Apportioning Credit Across Impressions and Platforms

Description: InnovidXP configures our models to apportion credit across multiple impressions (where a household has been exposed multiple times) and across platforms (where a household has been exposed to both linear and CTV impressions). This article gives an overview of how we do this as part of our approach to measuring outcomes.

Summary of approach

To apportion credit across multiple impressions, our model can credit each impression based on the following:

  • Recency against an algorithmically defined decay curve
  • Weighting based on statistical analysis against the appropriate weighting criteria (e.g., overall performance of the relevant linear spot for linear impressions)



The diagram below illustrates this approach in more detail:


How it works

1. Decay curve: We track responses to impressions over 7 days and calculate a decay curve.

2. Position on the curve: Each impression is plotted on the curve based on its timestamp relative to the target business event (e.g., registration or sale).

3. Weighting: Each impression is weighted algorithmically against the appropriately selected weighting criteria. For example, for linear impressions, each impression would be weighted algorithmically based on the overall performance of the spot that the impression relates to.

When a household has been exposed to both linear and CTV, we take that into account and use data science to appropriately assign values.

Related content

Viewing Attribution Results After Reloading Data

Approach to Linear TV Attribution

Extrapolation Overview

Was this article helpful?
0 out of 0 found this helpful