What Is a Device ID?
A device ID is a unique alphanumeric string assigned to a mobile device that allows advertisers, publishers, and analytics platforms to identify, track, and target that device across apps and ad networks. Unlike cookies, which are browser-based and easily cleared, device IDs persist at the operating system level, making them a more stable signal for mobile attribution and audience segmentation.
How Device IDs Work
Every smartphone operating system generates an advertising-specific identifier that apps can read with user permission. On iOS, this is the IDFA (Identifier for Advertisers). On Android, it is the GAID (Google Advertising ID), sometimes called the AAID (Android Advertising ID). Both are 32-character hexadecimal strings formatted as UUIDs, for example: A1B2C3D4-E5F6-7890-ABCD-EF1234567890.
When a user installs an app and consents to tracking, the app reads the device ID and passes it to ad networks, measurement partners, and analytics SDKs. Those platforms use the ID to match ad impressions and clicks to downstream installs or purchases, a process known as mobile attribution.
iOS vs. Android: Key Differences
| Platform | Identifier Name | Opt-Out Mechanism | Default State |
|---|---|---|---|
| iOS | IDFA | App Tracking Transparency (ATT) prompt | Opt-in required; ~25-35% consent rate |
| Android | GAID / AAID | User opt-out in device settings | On by default; opt-out rate ~10-15% |
What Device IDs Are Used For
Ad Targeting and Retargeting
Advertisers upload lists of device IDs to platforms like Meta, Google, and The Trade Desk to serve ads directly to known users. A retail brand can, for example, retarget users who abandoned a cart in its mobile app by matching those users’ device IDs against an audience segment on a demand-side platform.
Frequency Capping
Device IDs allow ad servers to limit how many times a single user sees a given creative. Without a persistent identifier, the same user could receive dozens of impressions of the same ad across different apps in a single day, which drives up costs and damages brand perception. Frequency capping tied to device ID typically targets 3 to 5 impressions per user per day as a baseline.
Attribution and Measurement
Mobile measurement partners (MMPs) such as AppsFlyer, Adjust, and Branch use device IDs as the primary key for matching an ad click to an app install. The standard attribution model works as follows:
- User sees an ad; the ad network records the device ID and a timestamp.
- User clicks the ad; click event logged with device ID.
- User installs the app; the MMP reads the device ID on first open.
- MMP matches the install device ID to the click record within the attribution window (commonly 7 or 30 days).
- The MMP credits the install to the originating campaign.
This process is a core component of last-click attribution on mobile. Accuracy depends heavily on ID availability, which is why the iOS ATT changes reduced measurable attribution rates by an estimated 30 to 50 percent for some advertisers after the April 2021 rollout.
Lookalike Audience Modeling
Platforms ingest device ID lists to build lookalike audiences. Meta’s Advantage+ audiences, for instance, use a seed list of high-value customer device IDs to find statistically similar users across its network. The quality of the seed list directly affects match rates and audience quality.
The Privacy Shift and Its Advertising Impact
iOS: The ATT Turning Point
Apple’s App Tracking Transparency framework, introduced in iOS 14.5 in April 2021, required apps to request explicit user permission before accessing the IDFA. Consent rates across the industry settled between 25 and 40 percent depending on the app category, according to data from AppsFlyer’s 2023 State of Partners Report. This cut the volume of deterministic device ID signals available to advertisers and accelerated adoption of probabilistic attribution and privacy-preserving alternatives.
Android: A Similar Shift in Progress
Google signaled parallel changes for Android through its Privacy Sandbox initiative, though the timeline for restricting the GAID has been delayed multiple times. As of early 2026, the GAID remains available on Android devices but is expected to face tighter restrictions. Advertisers who built measurement strategies around deterministic device ID matching are increasingly treating GAID availability as temporary, not permanent.
Alternatives When Device ID Is Unavailable
When a user opts out or resets their device ID, advertisers and measurement platforms turn to several fallback methods:
- SKAdNetwork (SKAN): Apple’s privacy-preserving attribution framework that reports campaign-level conversions without exposing individual device IDs. Aggregated, delayed, and limited in granularity.
