Skip to main content
All CollectionsAffise MMP for Marketers
Affise MMP attribution model
Affise MMP attribution model
Natalya Yefimenko avatar
Written by Natalya Yefimenko
Updated over 2 weeks ago

Affise MMP system defines a media source that encourages the user to install or re-engage with the application, and attributes a user's action to this source using the Affise MMP attribution model.

Definition of attribution

Attribution is the fact of determining what the user prompted: to install the application or perform any action after installation. Such post install actions can be re-engagement, as well as re-attribution.

The result of attribution can be:

  • Non-organic traffic: the user interacted (usually by showing or clicking) with the traffic source.

  • Organic: The user did not interact with the traffic source. For the ease of understanding, it's often referred to the user as organic. This is not entirely true, since users who install the application organically are not attributed at all, but only displayed as organic.

Mobile attribution is important for optimizing user engagement, re-engagement efforts, and the results of the application and advertising campaigns in general.

Mobile app installation

When the app is installed and running, Affise MMP receives event data via its SDK or S2S integration. The installation data package includes device identifiers, IP address, and user data (if permitted by privacy policies and user consent). This data is used to generate reports for advertisers or developers to assess data about ad campaign effectiveness and user behavior.

🔍 For this, it's important to use Affise MMP's format of unique user identifiers and events.

An attribution model defines how event scores are assigned to touchpoints in conversion paths. Each platform (Google Play, App Store, Facebook, etc.) and mobile measurement company has its own model, counting installations and events differently. The key is clear, unbiased rules help advertisers optimize campaigns and compare user experience quality.

Understanding the models and rules of partners, especially Affise MMP, is crucial. Affise MMP maps data only when specific events occur:

  • User Acquisition: Installation is recorded and assigned after the app is downloaded and launched. Affise uses the first launch timestamp, unlike ad networks (interaction time) or app stores (download time).

  • Retargeting: Includes re-engagement and re-attribution events.

Attribution methods

Affise MMP uses a number of attribution methods listed in the following table:

Method

Uses an ID

Technique

Attributed by

Android

iOS

Windows

Referrer

Yes

Deterministic

Affise MMP

Yes

No

Yes

Device ID matching

Yes

Deterministic

Affise MMP

Yes

Yes

Yes

Probabilistic modelling

Yes

Probabilistic

Affise MMP

Yes

Yes

No

Aggregated Advanced Privacy

No

Aggregated

Affise MMP

No

Yes

No

Preinstall

No

Deterministic

Affise MMP

Yes

No

Yes

SKaD Network

No

Deterministic

Affise MMP

No

Yes

No

Attribution methods are used in accordance with the priorities, decision-making factors are listed below. With a new installation (if more than one valid interaction), Affise MMP gives preference to clicks over impressions and deterministic rather than probabilistic methods.

Decision-making factors - Reconciliation Factors

Priority 1 — Lookback window

The Lookback Window determines how far back from the installation moment an action should be considered for attribution. It applies to device mapping and Fingerprint, for both clicks and installations. For example, a one-day lookback window means the click must occur within 24 hours before installation.

The lookback window is configured in the app settings or ad campaign settings. For users from ad networks, it is defined directly in the ad network's interface.

Priority 2 — Matching integrity

The Affise MMP approval system works in waterfall format. Attribution for installation is based on clicks and impressions in the following priority order:

Clicks Matching

🔍 Clicks have a higher priority for attribution than impressions.


Mapping scheme:

1. Device-Based Matches

  • Self-Attributing Network (SAN) claims including Facebook, Apple Search Ads and Google

  • Device ID (including progressive reconciliation)

  • Google Referrer

2. Standard Fingerprinting (IP Address + User Agent + Time)

  • No Device ID present on click

  • Unmatched Device ID present

3. IP-Only Fingerprinting

  • No Device ID present on click

  • Unmatched Device ID present

4. IP Range-Only Fingerprinting

  • No Device ID present on click

  • Unmatched Device ID present

Impression Matching

After attempting to match clicks, all installations without attributes are subject to matching with impressions.

Mapping scheme:

  1. Device-Based Match

  • Completed View (this is an impression type defined by each network often referring to a completed video view)

  • All other impression types (any other tracked impressions)

2. Standard Fingerprinting

  • Completed View

  • All other impression types

3. IP-Only Fingerprinting

  • Completed View

  • All other impression types

4. IP Range-Only Fingerprinting

  • Completed View

  • All other impression types

Parameters used in the mapping

Install referrer (Google, Android only)

Android apps from Google Play and other stores are typically attributed via the installation link method, using the original URL clicked by the user before redirecting to the store. This is the main Android attribution method. Currently, Google Play and Huawei App Store support install referrer attribution.

Device ID Matching

The advertising company that has access to the user's device sends the device ID to Affise MMP at the click URL or when notifying Affise MMP about the impression. This allows Affise MMP to match the click user ID with the device ID obtained using the Affise MMP SDK.

Available IDs:

  • Android devices: having Google Play services: GAID

  • Android devices without Google Play services: OAID, Android ID, IMEI, Fire ID

Probabilistic modeling

Probabilistic modeling is a statistical method that uses machine learning to evaluate the effectiveness of a campaign.

Probabilistic modeling parameters:

  • Collected initially when clicked or displayed (if used)

  • Data collected when the app is launched

Implementation characteristics:

  • Uses statistics is not based on unique identifiers.

