App Events Data Overview
Daily aggregated app event data is necessary for indicating app performance so a model can be trained to quantify the contributions of each marketing source and control variable. Data must always be cohorted by install date.
App event data can be aggregated for an app at the country level only or at a deeper level of granularity by including dimensions and metric breakdowns provided by last touch attribution. Including last touch attribution dimensions enables comparison between last touch and incrementality metrics in the Polaris dashboard and API. Supported dimensions are:
Install Date
Channel (last touch)
Campaign (last touch)
Country
Source App (last touch)
The only mandatory dimensions are Install Date and Country.
App event metrics must be cohorted by install date. The supported standard simple metrics are listed below. From those simple metrics, all other standard compound metrics can be derived.
Metrics with an aggregation of On Day are realized on the cohort day only. For example, Sessions D7 is the number of sessions realized on D7 only, not the cumulative sum of sessions through D7. Conversely, metrics with an aggregation of Cumulative (e.g., Revenue Amount) are the cumulative sum through the cohort day.
Incomplete cohorts are cohorts that do not have enough days since the install date to evaluate a specific cohorted metric. For example, if you are looking at Revenue Amount D1, only cohorts with install dates at least one day ago can possibly be complete. Incomplete cohorts are always automatically disregarded by Polaris. For example, if you are looking at Revenue Amount D1 and select yesterday and today as the date range, only the cohort from yesterday would be included in the value.
No metrics are mandatory, but you must at least provide one since otherwise there will be nothing to model.
The table below contains a sampling of the compound metrics available in Polaris. This list isn't comprehensive, but is meant to provide an overview of the most popular metrics.
Depending on the dimensions selected, ROAS may be interpreted as blended ROAS. Blended ROAS is all-in revenue (organic and marketing-driven) divided by spend. For example, if you are looking at ROAS at the country level, it will be blended ROAS. The same applies to the app level. Conversely, if you are looking ROAS at the channel level, it will only include revenue allocated to each channel.
The following cohort days are supported for each cohorted metric:
D0 (all except for Retained Users)
D1 through D7
D14
D30
D60
D90
D120
D180
D360
D0 is the install date, D1 is 1 day after the install date, and so forth.
Dimensions
App event data can be aggregated for an app at the country level only or at a deeper level of granularity by including dimensions and metric breakdowns provided by last touch attribution. Including last touch attribution dimensions enables comparison between last touch and incrementality metrics in the Polaris dashboard and API. Supported dimensions are:
Install Date
Channel (last touch)
Campaign (last touch)
Country
Source App (last touch)
The only mandatory dimensions are Install Date and Country.
Metrics
App event metrics must be cohorted by install date. The supported standard simple metrics are listed below. From those simple metrics, all other standard compound metrics can be derived.
Metrics with an aggregation of On Day are realized on the cohort day only. For example, Sessions D7 is the number of sessions realized on D7 only, not the cumulative sum of sessions through D7. Conversely, metrics with an aggregation of Cumulative (e.g., Revenue Amount) are the cumulative sum through the cohort day.
Incomplete cohorts are cohorts that do not have enough days since the install date to evaluate a specific cohorted metric. For example, if you are looking at Revenue Amount D1, only cohorts with install dates at least one day ago can possibly be complete. Incomplete cohorts are always automatically disregarded by Polaris. For example, if you are looking at Revenue Amount D1 and select yesterday and today as the date range, only the cohort from yesterday would be included in the value.
| Metric | Description | Cohorted | Aggregation |
|---|---|---|---|
| Installs | Number of installs | No | N/A |
| Sessions | Number of sessions on the cohort day | Yes | On Day |
| Retained Users | Number of users retained on the cohort day | Yes | On Day |
| Revenue Amount | Cumulative amount of total revenue through the cohort day | Yes | Cumulative |
| Ad Revenue Amount | Cumulative amount of ad revenue through the cohort day | Yes | Cumulative |
| Purchase Revenue Amount | Cumulative amount of purchase revenue through the cohort day | Yes | Cumulative |
| Purchases | Number of revenue events (purchases, ad views, etc.) on the cohort day | Yes | On Day |
| Unique Payers | Number of users with at least one revenue event on the cohort day | Yes | On Day |
No metrics are mandatory, but you must at least provide one since otherwise there will be nothing to model.
Compound Metrics
The table below contains a sampling of the compound metrics available in Polaris. This list isn't comprehensive, but is meant to provide an overview of the most popular metrics.
| Metric | Description | Formula | Cohorted | Aggregation |
|---|---|---|---|---|
| CPI | Cost per install | Spend divided by Installs | No | N/A |
| Retention | Retention rate | Retained Users divided by Installs | Yes | On Day |
| LTV | User lifetime value | Revenue Amount divided by Installs | Yes | On Day |
| ROAS | Cumulative return on ad spend | Revenue Amount divided by Spend | Yes | Cumulative |
| Paying Rate | Rate of paying users | Unique payers divided by Installs | Yes | On Day |
Depending on the dimensions selected, ROAS may be interpreted as blended ROAS. Blended ROAS is all-in revenue (organic and marketing-driven) divided by spend. For example, if you are looking at ROAS at the country level, it will be blended ROAS. The same applies to the app level. Conversely, if you are looking ROAS at the channel level, it will only include revenue allocated to each channel.
The following cohort days are supported for each cohorted metric:
D0 (all except for Retained Users)
D1 through D7
D14
D30
D60
D90
D120
D180
D360
D0 is the install date, D1 is 1 day after the install date, and so forth.
Updated on: 08/03/2023
Thank you!