- Probabilistic matching: Uses IP address, device type, OS version, and screen resolution to infer attribution without a deterministic ID. Less accurate than device ID matching, with false-positive rates typically between 5 and 15 percent.
- First-party identifiers: Hashed email addresses or phone numbers from logged-in users, passed via clean rooms or data collaboration platforms. More durable than device IDs in privacy-restricted environments.
- Modeled conversions: Meta’s Conversions API and Google’s enhanced conversions use machine learning to fill gaps in measured data when device IDs are absent.
Device ID Resets and Their Effect on Campaigns
Users can reset their advertising device ID at any time in device settings, generating a new random string. From the advertiser’s perspective, that user becomes a new, unknown device. This affects lifetime value calculations, suppression lists, and retargeting audiences. Campaigns built on audience lists with high reset rates will show inflated reach and reduced frequency efficiency. A standard hygiene practice is to refresh device ID lists every 30 to 90 days to remove stale identifiers.
Device ID in Programmatic Buying
In the programmatic advertising ecosystem, device IDs travel inside bid requests as standardized fields. The OpenRTB protocol, maintained by the IAB Tech Lab, includes dedicated fields for the IDFA and GAID so that demand-side platforms and ad exchanges can pass the identifier without custom integrations [VERIFY: OpenRTB field path for device identifier is device.ifa, not app.user.ifa as sometimes cited]. When a publisher’s app sends a bid request without a device ID, many DSPs bid lower or skip the impression entirely, because it cannot be targeted or attributed deterministically. That gap regularly produces effective CPMs 20 to 40 percent below ID-available inventory.
Key Metrics Tied to Device ID Availability
| Metric | With Device ID | Without Device ID |
|---|---|---|
| Attribution accuracy | Deterministic (near 100% match for consented users) | Probabilistic (70-90% depending on signals) |
| Retargeting reach | High (direct audience matching) | Low (limited to modeled or contextual) |
| Frequency cap reliability | Precise per-device capping | Estimated, often over-delivers |
| eCPM (publisher) | Baseline | 20-40% below baseline |
Frequently Asked Questions About Device ID
What is a device ID in mobile advertising?
A device ID is a unique identifier assigned to a mobile device at the operating system level, used by advertisers to track, target, and attribute users across apps. On iOS, it is called the IDFA; on Android, the GAID. Unlike browser cookies, device IDs persist across app sessions until the user manually resets them in device settings.
What is the difference between IDFA and GAID?
The IDFA (Identifier for Advertisers) is Apple’s advertising identifier for iOS devices. The GAID (Google Advertising ID), also called the AAID, is Android’s equivalent. Both are 32-character UUID strings, but they differ in default behavior: the IDFA requires opt-in consent as of iOS 14.5, while the GAID is on by default with an opt-out option in device settings.
What happened to device IDs after Apple’s iOS 14.5 update?
Apple’s App Tracking Transparency framework, launched in April 2021, required apps to ask users for explicit permission before reading the IDFA. Consent rates settled between 25 and 40 percent industry-wide. For some advertisers, measurable attribution rates dropped 30 to 50 percent as a direct result, pushing the industry toward probabilistic measurement and privacy-preserving alternatives like SKAdNetwork.
Can advertisers target users without a device ID?
Yes, but with reduced precision. When a device ID is unavailable, advertisers fall back on probabilistic matching, first-party identifiers like hashed emails, Apple’s SKAdNetwork for aggregate attribution, or modeled conversions from Meta’s Conversions API. None of these methods matches the accuracy or addressability of deterministic device ID matching.
How do device ID resets affect campaign performance?
When a user resets their device ID, the old identifier becomes inactive and the device appears as a completely new, unknown user to advertisers. This inflates apparent reach, breaks suppression lists, and degrades retargeting precision. Refreshing device ID audience lists every 30 to 90 days is standard hygiene to remove stale identifiers before they distort campaign data.
Related Terms
Device ID intersects with several adjacent concepts in mobile marketing. Understanding mobile attribution, programmatic advertising, and lookalike audiences provides fuller context for how device IDs function within the broader advertising stack. For measurement strategy, see also last-click attribution.