[Beta] Predict reports

For:
Marketers
Last update:

beta_feature.png

At a glance: Get Predict reports using either the Predict Pull API and/or Data Locker. Reports are available based on SKAdNetwork (SKAN) attribution or AppsFlyer (AF model) attribution.

 Related reading

For additional information about the methods for accessing Predict reports, be sure to read these articles:

Overview

Predict reports include install attributions, cost data, and predicted metrics for groups of newly acquired users (or “cohorts”). Different report types are available based on:

  • The attribution method selected for the report:
    • SKAdNetwork (SKAN) attribution; or
    • AppsFlyer (AF model) attribution
  • The level of data reported:
    • Aggregated reports: Data is aggregated using the postback arrival date (for SKAN attribution) or install date (for AF model attribution).
    • Raw data reports: Data is reported at the user level.

Accessing Predict reports

Predict reports are generated as CSV files and can be accessed via the Predict Pull API and/or AppsFlyer Data Locker.

  • With the Predict Pull API, you can generate reports on demand using API calls.
  • With Data Locker, your report data is written directly and automatically to cloud storage, enabling direct loading into your internal systems (such as BI or CRM systems).

The methods by which reports can be accessed vary by report type, as follows:

Report Aggregated Raw data
SKAN installs 

Data Locker

SKAN redownloads
  • Predict Pull API
  • Data Locker
 Data Locker

AF model installs

Data Locker

Additional details and instructions for accessing reports via these methods are included in the following tabs:

Predict Pull API

Your account admin can obtain the V2.0 token (bearer token) required for API access from the AppsFlyer dashboard.

Learn more about the Predict Pull API and how to retrieve the report.

Data Locker

To enable access to Predict reports via Data Locker:

  1. If you don't yet receive reports via Data Locker (for any AppsFlyer feature), complete first-time Data Locker setup. (If you do currently receive Data Locker reports, skip to step 2.)
  2. Add Predict reports to Data Locker:
    1. In AppsFlyer, go to Integration > Data Locker.
    2. Select all the reports in the Predict section.
    3. Click Save configuration.
      • Reports will be available the next day by 10:00 UTC. 

Predict report structure

The following tabs list the data fields included in Predict reports: 

Aggregated reports

Column name Dimension/metric Format
Install Date Dimension String
App ID Dimension String
Ad ID

Dimension

String
Ad Name

Dimension

String
Adset ID

Dimension

String
Adset Name

Dimension

String
Agency

Dimension

String
Campaign ID

Dimension

String
Campaign (c)

Dimension

String
Channel

Dimension

String
Platform

Dimension

String
Geo

Dimension

String
Media Source (pid)

Dimension

String
Region Dimension String
Site ID Dimension String
Total Users Metric Integer
Users without a conversion value Metric Integer
Users without sufficient data Metric Integer
Predicted Revenue Metric Number
pARPU Metric Number
Predicted Paying Users Rate Metric Number
Predicted Whale Rate Metric Number
eCPI Metric Number
pROAS Metric Number
1 AppsFlyer Attribution Found Dimension Boolean
1 SKAN Redownload Dimension Boolean
1 SKAN Source App ID Dimension Boolean
2 p-retention day {x} Metric Number
2 p-retention day {y} Metric Number
2 p-retention day {z} Metric Number

Notes:

1 Column appears only in reports using the SKAN attribution method
2 {x},{y},{z} represent the 3 retention measurement dates selected during the onboarding process

Raw data reports – SKAN attribution

Field name 1 Data source Description Format In postback copy? 

ad_network_campaign_id

 Ad network Campaign ID reported by the ad network to AppsFlyer String No

ad_network_campaign_name

 Ad network

Campaign name reported by the ad network to AppsFlyer

 String No

ad_network_timestamp

 Ad network

Time the ad network received the postback.

  • In the case of Google Ads, the time is derived from the last-click time reported by Google.
 Timestamp No

app_id

 AppsFlyer App ID (advertiser app); includes the prefix id String Yes

city

 AppsFlyer Starting August 30, 2021, always null  String No

country_code

 Ad network   String No
dma   Starting August 30, 2021, always null  String No
install_date iOS Estimated by AppsFlyer based on the updated conversion window Time string No
install_type iOS

New install/redownload (corresponds to skad_redownload)

 String No
ip  

IP address in the HTTP message of the iOS device sending the postback.

