Testing Predictive Scores
Predictive Scores help you identify high intent vs low intent users. The next step is to create a targeted segment to increase your campaign performance. This can be done by either filtering your existing campaign audience to users most likely to respond on a campaign or testing hypothesis on high intent or low intent audiences to increase their conversion rates.
Improving existing campaign performance
If existing campaigns are targeted on a large user base, Predictive Scores can be used to filter down your audience to increase the performance of your campaign by targeting a subset of users. This could reduce your ad spend without impacting the overall yield, reduce your unsubscribe rate or build a better reputation with the ESPs.
To measure the conversion attribution, create a control group based on bsft_control_bucket and test it against different cohorts of scores using vertical branching feature in campaigns.
Control bucket - use user attribute bsft_control_bucket (example 0-25 to use 25% traffic for control bucket)
You can use the performance tab to view your scores distribution(i.e number of users whose score is between X and Y) and come up with different thresholds for cohorts of Predictive Scores to identify.
- Low intent users (e.g 0-33)
- Mid intent users (e.g. 34-66)
- High intent users (e.g. 67-100)
The overall setup will look as follows:
Once the campaign is launched the metrics for each vertical branch can be compared to see the efficacy of different deciles vs control population.