Product Understanding
DataSenses's attribution methods
Overview
DataSenses's attribution matches your app users to the source that drove their install. You can use this attribution data to measure campaign performance, run effective retargeting campaigns and more.
DataSenses supports two attribution methods: deterministic attribution and probabilistic modeling for both clicks and impressions. The type of conversion and quality of user engagement determine which method we use.
Deterministic attribution
Deterministic attribution is DataSenses's main attribution method and involves device matching. We collect a unique identifier from recorded engagements and installs, and if both IDs match, we can attribute that engagement to the install. With a 100% accuracy rate, click-based device matching is the most reliable attribution method.
We use deterministic attribution to attribute installs (first app opens) and reattribute (assign new attribution sources to) inactive users.
DataSenses uses the following identifiers for deterministic attribution:
| Identifier | Note |
|---|---|
| Advertising IDs | Used explicitly for advertising purposes. Device users have the option to reset the ID or can refuse to share it (such as limit ad tracking settings). DataSenses stores advertising IDs and they can be used for retargeting purposes. iOS example: IDFA, Android example: GPS ADID (for Android) |
| Device IDs | Permanently attached to the device without users having the option to reset it or deny sharing rights. DataSenses does not record device IDs by default, nor do we store raw device IDs; we only use them for attribution purposes. iOS example: IDFV (for iOS), Android example: Android ID |
| DataSenses tag | Unique IDs created by DataSenses for every click or impression on both iOS and Android. DataSenses only uses Android reftags for attribution matching. iOS example: IDFV (for iOS), Android example: Android ID |
Probabilistic modeling
Probabilistic modeling is DataSenses's secondary attribution method, and uses machine learning to support a statistical approach to measurement.
On iOS 14.5+, probabilistic modeling can be used for owned media, cross-promotion, and consented web-to-app flows.
Platform support
Since iOS and Android operating systems handle user data in different ways, DataSenses may use a different attribution method and fallback depending on the user's device, the advertising channel and engagement source.
Attribution on Android
| Advertising channel | Engagement source | Attribution methods on Android |
|---|---|---|
| Owned Channels (CRM, website, etc) | Mobile Web | Deterministic matching using the Google Play Store Referrer. Fallback: probabilistic modeling |
| In-app | Deterministic matching using the Google Play Store Referrer. Fallback: probabilistic modeling | |
| Self-Attributing Networks | Mobile web & In-app | SANs claim installs based on their own attribution |
Attribution on iOS 14.4 and earlier
| Advertising channel | Engagement source | Attribution methods on Android |
|---|---|---|
| Owned Channels (CRM, website, etc) | Mobile Web | Probabilistic modeling |
| In-app | Deterministic matching. Fallback: probabilistic modeling | |
| Self-Attributing Networks | Mobile web & In-app | SANs claim installs based on their own attribution |
Attribution on iOS 14.5+
| Advertising channel | Engagement source | Attribution methods on Android | SKadNetwork attribution |
|---|---|---|---|
| Owned Channels (CRM, website, etc) | Mobile Web | Probabilistic modeling | Not available |
| In-app | Deterministic matching. Fallback: probabilistic modeling | Not available | |
| Self-Attributing Networks | Mobile web & In-app | SANs claim installs based on their own attribution | Yes, with limited reporting |