SKAdNetwork Attribution

For each app that you are marketing, you can choose one of the following conversion model types:

– Revenue: Lets you optimize your campaigns based on revenue gained during the measurement period following the install/reinstall. A revenue model can measure three types of revenue, depending on how you set it up: in-app purchase revenue, ad revenue, or both (all revenue).
See also: SKAN Optimized Models FAQ

– Conversion Events: Lets you optimize campaigns based on specific post-install user activity. The model encodes user events into the conversion value if they occur at least once during the measurement period.

– Engagement: Lets you optimize your campaigns by how much the users engaged with the app during the measurement period. The model encodes into the conversion value how many times various events occurred during the measurement period.

– Mixed Models (NEW): Mixed models allow you to get both revenue information and one other type of information about the same SKAdNetwork campaign. Singular offers three mixed models: Conversion Event.

For SKAdNetwork campaigns with partial data (due to censored conversion values), Singular can show modeled (extrapolated) metrics for the entire campaign.

Modeled metrics are derived from the installs for which we do have data.

For example:
– A SKAN campaign has 50 installs, out of which 25 came with a conversion value (the Conversion Value Ratio is 50%).
– Based on the existing conversion values, the SKAN revenue is $30.
– The modeled revenue for the campaign, based on the assumption that the installs with censored conversion values behave like the installs with available conversion values, is $60.

SKAdNetwork revenue models are only as effective as the revenue buckets defined for them (see How does a revenue model work?).

Optimized models solve the dilemma of how to define the revenue buckets (how many buckets, how big/small, etc.) by generating the best revenue buckets for your app automatically.

Optimized models are calculated based on real user-level revenue event data for the specific app (as measured by the Singular tracker).

Singular has developed estimation and statistical algorithms that improve the SKAdNetwork dataset by leveraging additional datasets we are collecting as the MMP. We use ad spend, IDFA, and in-app event data to estimate cohort metrics, for example d7 and d30 revenue.

The estimation benefits from the existing SKAdNetwork conversion model used by your app — better models creating more accurate data sets will yield better accuracy in estimating cohort metrics.

Against each metric, Singular will provide the confidence interval to help marketers evaluate the accuracy and in order to make informed decisions around campaign optimization.

Here’s an example:

– Let’s take a marketer who is using a measurement period of 24 hours for their conversion model.

– At the end of the first 24-hour window, which starts when SKAdNetwork is first called, every iOS device will get an additional random timer, and after which Singular will collect all captured conversion values, encoded via SKAdNetwork and sent to Singular by the ad networks.

– Singular will then decode the conversion values back to the original metric, for example Revenue, which at this stage is not cohorted.

– Now, comes the data science – Singular will then take the original Revenue metric and using technology will estimate the d7 Revenue. Each value will be provided with a confidence interval to allow the marketer to get a sense of accuracy.

– For example, when $100 is estimated with a confidence interval of $10, this means we estimated the d7 Revenue at $100 with 90% accuracy!