  • Note: This field is populated if the ad network provides AppsFlyer with the address. The AppsFlyer SDK is not involved in obtaining the IP address.
 String No
media_source iOS Ad network attributed using AppsFlyer network names String No

postal_code

   Starting August 30, 2021, always null   String No

region

 AppsFlyer According to the country_code reported String No

skad_ad_network_id

 SKAN 1.0+
  • ID of the ad network to credit. Use did_win to identify if the ad network is the winner or the contributor. 
  • Some SRNs don't share their IDs (for example, Twitter). When this is the case, the field is left blank.
 String

Yes 

skad_attribution_signature

 SKAN 1.0+   String Yes
skad_campaign_id SKAN 1.0+

Campaign ID used to publish the ad

  • Range: 0-100
 String Yes
skad_conversion_value SKAN 2.0+

Conversion value reported by iOS

  • Range: 0-63

2 SKAN privacy threshold determines whether field is populated

 Integer Yes
skad_mode Predict

Conversion measurement mode used to decode the iOS postback

 String No
skad_redownload SKAN 2.0+ If TRUE, the app was redownloaded

Boolean

Yes
skad_source_app_id SKAN 2.0+

Site ID (the app publishing the ad)

  • This field is identical to source_app_id

2 SKAN privacy threshold determines whether this field is populated

 String Yes
skad_transaction_id SKAN 1.0+   String Yes
skad_version SKAN 2.0+   String Yes
timestamp Ad network

Time the iOS device sent the postback.

In the case of Google ads, this time is derived using the last-click time reported by Google Ads. 

 Time string Yes

source_app_id

 AppsFlyer

Site ID (the app publishing the ad)

  • This field is identical to skad_source_app_id
 String No

ad_network_adset_id

 Ad network   String No

ad_network_adset_name

 Ad network   String No

ad_network_ad_name

 Ad network   String No

ad_network_ad_id

 Ad network   String No

skad_did_win

 SKAN 3.0+ (iOS 14.6+)

Used to distinguish between campaigns contributing to or attributed with the install:

  • If TRUE, the campaign is attributed for bringing the install.
  • If FALSE, the campaign is recognized as contributing to the install
 Boolean Yes

fidelity_type

 SKAN 2.2+ (iOS 14.5+)

Differentiates between ad presentation options that affect attribution:

  • 0 = view-though ads
  • 1 = StoreKit-rendered ads

Disregard the field value if SKAN version is prior to 2.2.

 Integer Yes

ad_network_source_app_id

 Ad network Publisher app ID reported by the ad network. Used to help reconcile cost; reported by the ad network. The publisher app gives the basis for this data.  String No

ad_network_channel

 Ad network   String No

af_prt

 Ad network   String No

measurement_window

 AppsFlyer

Duration of the activity window

  • Default: 24 (hours)
 Integer No

af_attribution_found

 AppsFlyer Currently null for Predict Boolean No

p_revenue

 Predict Mean predicted revenue (Day 30) of the revenue bucket to which the user is assigned Number No

is_p_payer

 Predict
  • 0 = user will NOT be a paying user as of day 30
  • 1 = user will be a paying user as of day 30
 Integer No

is_p_whale

 Predict Integer No

p_retention

 Predict The last of the 3 selected retention measurement dates on which the user will be a retained user Integer No

no_conversion_value

 Predict
  • 0 = postback reported a conversion value
  • 1 = no prediction because postback did not report a conversion value (due to SKAN privacy thresholds)
 Integer No

insufficient_data

 Predict
  • 0 = sufficient data for prediction
  • 1 = no prediction as a result of insufficient data
 Integer No

Notes:
1 When SKAN is the data source, the first SKAN version that reports this data is specified.
2 SKAN determines whether the field is populated using a privacy-threshold mechanism.

Raw data reports – AF model attribution

Field name Description Format

af_ad_id

Ad ID

String

platform

Device platform (iOS)

Enum string

af_adset_id

ID of the adset (a group containing one or more ads)

String

af_adset

Name of the adset

String

media_source

Ad network attributed (using AppsFlyer ad network names)

String

af_ad

Ad name

String

region

Geographic region, based on the device IP address reported by the SDK

String

campaign

Campaign name

String

af_channel

Media source channel 

  • For example, YouTube is a Google channel; Instagram is a Meta channel, etc.

Dynamic enum string

campaign_id

Campaign ID

String

Country

Country code using ISO 3166 (alpha-2), based on the device IP address reported by the SDK – for example: US, CN

  • The United Kingdom is shown as UK (not GB). 

Enum string

site_id

Publisher ID

String

attribution_type

The type of attribution:

  • Install
  • Reinstall
  • Re-engagement 
  • Unknown

Currently, Predict supports only Install attributions.

 String

af_prt

Agency

String

install_time
  • UA: First launch after install
  • Retargeting: First launch after re-engagement/re-attribution

Currently, Predict supports only UA, so install_time will always be the first launch after install.

Date Time

app_id

App ID assigned by the app store

  • For iOS, this always includes the prefix id.

String

ip

IP address in the HTTP message of the iOS device sending the postback, as reported by the SDK

String

attributed_touch_type

Possible values:

  • UA: click, impression, TV, pre-installed
  • Retargeting: click, impression

Currently, Predict supports only UA, so attributed_touch_type will be one of the possible values for UA.

Enum

city

The most granular location of the user, based on the device IP address reported by the SDK

  • Usually, the value is a city name, but it can also be districts, boroughs, etc., which are more precise.

String

dma

Designated Market Area or geographic area, based on the device IP address reported by the SDK

  • These values are defined by the Nielsen company.

String

postal_code

 Postal code, based on the device IP address reported by the SDK

String

state

State, based on the device IP address reported by the SDK

String

user_agent

URL user agent

String

is_lat

Limit ad tracking (LAT):

  • For users on iOS versions prior to iOS 14, TRUE indicates that the user has elected to limit ad tracking (LAT).
  • For users on iOS 14 or later, TRUE indicates that the user has elected to disallow tracking based on ATT status.
 Boolean

app_name

App name as set by the advertiser

String

source

Organic or non-organic

 Enum string

p_retention

The last day (during the first 30 days post-install) on which the user will use the app

Integer

p_purchase_revenue

 Predicted revenue as of Day 30 from user purchases/in-app events String

p_ad_revenue

 Predicted revenue as of Day 30 from served ad impressions String

bucket_revenue_0

Integer

bucket_revenue_1

Integer

bucket_revenue_2

Integer

bucket_revenue_3

Integer

bucket_revenue_4

Integer

bucket_revenue_5

Integer

bucket_revenue_6

Integer

bucket_retention_0

Integer

bucket_retention_1

Integer

bucket_retention_2

Integer

bucket_retention_3

Integer

is_p_whale

Integer

is_p_payer
  • 0 = user will NOT be a paying user as of Day 30
  • 1 = user WILL be a paying user as of Day 30

Integer

insufficient_data
  • 0 = sufficient data for prediction
  • 1 = no prediction as a result of insufficient data

Integer

install_date

Date of the first launch after install based on UTC time zone

  • Reported in MM/DD/YYYY format

Date string

install_date_tz

Date of first launch after install, adjusted based on app-specific time zone

  • Reported in MM/DD/YYYY format
 Date string

Feature specifications

  • Ad network access: Ad networks are provided with postbacks
  • Agency access: Not available
  • Agency transparency: N/A
  • Currency: USD
  • Timezone: UTC
  • Data freshness: Daily
    • For reports based on the SKAN attribution method, SKAN timing means that iOS reports on an install 48-72 hours after the first app is opened.
  • Historical data: Data is available from the date Predict was enabled in your AppsFlyer account.
  • Source (organic/non-organic): Reports based on the SKAN attribution method include only non-organic data, calculated using SKAN postbacks received by AppsFlyer.