reported a period of lower and more volatile iOS eCPMs, followed by partial recovery as networks improved their modelling. We describe it in general terms because the details differ by network, format and month. If the term eCPM itself is hazy, the eCPM explainer covers the arithmetic.

The benchmarks on this site still give iOS a premium, using a multiplier of 1.3 over the Android baseline. For a Tier 1 AdMob interstitial that means $8.00 on Android and $10.40 on iOS, and for a Tier 1 rewarded video it means $12.00 and $15.60. Those are rounded estimates meant for planning, not statements about any one app. The eCPM benchmarks guide shows the full ranges, and the AdMob calculator lets you switch the platform setting to see the gap for your own audience.

SKAdNetwork in plain terms

SKAdNetwork is Apple's system for measuring installs from ads without revealing who the user is. An ad network shows an ad, the user installs the advertised app, and Apple sends a delayed, limited report to the network confirming the install. The report is deliberately coarse. It can include a small value that the advertiser's app sets to hint at early behavior, but it never includes a device identifier, and Apple may withhold details when too few installs came from one campaign. Apple's SKAdNetwork documentation describes the current version, and Apple has since introduced AdAttributionKit as a successor, so check which one your SDKs use.

What does this have to do with a publisher? Your app is where the ad is shown, and it takes part in the process. Each ad network that wants credit for installs has a SKAdNetwork identifier, and your app's configuration file needs to list the identifiers of the networks you work with. If you leave them out, those networks' advertisers cannot measure installs from your app, and they bid lower or not at all.

What publishers can control

You cannot undo the policy, but several things remain in your hands.

Add the identifiers. Your mediation provider's documentation lists the SKAdNetwork entries to copy into your app. Repeat the check whenever you add a network. Keep SDKs current, since older versions may predate newer privacy handling. Ask for consent properly: use the system prompt with a short pre-prompt screen that explains why you ask, shown after the user has seen some value in the app rather than on the first launch. Apple's rules generally forbid pushing people to say yes, so check the current terms before attaching any incentive.

Use competition. Mediation with bidding lets networks with the best privacy-safe modelling win the impression, which matters more when individual signals are weaker. The mediation guide explains the setup and the AppLovin MAX calculator shows what that competition might add. Finally, improve what you can see: contextual signals such as app category, region and session depth are untouched by ATT, and the levers in seven ways to earn more work regardless.

Mistakes that cost iOS revenue

A few errors come up often. The first is judging iOS by a blended number. Report iOS and Android separately, because the blend hides which one moved. The second is prompting at the wrong moment, such as before the user knows what the app does, which tends to lower opt-in. The third is a forgotten network entry after adding an SDK, which silently narrows demand.

The fourth is mixing up two separate rules. ATT covers tracking across apps on iPhones and iPads. Consent requirements for people in the EEA, the UK and some other regions are a different system, handled with a consent message from your ad network, and you may need both. A gap in either can reduce what advertisers will pay. For a first estimate of your own iOS income, set the platform in the main calculator, read the methodology for how the multiplier is applied, and use the glossary for any term that is unfamiliar.