Question
When is it appropriate to use more than 1 predictive model in a campaign?
Answer
- Distinct, non-overlapping goals: If audiences have clearly separate objectives and channels, they may legitimately use different models.
- Experimental work: For exploratory tests you may temporarily combine models to discover signals, but treat this as a test and avoid using it for definitive campaigns.
- Very small subsegments: If the target population is tiny, you might need supplemental models to reach statistical power — but document this decision and measure carefully.