  • It is a backup method used when there's no referent or advertising identifiers. Priority is given to deterministic attribution methods.

  • Loses to clicks with matching referrer or ID if they occur in the lookback window.

  • The attribution window is determined dynamically by Affise, depending on the user's network. The window duration is adaptive, but shorter than other methods (up to 24 hours).

  • Click-through - end-to-end probabilistic modeling is always enabled.

  • View-through - end-to-end probabilistic modeling should be enabled on the application settings page and on the non-SRN integration tab.

Pre-install

Affise MMP attributes app installations even if the app was pre-installed before device purchase. Installations are tracked at first launch via the SDK API, without prior user interactions.

⚠️ If there is interaction from the other source before the first launch, the pre-installation has a higher priority.

Attribution methods per media source

Owned Media

Media source

Feature

Attribution method Android

Attribution method iOS

Owned media: email (including ESPs), SMS, social media posts, influencers/affiliates, print, etc.

SmartLink

  • Install referrer

  • Probabilistic modeling

  • Probabilistic modeling

Owned mobile website/landing page with paid or organic incoming traffic

Banners, Adv

  • Install referrer

  • Probabilistic modeling

  • Probabilistic modeling

Owned mobile apps

User invites/referral

  • Probabilistic modeling

  • Probabilistic modeling

Owned mobile apps

Cross-promotion

  • Device ID matching

  • Probabilistic modeling

Paid media

Media source

Attribution method Android

Attribution method iOS

SRNs

  • Device ID matching

  • Device ID matching

  • AAP (iOS 14.5+)

  • SKAdNetwork (iOS 14+)

Ad-Networks (non-SRNs)

  • Install referrer

  • Device ID matching

  • Probabilistic modeling

  • Device ID matching

  • Probabilistic modeling (before iOS 14.5)

  • AAP (iOS 14.5+) - SKAdNetwork (iOS 14+)

Pre-Installed on device

  • Pre-installs


User engagement attribution types - time of engagement

The last click or impression time determines attribution when other factors are equal. Also, device identifier matching and probabilistic modeling are used for media source attribution.

Click-through attribution

Most installations are the result of user clicks ads such as banners, videos and interstitials.

After clicking the ad, the click lookback window opens (7 days by default). Installations outside the window period are non-organic and attributed to media source.

Attribution type

Attribution method

Range

Default

Click-Through (All Integrated Partners)

  • Referrer

  • ID Matching

1 – 30 Days

7 Days

Click-Through (All Integrated Partners)

  • Probabilistic modeling

0-24 Days

  • Adaptive

  • Determined by Affise

View-through attribution

Users who view mobile ads but do not click can be attributed to the advertising network where the ad is displayed.

Lookback window for end-to-end attribution:

  • Less than the attribution by click

  • Personal customization is possible

To enable view-through attribution, it's needed to set the look back window in the configuration window. This is especially useful for video advertising networks that traditionally have low CTR in their advertising materials, but it will also be useful for traditional advertising networks that show regular ads.

Attribution type

Attribution method

Range

Default

View-Through

  • ID Matching

1 Hour - 2 Days

1 Day

View-Through

  • Probabilistic modeling

0-24 Hours

  • Adaptive

  • Determined by Affise

⚠️ In cases where both сlick and view occur, click always has priority since it is an active interaction.

Advanced attribution terms

Assisted Install

Affise MMP fully defines only one media source for each installation, usually using the last click or the last view (if there were no clicks).

Auxiliary installations (also called multi-interaction attribution) are installations in which the media source/campaign was not the last point of contact, but interacted with the user before installation, and the interaction took place in the Affise MMP l lookback window. Auxiliary networks are shown as installation participants in Affise MMP.

Reinstall

Reinstall occurs when a user installs an application, deletes it, and then installs it again. The reinstall attribution is regulated by the re-attribution window as follows:

  • If the reinstallation occurs after the expiration of the re-attribution window: a new installation is recorded.

  • If the reinstallation occurs during the re-attribution window, one of the following actions is applied:

    • If the user participated in a retargeting campaign before reinstalling: re-targeting (re-attribution) is recorded.

    • If the user did not participate in the campaign or did not participate in the UA campaign: the installation is not registered. The events in the application of these users who have completed the re-installation are organic.

To test the device and install it multiple times, register the device in the application settings in the application card. If you don't register the device, only the first installation will be recorded.

Reinstalling iOS apps backed up in iCloud

When an application is backed up using iCloud and then restored (on the same or other device), Affise MMP does not consider this a new installation or reinstallation. A user restoring an application from iCloud saves their Affise MMP ID and attribution data.

Retargeting attribution

A user who re-installs the application within the re-attribution window (90 days by default) is considered a re-attribution. If this installation occurs after participating in a retargeting campaign, it is registered as a retargeting reinstall or re-attribution, and is available in the retargeting report.

App updates

When existing users update the application to the next version, it is not considered as an attribution event if the user was previously attributed to Affise MMP.

⚠️ When you switch to Affise MMP from other MMPs, existing users are classified as organic when the application is opened after migration for the first time.

Metrics related to the number of users per application version are available in the Affise MMP SDK reports.


Please contact the Affise Customer Support team regarding all raised questions via the e-mail: [email protected].

Did this answer your question?