Modelling what
drives engagement.
AI-led analysis of creative features and engagement across P&G Baby Care and competitor influencer content.
My role
At Stagwell Marketing Cloud, I translated AI-led content analysis into recommendations for influencer selection and creative briefs. My role focused on interpreting model outputs, assessing the strength of the evidence and connecting performance patterns to specific creative decisions.
The analysis
The study evaluated 581 Instagram posts from P&G and 11 competitor brands against more than 100 features, including creator demographics, editing pace, messaging and child presence, with engagement rate as the target metric.
I interpreted feature-importance rankings and partial dependence plots to examine how individual features related to predicted engagement. Post counts and feature coverage informed confidence in each finding; comparison with competitor content helped test whether patterns extended beyond P&G.
The finding
Baby or child presence showed no meaningful observed relationship with engagement rate across P&G and competitor content, with medium confidence. This supported testing less child-centric creative, focusing instead on parents’ experiences, advice and routines.
An observational model finding, rather than evidence of causation.
