Answers are grounded in this paper and cite their sources.
Retail media networks need ad decisions that react to temporally shifting shopping behavior rather than only to static transaction summaries. This paper tries to infer purchase intention from ever-changing clickstreams as an operational identification problem for a deployed advertising system. The proposed method fits time-dependent data to a covariate-dependent hidden semi-Markov model whose latent states are anchored by the Theory of Planned Behavior, the Consumer Decision Journey, and the Persuasion Knowledge Model. A retrieval-augmented language model supplies a theory-coherent reference posterior γ to stabilize the categorization. Semi-Markov inference uses this reference as a Kullback-Leibler regularizer in the expectation step while mapping retrieved context into the initial, transition, emission, and duration kernels. Closed-loop directional validation and revalidation are conducted through a 3-campaign experiment on a live platform. For illustrative purposes, the experiment, conducted with an anonymized omni-channel retailer covering 5.2 million active members, confirms an AUC of 0.717 for our theory-identified model, compared to 0.635 for an action-only covariate baseline on n=1008 paired test consumers. The results show that segment-level conversion lift, opt-out reduction, and 60-day retention lift can be retained in future deployments under a drifting environment.
No takes yet. Share an insight, caveat, or question.
Ou et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: